e ective development of haptic devices using a model-based and...

Effective development of haptic devices using amodel-based and simulation-driven design approach

AFTAB AHMAD

Doctoral Thesis in Machine DesignStockholm, Sweden, 2014

TRITA - MMK 2014:02ISSN 1400-1179ISRN/KTH/MMK/R-14/02-SEISBN 978-91-7595-063-1

KTH School of IndustrialEngineering and Management

100 44 StockholmSweden

Academic thesis, which with the approval of the Royal Institute of Technology,will be presented for public review in fulfillment of the requirements for aDoctorate of Technology in Machine Design. The public review is held inRoom F3, Lindstedtsvagen 26, Royal Institute of Technology, Stockholm on2014-04-23 at 10:00.

c© Aftab Ahmad, April 2014

Print: US-AB

Department of Machine Design Royal Institute of Technology S-100 44 Stockholm SWEDEN

TRITA - MMK 2014:02 ISSN 1400 -1179 ISRN/KTH/MMK/R-14/02-SE ISBN 978-91-7595-063-1

Document type Thesis

Date 2014-04-23

Author Aftab Ahmad ([email protected])

Supervisor(s) Kjell Andersson, Ulf Sellgren Sponsor(s)

Higher Education Commission HEC, Pakistan

Title Effective development of haptic devices using a

model-based and simulation-driven design approach Abstract Virtual reality (VR)-based surgical simulators using haptic devices can increase the

effectiveness of surgical training for surgeons when performing surgical procedures in hard tissues such as bones or teeth milling. The realism of virtual surgery through a surgical simulator depends largely on the precision and reliability of the haptic device, which reflects the interaction with the virtual model. The quality of perceptiveness (sensation, force/torque) depends on the design of the haptic device, which presents a complex design space due to its multi-criteria and conflicting character of functional and performance requirements. These requirements include high stiffness, large workspace, high manipulability, small inertia, low friction, high transparency, and cost constraints.

This thesis proposes a design methodology to improve the realism of force/torque feedback from the VR-based surgical simulator while fulfilling end-user requirements.

The main contributions of this thesis are: 1. The development of a model-based and simulation-driven design methodology,

where one starts from an abstract, top-level model that is extended via stepwise refinements and design space exploration into a detailed and integrated systems model that can be physically realized.

2. A methodology for creating an analytical and compact model of the quasi-static stiffness of a haptic device, which considers the stiffness of actuation systems, flexible links and passive joints.

3. A robust design optimization approach to find the optimal numerical values for a set of design parameters to maximize the kinematic, dynamic and kinetostatic performances of a 6-degrees of freedom (DOF) haptic device, while minimizing its sensitivity to variations in manufacturing tolerances and cost, and also satisfying the allowed variations in the performance indices.

4. A cost-effective approach for force/torque feedback control using force/torque estimated through a recursive least squares estimation.

5. A model-based control strategy to increase transparency and fidelity of force/torque feedback from the device by compensating for the natural dynamics of the device, friction in joints, gravity of platform, and elastic deformations.

Keywords Haptic device, model-based design, design optimization

Language English

Acknowledgements

This journey would not have been possible without the support and encour-agement of numerous people including my well-wishers, friends and colleagues.It is a delight to acknowledge those who have contributed in many ways tothe success of this thesis and made it an unforgettable experience for me.

First and foremost, I owe my deepest gratitude to my supervisors KjellAndersson and Ulf Sellgren for their continuous support of my study andresearch, for their patience, motivation, enthusiasm and valuable discussions.Their guidance helped me in research and writing of this thesis.

Beside my advisors, I would also like to thank Lei Feng and Bengt Erikssonfor valuable discussion about control design. My deepest appreciation goes toVicki Derbyshire for reviewing my manuscript with excellent suggestions. Iam particularly thankful to Staffan Qvarnstrom and Tomas Ostberg for theirassistance in prototype development.

My sincere thanks also go to Suleman Khan, Magnus G. Eriksson and XuanSun for sharing their ideas about the project. I am indebted to my many col-leagues at the Department of Machine Design including Xinhai Zhang, Yi Zhu,Saeed Abbasi, Abdurasul Pirnazarov, Matthew Cha, Abbos Ismoilov, PatrickRohlmann, Xinmin Li, and Baha Alhaj Hasan for their support, motivation,and providing a stimulating and fun filled environment.

I would also like to express my gratitude to Higher Education Commission(HEC), Pakistan and Department of Machine Design for funding my studiesat KTH and my stay in Sweden.

Last but not the least; I would like to thank my family back at home fortheir continuous support.

Aftab Ahmad Stockholm, April 2014

v

Contents

Acknowledgements v

Contents vii

List of appended publications xi

List of additional publications xiii

1 Introduction 11.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 Problem description . . . . . . . . . . . . . . . . . . . . . . . . 31.3 Research objective and questions . . . . . . . . . . . . . . . . . 51.4 Research approach . . . . . . . . . . . . . . . . . . . . . . . . . 51.5 Delimitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71.6 Thesis outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71.7 Chapter summary . . . . . . . . . . . . . . . . . . . . . . . . . 7

2 State of the art 92.1 Haptic devices . . . . . . . . . . . . . . . . . . . . . . . . . . . 92.2 Haptic device design . . . . . . . . . . . . . . . . . . . . . . . . 10

2.2.1 Optimization . . . . . . . . . . . . . . . . . . . . . . . . . 112.2.2 System models . . . . . . . . . . . . . . . . . . . . . . . . 122.2.3 Control design . . . . . . . . . . . . . . . . . . . . . . . . 13

2.3 Chapter summary . . . . . . . . . . . . . . . . . . . . . . . . . 14

3 Model-based and simulation-driven design 153.1 Model development in the conceptual design phase . . . . . . . 18

3.1.1 Kinematic model . . . . . . . . . . . . . . . . . . . . . . 183.1.2 Dynamic model . . . . . . . . . . . . . . . . . . . . . . . 193.1.3 Stiffness model . . . . . . . . . . . . . . . . . . . . . . . . 193.1.4 Model verification and validation . . . . . . . . . . . . . . 213.1.5 Robust optimum design . . . . . . . . . . . . . . . . . . . 23

3.2 Control design in the detailed design phase . . . . . . . . . . . 263.2.1 Friction estimation . . . . . . . . . . . . . . . . . . . . . 273.2.2 Force torque estimation . . . . . . . . . . . . . . . . . . . 28

vii

viii CONTENTS

3.2.3 Transparency measurement . . . . . . . . . . . . . . . . . 283.3 Chapter summary . . . . . . . . . . . . . . . . . . . . . . . . . 29

4 Prototypes 314.1 The Ares haptic device . . . . . . . . . . . . . . . . . . . . . . . 31

4.1.1 Model development and analysis based on the Ares hapticdevice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

4.2 The TAU haptic device . . . . . . . . . . . . . . . . . . . . . . 324.2.1 Transmission system . . . . . . . . . . . . . . . . . . . . 334.2.2 Spherical joint design . . . . . . . . . . . . . . . . . . . . 334.2.3 Model development and analysis based on the TAU haptic

device . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344.2.4 Performance evaluation of the TAU haptic device . . . . 344.2.5 Performance characteristics . . . . . . . . . . . . . . . . . 35

4.3 Chapter summary . . . . . . . . . . . . . . . . . . . . . . . . . 37

5 Summary of appended papers 395.1 Paper A: a model-based and simulation-driven methodology for

design of haptic devices . . . . . . . . . . . . . . . . . . . . . . 395.2 Paper B: a stiffness modeling methodology for simulation-driven

design of haptic devices . . . . . . . . . . . . . . . . . . . . . . 405.3 Paper C: an optimization approach towards a robust design of

6-DOF haptic devices . . . . . . . . . . . . . . . . . . . . . . . 405.4 Paper D: evaluation of friction models for haptic devices . . . . 415.5 Paper E: model-based control strategy for 6-DOF haptic devices 42

6 Discussion, conclusions and future work 436.1 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 436.2 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 446.3 Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 456.4 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

References 47

Paper A: a model-based and simulation-driven methodologyfor design of haptic devices 57

Paper B: a stiffness modeling methodology for simulation-drivendesign of haptic devices 73

Paper C: an optimization approach towards a robust design of6-DOF haptic devices 93

Paper D: evaluation of friction models for haptic devices 113

CONTENTS ix

Paper E: model-based control strategy for 6-DOF hapticdevices 125

List of appended publications

• Paper AA. Ahmad, K. Andersson, and U. Sellgren, “A model-based and simulation-driven methodology for design of haptic devices,” Journal of Mechatron-ics, Jan 2014.

Ahmad, Andersson and Sellgren developed the methodology, and Ahmadverified through experimental test.

• Paper BA. Ahmad, K. Andersson, U. Sellgren, and S. Khan, “A stiffness modelingmethodology for simulation-driven design of haptic devices,” Engineeringwith Computers, pp.1-17, 2012.

Ahmad developed the stiffness model, and developed a methodology forstiffness evaluation with Andersson and Sellgren. Khan helped in exper-imental verifications.

• Paper CA. Ahmad, K. Andersson, and U. Sellgren, “An optimization approachtowards a robust design of 6-DOF haptic devices,” Journal of MechanicalDesign. Submitted Jan. 2014

Ahmad developed the robust design optimization approach, Anderssonand Sellgren provided feedback and corrections.

• Paper DA. Ahmad, K. Andersson, U. Sellgren,and M. Boegli,“Evaluation of fric-tion models for haptic devices,” In proceeding of DSCC 2013 ASMEDynamic Systems and Control Conference October 21-23, 2013, Stanforduniversity, Munger Centre, Palo Alto, CA, USA, at Stanford university,Munger Centre, Palo Alto, CA, USA.

Ahmad evaluated the friction models, Andersson, Sellgren and Boegliprovided feedback and corrections.

• Paper EA. Ahmad, K. Andersson, and U. Sellgren, “Model-based control strategyfor 6-DOF haptic devices,” to be submitted to IEEE Transaction onrobotics.

xi

xii LIST OF APPENDED PUBLICATIONS

Ahmad developed and verified the model-based control strategy, Ander-sson and Sellgren provided feedback and corrections.

List of additional publications

• A. Ahmad, K. Andersson, and U. Sellgren, “Design and performanceevaluation of a new 6-DOF haptic device based on TAU”, to be submittedto IEEE Transaction on Haptics.

• A. Ahmad, K. Andersson, and U. Sellgren, “A Deterministic and Proba-bilistic Approach for Robust Optimal Design of a 6-DOF Haptic Device”,In proceeding of: ASME 2013 International Design Engineering Techni-cal Conferences & Computers and Information in Engineering ConferenceIDETC/CIE 2013, Portland, Oregon, USA, 2013.

• A. Ahmad, K. Andersson, and U. Sellgren, “ A Comparitive study of fric-tion estimation and compensation using Extended, Iterated, Hybrid andUnscented Kalman Filter”, In proceeding of: ASME 2013 InternationalDesign Engineering Technical Conferences & Computers and Informationin Engineering Conference IDETC/CIE 2013, Portland, Oregon, USA,2013.

• A. Ahmad, K. Andersson, and U. Sellgren, “A model-based and simulation-driven design approach for haptic device”, In proceeding of: ASME 2013International Design Engineering Technical Conferences & Computersand Information in Engineering Conference IDETC/CIE 2013, Portland,Oregon, USA, 2013.

• A. Ahmad, K. Andersson, and U. Sellgren, “An approach to stiffnessanalysis methodology for haptic devices”, 3rd International Congress onUltra Modern Telecommunications and Control Systems and Workshops(ICUMT), 2011, pp.1-8, 5-7 Oct. 2011.

• S. Khan, A. Ahmad, and K. Andersson, “Design optimization of the Tauhaptic device”, 3rd International Congress on Ultra Modern Telecommu-nications and Control Systems and Workshops (ICUMT), 2011, pp.1-8,5-7 Oct. 2011.

• A. Ahmad, S. Khan, and K. Andersson, “Kinematics and dynamics ofa novel 6-dof tau haptic device,” in IEEE International Conference onMechatronics (ICM), 2011. IEEE, 2011, pp. 719-724.

xiii

Chapter 1

Introduction

Haptics-based virtual reality (VR) simulators for teaching and evaluating sur-gical skills have gained considerable attention in recent years. The virtualreality (VR) based simulators provide a 3D scene of a virtual model and ahaptic feedback at the tool handle of a haptic device. A haptic device acts asa communicating bridge to reflect forces and torques to a user from interac-tions with a virtual environment or teleoperation tasks so that human userscan indirectly feel several physical sensations in real situations. The hapticdevice works on haptic feedback and a haptic collision detection algorithm.The haptic feedback refers to sense and manipulation through touch.

The haptic feedback is active, when the tool center point (TCP) of theend effector of the device is moved and mapped to a position/orientation in avirtual environment. The haptic collision detection algorithm generates force/torque through interaction with a virtual model. The force/torque is generatedby the actuators on the haptic device in real-time, so that appropriate reactionforces are applied to the user, leading to haptic perception of virtual objects[1].

1.1 Background

This thesis is part of a research project on “Haptic milling surgery simulator”for bone milling operations. In the earlier part of the research a master-slave for tele-robotic surgery was developed by Flemmer [2], followed by adevelopment of a 6-degrees of freedom (DOF) haptics-based virtual reality(VR) simulator based on Stewart-platform by Khan [3] and Eriksson [4]. Thework presented here is an extension of this work, in terms of improving therealism of force/torque feedback of an existing haptic device (Ares hapticdevice [5]) and development of a new 6-DOF haptic simulator based on TAUconfiguration.

An evaluation of Ares haptic device using face validity tests were performedby [3] and [4], where it was observed that during manipulation in free space,the motion did not feel free. Many of the participants in the face validity test

1

2 CHAPTER 1. INTRODUCTION

had that opinion about the perception of force/torque feedback while movingin free space and during interacting with the virtual environment. The qualityof perception of haptic devices is mostly measured by transparency, whichquantifies the correspondence between the desired Fg and actually observedvalues of force/torque feedback FT as shown in Figure 1.1.

Force & torque model

Visual display

Haptic device

Visual model

+collision

User Human

operator

Haptic

device

Haptic

rendering( )TF t ( )gF k

( )mx k( )px t

( )px t

( )TF t

( )mx k

( )gF k

Figure 1.1: A closed loop between user and virtual rendering environment.

For transparent force/torque feedback, the motion in free space should feelfree; while the interaction with a virtual object should result in a feedbackof force/torque as close as possible to real physical contact. The quality ofperception of free space and force/torque feedback was affected by the frictionin the active and passive joints, and inertia of the device.

Apart from the developed simulators in the aforementioned research project,several prototypes of haptic devices as surgical simulators have been developedcommercially and in academia for soft and hard tissues over the past coupleof decades, but very few have offered a convincing solution for hard tissues.One reason is because of the use of parallel kinematic structures to obtain thewanted high stiffness for interacting with hard tissues. This results in a com-plex design solution and a need for multi-domain considerations to obtain arealistic force/torque feedback from the device. This thesis proposes a model-based and simulation-driven design approach to deal with this complexity.

Designing a haptic device is a non-trivial task due to its sophisticatedperformance requirements from users, and the integration of software andhardware. The area of haptic research is an interdisciplinary field and isgenerally divided into three main fields: computer haptics, machine hapticsand human haptics [1].

• Human haptics: the study of how people sense and manipulate the worldthrough touch.

1.2. PROBLEM DESCRIPTION 3

• Machine haptics: the design, construction, and use of machines to replaceor augment human touch. Haptic interfaces are devices composed ofmechanical components in physical contact with the human body for thepurpose of exchanging information with the human nervous system.

• Computer haptics: algorithms and software associated with generatingand rendering the touch and feel of virtual objects analogous to computergraphics.

1.2 Problem description

The training of a surgeon is a complex and multi-dimensional process [6] par-ticularly in the case of hard tissues like bone and dental procedures. Surgicalstudents are trained to develop skills by performing surgery using a hand-heldmill on real patients, cadavers and plastic models, which are questionable fromand ethical, training and cost-effectiveness point of view [4], [7]. The plasticmodels are expensive to produce and cannot provide a sufficient level of detailand material properties. Also, the repetitive usage of these artificial models isnot possible, which is ultimately not cost-effective. The high risk of trainingon real patients and the high cost have motivated research and developmentof haptic devices and virtual reality simulators for training surgeons. In con-trast to the classical training techniques, training with medical simulators hasproved to increase patient safety and reduce risk associated with human errorsin hospitals by allowing medical students to develop skills more efficiently andin a shorter time period [7].

Simulators will create new training opportunities for surgical procedures,which are impossible to train with traditional methods. Moving training forsurgical procedures from the operating room to a simulator will offer consid-erable economic advantages. Besides, with the help of virtual simulation envi-ronments, we can better assess on the trainee’s performance. It is possible todefine performance metrics for a specific surgery in simulations. For instance,during a bone removal operation, simulations are able to figure out the regionsthat are removed when they are not visible by the user. Furthermore , it ispromising to define maximum velocities and forces for some critical anatomicregions and observe whether the user exceeds these thresholds or not. Withthese metrics, it is also possible to obtain a visual feedback, which can showthe bone regions in different colors according to the performance of the userfor the tasks on that region.

Although complex and realistic medical simulators are currently being de-veloped by a large number of universities and medical companies, the fieldof hard tissues interaction like orthopedics and dental surgery has not beenexploited yet [7].

Considering the application context of a haptic device in orthopedics anddental surgery, the surgeon needs to perform various tasks like cutting, milling


and drilling, for example, as shown in Figure 1.2.

Figure 1.2: Typical tasks involved in a surgery of hard tissues, [8],[9].

In order to create a haptically enabled virtual reality simulator to trainsurgeons for these skills, the main user requirements are given in [10], whichare as follows:

• Haptic feedback in 6-DOF to allow both force and torque feedback, as wellas translational and rotational capabilities of the virtual tool operatingin a (narrow) channel or cavity.

• The whole device should fit in a space of 250 × 250 × 300 mm.

• The minimum translational and rotational workspace should be 50 × 50× 50 mm and ±40o, respectively, in all directions at the nominal position,with no singularities in the workspace.

• The tool center point (TCP) should be able to render high force andtorque up to at least 50 N and 1 Nm, respectively.

• Low back-drive inertia and friction, and no constraints on motion imposedby the device kinematics, so that free motion feels free.

• Ergonomics and comfort for the user.

• Transparency and stability of the complete system.

Some of the device requirements expressing common desirable character-istics for force/torque feedback of haptic devices include, according to Gogu[11]:

• Isotropic behavior.

1.3. RESEARCH OBJECTIVE AND QUESTIONS 5

• Large workspace.

• Low effective mass.

• Symmetric inertia, friction, stiffness, and resonate frequency propertiesthereby regularizing the device so users do not have to compensate forthese forces.

• Robustness.

• Transparency of the force/torque feedback from the virtual environment.

• Balanced range, resolution, and bandwidth of position sensing and forcereflection.

• Proper ergonomics that eliminate pain and discomfort when manipulat-ing the haptic interface.

1.3 Research objective and questions

The main research objective of this thesis is to develop a 6 DOF haptic devicethat can provide realistic force/torque feedback from the virtual environment.The focus will be on the investigation of a model-based and simulation-drivendesign approach as an effective tool for the design of haptic devices.

This main objective can be divided into the following research questions:

• Q1: How to accurately estimate stiffness of a 6-DOF haptic devicethrough an analytical model, which is compact and can be used in real-time control?

• Q2: How to find a robust optimum design of haptic devices, whose per-formances are less sensitive to variations in manufacturing tolerances?

• Q3: How to estimate friction in haptic devices that can mimic realisticbehavior and be used in real-time control?

• Q4: How to evaluate force/torque capability of a haptic device withouta force/torque sensor?

• Q5: How to develop a control strategy that can be used to compensatefor unwanted effects and maximize the transparency of a haptic device?

1.4 Research approach

The development of haptic devices is a challenging task because of the multi-criteria design consideration. As stated earlier, the main objective is to providea realistic force/torque feedback from the virtual environment via a hapticdevice. This directed the research towards the investigation of how the real


sensation of using the system can be improved when using it in free space(increase transparency) and when in contact (stiffness representation) duringsurgery can be increased, as shown in Figure 1.3.

Constrained motionFree space

Realisticfeedback

Figure 1.3: Research focus.

The following research approach has been taken while conducting this thesiswork.

• Literature review to explore the area of haptics and haptic interfaces andto identify the required performance specifications.

• Elicitation of requirements from the end user and transformation of theserequirements together with the performance requirements to design spec-ifications.

• Development of an approach for robust optimization of haptic devices

– Kinematic models

– Dynamic model

– Stiffness model

• Development of a model-based control strategy for haptic devices

– Compensation for friction

– Compensation for deflections

– Compensation for dynamic effects

• Development of a methodology for design of haptic devices

• Experimental verification and validation

1.5. DELIMITATIONS 7

1.5 Delimitations

This thesis aims to improve transparency and stability of haptic simulatorsfrom the perspective of device performance. Performance of the virtual modeland haptic collision detection algorithm has not been considered in this thesis.

This thesis proposes model-based and simulation-driven design approachto develop haptic devices with a model-based control strategy to further im-prove transparency. In a model-based control strategy, dynamic, stiffness andfriction models have been considered for compensating inertial effects, elasticdeformations and friction in active joints. However, the friction model is notconsidered in the multi-objective optimization, as it has only been consideredin the control strategy.

1.6 Thesis outline

The thesis is divided into six chapters and a brief overview of each chapter isgiven below.

Chapter 1 provides an introduction to the thesis, research background,and discusses the problems that have been identified. Furthermore, researchquestions, research approach, and delimitations of the thesis are given.

Chapter 2 presents state of art in the design of haptic devices. The reviewincludes characteristics of different architectures used for haptic devices.

Chapter 3 presents the model-based and simulation-driven design approachand its verification using both the Ares and TAU 6-DOF haptic devices.

Chapter 4 describes the two haptic devices (Ares and TAU) that have beendeveloped and used as test cases during this thesis work.

Chapter 5 gives a brief summary of the appended papers.Chapter 6 summarizes the findings, contributions, and conclusions of the

thesis work and outlines future work.

1.7 Chapter summary

This chapter summarizes the motives of the thesis and the research frame ofthe virtual reality-based surgical simulators. The research problem is formu-lated in the form of research questions followed by the taken research approach,delimitations and thesis outline.

Chapter 2

State of the art

Significant research efforts have been devised in the development of hapticdevices, and several studies exist on the design and devolvement in commercialand academic literature. This chapter provides an overview of the currentstate of the art in design of haptic devices. The review includes an overviewof different architectures used for haptic device and its design, optimization,system models, and control design.

2.1 Haptic devices

A state of the art about different commercially available haptic devices andtheir application areas have been reviewed in [12]. Furthermore, a comparativestudy has been conducted by Khan [3] in which he compared different hapticdevices based on kinematic structure, DOF, workspace, stiffness, maximumforce and cost. Currently, there are haptic devices available in the forms of 2,3, 4, 5, 6, and 7-DOF, reviewed by for example, Gogu [11] and Lee [9]. Leehas compared different haptic devices based on serial and parallel kinematicstructures.

Serial haptic devices have a kinematic structure, which consists of a singleopen-loop chain where the TCP is connected to the fixed based. The kinematicchain is composed of a group of rigid links where each pair of adjacent linksis interconnected by an active kinematic pair (controlled joint).

Serial devices have some advantages like a large working volume and highdexterity; however, numerous disadvantages such as low precision, poor forceexertion capability, low payload-to-weight ratio, and high inertias tend to limitits use in many applications.

Parallel haptic devices have a kinematic structure, where the TCP is con-nected to a fixed based through several kinematic chains. In contrast withthe open chain serial devices, the parallel devices are composed of closed kine-matic chains, and every kinematic chain includes active and passive kinematicpairs.

9

10 CHAPTER 2. STATE OF THE ART

Parallel devices possess high stiffness, low inertia, high rigidity and accu-racy, and high payload-to-weight ratios, which enables the large bandwidthtransmission of forces but with some disadvantages like a small workspace,the possibilities of singularity due to link collision, and low dexterity.

Recently, many 6-DOF haptic devices have been developed, some of whichhave been commercialized, which include both serial and parallel haptic de-vices. The most widely used haptic devices are PHANTOM R© by Massie andSalisbury [13], Omega.3 by Force Dimension [14], Novint Falcon by NovintTechnologies [15], Haptic master [16], Virtuose by Haption [17], Freedom 7Sby MPB Technologies [18], HD2 by Quanser [19], Xitact HP by Mentice [20]and Meglaev by CAE Healthcare [21]. A performance comparison of thesedevices is given in Table 2.1.

Table 2.1: Commercially available force feedback haptic devices [12]

PhantomOmni

Omega3

NovintFal-con

HapticMas-ter

Virtu-ose

Freedom7S

HD2 IHP Meglaev

Structure Serial Parallel Parallel Serial Serial Serial Parallel Hybrid MagneticWorkspace(10−3m3)

1.3 2.2 1.1 80.0 91 12 70 1.9 0.01

DOF 6//3 3 3 3 6.0 7 6//5 4 6PositionRes (µm)

55.0 10.0 60.0 4.0 6.0 2.0 51.0 57.0 2.0

ContinuousForce (N)

0.9 12.0 9.0 100.0 10.0 0.6 10.8 20.0 n/a

Peakforce (N)

3.3 n/a n/a 250.0 35 2.5 19.7 30.0 40.0

Stiffness(N/mm)

2.0 14.5 n/a 50.0 n/a 2.0 3.0 n/a 50.0

Stiction(N)

0.26 n/a n/a n/a n/a 0.04 0.35 n/a 0.0

ForceReso-lution(N)

n/a n/a n/a 0.01 n/a n/a n/a n/a 0.02

ForceBand-width(Hz)

n/a n/a n/a n/a n/a n/a n/a n/a 2000

2.2 Haptic device design

As stated in Chapter 1, the design of haptic devices is based on user and de-vice requirements, for example: a large workspace for a human operator, lowapparent mass/inertia, low friction, high structural stiffness, backdriveability,

2.2. HAPTIC DEVICE DESIGN 11

low backlash, high force bandwidth, high force dynamic range, absence of me-chanical singularities, compactness, and an even “feel” through the workspace[22]. These requirements reflect considerations of different disciplines such asmechanical, electrical and control design. These considerations benefit fromthese systems being designed in parallel and are integrated to form the com-plete system as a mechatronic product.

Fathy et al. [23] identified four different design approaches for the inte-grated optimization of mechanical and control system design: sequential, iter-ative, nested, and simultaneous. Khan et al. [24] presented a design method-ology for haptic devices based on a combination of the “nested” approach andthe general design process by [25]. Another approach for the design of mecha-tronic systems is to use the V-model [26] in combination with a stage-gatesequential process model (e.g. [27], [28]).

2.2.1 Optimization

Many researchers have addressed the challenge of finding an optimal hapticdevice design using different approaches depending on specific objectives, forexample minimizing or maximizing one or several performance indices such asmaximum kinematic dexterity and transmission ability [29], [30], maximumstiffness [31], minimum position error of the movable platform [32], maximumvelocities in the platform [33], and maximum required workspace [34], [35],[36].

Many methods and approaches have been proposed to achieve an opti-mum design that satisfies multi-criteria requirements and constraints. Gen-erally, these approaches can be classified into two types: deterministic de-sign optimization (DDO) and robust design optimization (RDO). A DDOalways returns the same result for a specific set of design values, while a non-deterministic optimization may return different results for the same set ofdesign inputs.

Finding an optimal solution using a DDO approach for these devices entailshandling a multi-criteria optimal design problem, that is a multi-objective con-strained nonlinear optimization problem with no explicit analytical expression.The gradient and Hessian algorithms that generally converge on a local mini-mum are not suitable for solving this problem. An interval analysis-based ap-proach has recently been used to solve a multi-criteria design problem for par-allel manipulators by Hao and Merlet [37]. This approach determines a designparameter space that satisfies all design constraints. However, this approachrequires explicitly analytical expressions of all constraints. The performance-chart-based design methodology (PCbDM) proposed by Xin-Jun Liu et al.[38] is an optimal kinematic design methodology for parallel mechanisms withfewer than five linear parameters, which is unsuitable for our optimizationproblem with its five design parameters. In this regard, the genetic algorithm(GA) [39] approach seems a good option for multi-criteria problems due to its


good convergence property and robustness. Stan et al. [40], Lee et al. [22],Hwang et al.[41], Raza. R et al. [42], and Khan et al. [43] have used a GAapproach for the multi-criteria optimization of parallel kinematics machinesand parallel haptic devices.

An optimum design using a DDO approach, however, may not alwaysachieve the desired performance due to significant uncertainties that existin the material and geometrical parameters and the component assembly [44].Due to variations, uncertainties, and scatter, most real-world designs do notbehave in a deterministic manner; rather, they behave stochastically, influ-enced by chance or probability. Variation is inherent in, for example, geo-metric properties, manufacturing precision, and actual use of the product.Applying DDO strategies without incorporating uncertainty and measuringits effects leads to designs that cannot be called “optimal”.

A robust optimum design makes the performance minimally sensitive byoptimizing mean performance and minimizing its sensitivity to various sourcesof variation. By accurately assessing the robustness of the design RDO ensuresthat a product performs its intended function with minimal variation in theperformance indices by controlling design parameters without eliminating thecause of the uncertainties [45],[46].

Several authors have contributed to the formulation of robust design prob-lems with the aim of improving the quality of a product whose manufacturingprocess involves variability [47]. A comprehensive review of the state of artin RDO is presented by Hans et al. [48]. Xiaoping et al. [46] developed astrategy for assessing robustness and synthesizing robust mechanisms whenrandom and interval variables are involved, using a Monte Carlo simulation(MCS) to quantify the variability of the performance. A simulated annealingalgorithm was used to maximize robustness by Zhang et al. [49]. A new robustmulti-objective GA (RMOGA) that optimizes two objectives, a fitness valueand a robustness index, was used by Li et al. [50].

2.2.2 System models

To formulate the performance indices we need to develop a number of focusedsystem models, which include kinematic, dynamic and stiffness models.

For kinematic modeling of haptic devices, both closed loop and open archi-tectures have been studied, for example, [51], [52], [22]. Cui H et al. [53] andZhu Z et al. [54] have worked on the kinematic analysis, error modeling, anddynamic modeling for real-time control of parallel (TAU) robots.

Most of the different types of haptic devices have been dynamically in-vestigated by researchers using different approaches such as the Newton Eulerformulation [55], the Lagrangian formulation [56], the principle of virtual work[57], and the generalized momentum approach [58]. Nguyen et al. [59] usedgeneralized momentum methods to model a Stewart platform with the legs as

2.2. HAPTIC DEVICE DESIGN 13

point masses. Ahmadi et al. [60] studied dynamic modeling and simulationof a parallel robot Hexa using Lagrangian equations of the first type.

Aiming for high dynamic performance with relatively small actuators andlow energy consumption motivates the design of lightweight structures [61].A light-weight design is a compromise or trade-off between low weight andhigh stiffness. There are several methods available for computing the stiffnessmatrix, including the virtual joint method often called the lumped model-ing method), finite-element analysis (FEA), and matrix structural analysis(MSA).

Stiffness analysis has been widely investigated in the literature. Severalmethods exist for computation of the stiffness matrix: the virtual joint method(VJM), which is often called lumped modeling [62], [63], [64], [65], finite el-ement analysis (FEA) [66], [67] , and matrix structural analysis (MSA) [68],[69], [70], [71]. Uchiyama [72] has derived an analytical model for the stiff-ness of a compact 6-DOF haptic device based on static elastic deformation ofcompliance elements.

Various prototypes have been tested experimentally: Pinto et al. [73] eval-uated static stiffness mapping of their Lower Mobility Parallel Manipulatorusing pre-loading in the experimental testing to eliminate backlash in the sys-tem. In experimental testing [74], the static behavior evaluation of a robot wasanalyzed by norms, for example ANSI/RIA R15.05-1-1990 [75], ISO 9283:1998[76], which were established for serial manipulators.

2.2.3 Control design

Various control strategies such as open-loop impedance control, impedancecontrol with and without force feedback [77], [78], and admittance controlfor model-based compensators [79] are employed in haptic interfaces to maskthe inertia and friction, and improve transparency and fidelity of force/torquefeedback of the system. A force-based impedance control approach has beenpresented in [80] and [81], and showed that this approach can improve trans-parency and fidelity of force torque feedback. A closed-loop impedance control(CLIC) with model-based compensator (MBC) and torque compensator basedon motor current (TCBMC) is used by [80] and [81] to improve transparencyof haptic devices and mask the parasitic force/torque. These compensatorsrequire models, which can accurately predict inertia, stiffness and friction phe-nomena.

For inertia compensation Carignan et.al [77] used model-based compen-sation. The dynamic model for inertia compensation should be linear in itsparameters. Honegger et.al [82] used dynamic equations in linear form fornon-linear adaptive control of a Hexa-glide type parallel robot.

High quality haptic interaction requires compensation of frictional effects,which relies on a friction model that enables accurate prediction of the frictionforces. A friction model for friction compensation in real time must also be


computationally efficient. Different friction models exist in the literature, bothsimple and advanced to predict friction phenomena from simple to complexbehavior. These models are investigated in terms of realistic friction phenom-ena and computational efficiency in a comprehensive survey in [83], where theSmooth Generalized Maxwell slip model (S-GMS) [84] was suggested. Jianjunet.al [85] used a model-based compensation to compensate for robot deflectioncaused by external forces.

As stated in the previously mentioned literature, there are two types ofcontrol structures, as impedance and admittance control are widely used forhaptic devices [77], [86], [87]. To implement impedance control an open-loopcontrol structure is traditionally used [88]. It is then assumed that the dynam-ics of the device, for example time dependent disturbances, friction, inertialand gravitational forces are negligible. However, it is also reported in theaforementioned literature that the quality of haptic feedback (forces/torques)in open-loop is susceptible to environment model errors and time dependentdisturbances [81]. Therefore, for high transparency of a device, it is well moti-vated to compensate for the dynamics and frictional forces of the device usingimpedance control with force feedback. In these approaches the force is feed-back through commercially available force/torque sensors, which are ratherbulky and expensive and can increase system inertia. Furthermore, they canincrease inertia at the TCP of the platform. An alternative solution for theforce feedback is an estimated force/torque using motor torque based on cur-rent measurement transformation using the Jacobian matrix to TCP of theend effector. Another solution is the use of dynamic force observers, whichestimate external forces using a system dynamic model. For such an approachVan Damme et al. [89] proposed recursive least squares estimation with expo-nential forgetting to estimate the end-effector force from noisy actuator torquemeasurement.

2.3 Chapter summary

This chapter describes state of the art for the development of haptic devices.An overview of the characteristics of existing haptic devices is given, followedby an overview of haptic device design, including optimization, system modelsfor performance indices and control design.

Chapter 3

Model-based and simulation-drivendesign

This chapter presents an overview of a model-based and simulation-driven de-sign approach for 6-DOF haptic devices design in order to have a high trans-parent force/torque feedback from the virtual environment. Transparencyquantifies the correspondence between the desired and actually observed val-ues of force/torque feedback. In an ideal scenario, the force/torque feedbackobserved at the TCP of the end effector of the haptic device must reflect theactual force/torque generated by interacting with the virtual tool in the vir-tual environment. Haptic device transparency is restrained by device naturaldynamics, gravitational and frictional forces, which can have a significant im-pact on the kinematic, dynamic and kinetostatic performances. An adequatecontrol system must be implemented for compensation of unwanted effectscaused by the mechanical design.

Haptic device transparency depends on the device dynamics and controlalgorithm, and in this thesis the focus has been to improve transparency byminimizing effects from natural dynamics of the device and by using com-pensation in the control design. The approach is formulated as a two-stageapproach, where a basic level of transparency is achieved during the concep-tual design phase and further improved during the detailed design phase. Inthe conceptual design phase, the suggested approach is to use a model-basedand simulation-driven design approach, while in the detailed design, a model-based control strategy is suggested. The approach has been applied to twodifferent case studies to verify its effectiveness. It has been applied on the de-velopment of a new haptic device based on the TAU configuration and partsof it have been validated on the Ares device, which is based on a Stewartplatform.

Numerous design aspects based on the natural dynamics of haptic devicescontribute to the performances. These performances can be described in termsof a large workspace, high manipulability/isotropy torque requirements on anactuator for unit force/torque applied, and high stiffness for precise position-

15

16 CHAPTER 3. MODEL-BASED AND SIMULATION-DRIVEN DESIGN

ing, low inertia and low resonant frequency. Furthermore, these performanceare evaluated numerically through performance indices like workspace volumeindex (VI), kinematic isotropy index (II), quasistatic force/torque requirementindex (TRI), dynamic dexterity index (MI), stiffness index (KI), and naturalfrequency index (NFI) (described in Papers A and C).

The performance requirements are extracted from the requirement speci-fication of the device, and used to identify the main function of the deviceand basic constraints on the solution space. The main function is described interms of functional requirements, which are derived from device and user re-quirements described in section 1.2, and are the basis for generating potentialdesign concept, as shown in Figure 3.1

Workspace

Manipulability

Payload

Stiffness

Inertia

Friction

Backlash

Design concept

DOF

Functional requirements

Workspace

Kinematic Model

Inverse kinematic

modelForward

kinematic model

Friction ModelStiffness

Model

Dynamic ModelJoint

clearance Model

Kinematic isotropy

Quasi-static force

Stiffness isotropy

Dynamic isotropy

Jacobian matrix

Control design

Transparency

Backlash

User requirements

Devicerequirements

Figure 3.1: Dependencies between functional requirements, design concept and modelsneeded for evaluation of product performance.

The multi-criteria nature of performance requirements of haptic devicesmotivates the use of a model-based and simulation-driven design approach.Furthermore, the design and development of haptic devices involves a greatdeal of uncertainties and project risk. This dilemma can be solved by using aprocess model like the V-model, which focuses on minimizing technical risksby dividing the product development process into a sequence of decompositionstages. This decomposition from abstract idea to realization of the physicalproduct is often referred to as the left leg of the V-model, which is followedby integration stages, that is the right leg.

The decomposition of design tasks for a product with hybrid technologicalconstituents, like a mechatronics product, is not obvious, but one attemptto adapt and formalize the generic V-model for mechatronics development isthe German standard VDI 2006 [26]. A coarse representation of VDI 2006is shown in Figure 3.2. On top of V-model is also represented a stage-gatesequential process model (e.g. [27], [28]), which has as its main purpose toreduce the economical project risks by providing a mechanisms to a company

17

management to make a go/no-go/wait decision at each main decision gate.In many industrial organizations, the V- and extended stage-gate models aresynchronized.

Mechanical engineering

Electronics engineering

Control engineering

Product requirements Product/unit

Domain specific design

G0 G-1 G1 G4 G5 G6 G7 G2 G3

Figure 3.2: The mechatronics development V-model combined with the stage-gate model.

The design process is evaluated through an iterative design using a designpyramid as shown in Figure 3.3 (Paper A), which focuses on model-basedand simulation-driven (e.g., [90] , [91]) design exploration, where one startsfrom an abstract top-level model, that is extended via stepwise refinementand design space exploration, into a detailed and complete physically realizedsystem.

Detail Design

Industralization

Ab

stra

ctio

n L

evel

Lev

el o

f d

etai

l

Product idea

Requirements specification

Conceptual design

Market analysis

(a) (b)

Feasible design space

Solutions

· Alternatives

exploration

· Different design,

different solutions

Figure 3.3: Design pyramid with different abstraction levels [91].

The design pyramid is evaluated both at the system and individual modellevel, where the models were developed using stepwise refinement by addingmore details to the models, which are quite coarse in the first instance towards


such a detail that the models can predict the complete behavior of the systemand investigate whether it fulfills the required performances or not (Papers Aand C). In this thesis, the developed models are presented in the context ofthe Ares and TAU 6-DOF haptic devices, and a brief overview of these modelsis given in the coming subsections.

3.1 Model development in the conceptual design phase

The first step of the proposed design approach concerns activities during theconceptual design phase. The activity that has a major impact on deviceforce/torque feedback transparency is the activity that determines the ba-sic kinematic structure parameter and basic shape parameters. This task isaddressed by a new RDO approach, which is based on hybrid design opti-mization, which combines a GA and MCS. This optimization must balancemulti-criteria requirements and constraints expressed in the form of designindices. These performance indices are expressed based on a number of be-havior models for the studied TAU 6-DOF haptic device. The formulationand verification of these models are given in the coming sections.

3.1.1 Kinematic model

In the kinematic model development, the inverse and velocity kinematics havebeen considered. Solving the inverse kinematic problem for the 6-DOF TAUconfiguration as shown in Figure 3.4, determines all orientations (i.e., joint an-gles) [θi1, θi2] of all active joints when the pose (i.e., position and orientation)q = [px, py, pz, α, β, γ] of the platform are given.

31

XN N

1.5 d

1.5 d L1

L2

1112

2221

32

1a

1b

1c

1d

1e

2e

2c

2d

2b2a

3c

3b3a

12

11

1314

21

2223

24

31

YN

ZN

P

XP

YP

ZP3 1 5L d .

1L

2L

1.5 d

3L

Figure 3.4: Kinematic structure of TAU haptic device.

3.1. MODEL DEVELOPMENT IN THE CONCEPTUAL DESIGN PHASE 19

A closed-form solution for the inverse kinematics can be found based onthe constrain equations. There are two constrain equations for symmetricalchains according to the kinematic path (ai, bi, ci) and (ai, bi, di, ei) shown inFigure 3.4, as given in equation (3.1) and (3.2).

(cix − bix)2 + (ciy − biy)2 + (ciz − biz)2 = L22 (3.1)

(dix − eix)2 + (diy − eiy)2 + (diz − eiz)2 = L22 (3.2)

3.1.2 Dynamic model

The dynamic model describes the inherent property like inertia of a system.The Lagrange dynamic formulation is used to develop the equation of motionof the form as in (3.3),

M(θ)θ + V (θ, θ) +G(θ) = τJ (3.3)

where the first term M(θ) in equation represents inertial forces, while the

second term V (θ, θ) accounts for the Coriolis and centrifugal forces. The lastterm G(θ) on the left side is the gravity force, while τJ is the generalized Carte-sian forces/torques at the end-effector. A detailed formulation of dynamicsequations is given in (Paper C).

3.1.3 Stiffness model

As it has been previously pointed out in Chapter 1, the design of haptic de-vices is not trivial due to its complex design, multi-domains consideration andin finding the best compromise between several properties such as workspace,dexterity, manipulability, and stiffness. Stiffness is an essential performancemeasure since it is directly related to the positioning accuracy and payloadcapability of haptic devices. To make these devices compatible with their ap-plications it is necessary to model, identify and compensate all of the effectsthat degrade their accuracy. Stiffness can be defined as the capacity of a me-chanical system to sustain loads without excessive changes of its geometry, orthese produced changes on geometry, due to the applied forces, are known asdeformations or compliant displacements [92]. Compliant displacements in arobotic system produce negative effects on static and fatigue strength, wearresistance, efficiency (friction losses), accuracy, and dynamic stability (vibra-tion). In this thesis, a systematic procedure is proposed in order to developa generalized stiffness model within the workspace of the manipulator and itsevaluation with virtual (FEM analysis) and physical experiments. To developthis, a work procedure defined by the flowchart in Figure 3.5 is presented. Theprocedure established by this chart is outlined below.

This methodology starts with the development of a simplified analyticalmodel. This is verified by a simplified FEM model in an iterative way until


Simplified Analytical Model

Start

Simplified FEM Model

FEA=Simplified Anaytical

Simplified Physical experiments

Yes

Physical experiments= Simplified FEA=Simplified Anaytical Model

Detailed Analytical Model

No

Detailed FEM

Detailed Physical Experiments

No

Detailed FEA=Detailed Anaytical Model

Physical experiments= Detailed FEA=Detailed Anaytical Model

Sensititvity Analysis

yes

Simplification based on contributions

Design Optimization and RoubustnessStiffness compensation in control

Model parameters adjustment and

hypothesis revision

No

Model parameters adjustment

No

Simplified Modelling

Detailed Modelling

Figure 3.5: Stiffness modeling methodology.

these two models give the same result. Thereafter, a simplified physical ex-periment is made to validate the analytical model. If the difference betweenthe analytical and the experimental results is not acceptable, in other words itdoes not validate the analytical model, a more detailed analytical model mustbe developed.

For the detailed analytical model the passive joints and actuation systemare also included. This is further verified using a detailed FEM model withcorresponding detailing level in the same way as for the simplified model.Thereafter, this detailed analytical model is validated by means of a detailedphysical experiment. After validating the proposed model, a sensitivity anal-ysis is performed to map the variation of static stiffness in the workspace. Inthese maps, the engineering stiffness is visualized as a function of the general-ized coordinates of the workspace. A detailed description of this methodologyis given in Paper B.


For the case of N number of serially connected compliant elements withlocal stiffness i−1Ki, a general stiffness 0KN of the TCP of the end-effectorcan be obtained by force/displacement transformation Jacobian i−1Ji, whichcan be expressed in the base frame 0, as given in equation (3.4):

0KN = 0JN−1N−1SN

0JTN−1 + ......

0NJN−1,disp(

0JN−2N−2SN−1

0JTN−2)

0NJN−1,force + .....

+0NJi,disp(

0Ji−1i−1Si

0JTi−1)

i0JN,force + .....

+0NJ1,disp(

0J10S1

0JT1 )01J1,disp

(3.4)

3.1.4 Model verification and validation

To verify the developed kinematic, dynamics and stiffness models, the firsttwo models were verified in MapleSim [93], while the third model was verifiedby FEM analysis (ANSYS [94]) and validated by physical experiments. TheMapleSim model is shown in Figure 3.6.

X

Z

Y

Figure 3.6: MapleSim model.


Kinematic model

In order to verify the inverse kinematic model, an input trajectory of wasapplied in the rotational degree of freedom along the x-axis. The results fromboth the analytical and MapleSim models are shown in Figure 3.7.

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5−52

−51

−50

−49

−48

−47

−46

−45

−44

Time(s)

θ 11 (

deg)

AnalyticalMaple

(a) Rotation of actuator 1

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5−15

−10

−5

0

5

10

15

Time(s)

θ 12 (

deg)

AnalyticalMaple

(b) Rotation of actuator 2

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 544

45

46

47

48

49

50

51

52

Time(s)

θ 21 (

deg)

AnalyticalMaple

(c) Rotation of actuator 3

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5−15

−10

−5

0

5

10

15

Time(s)

θ 22 (

deg)

AnalyticalMaple

(d) Rotation of actuator 4

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5−6

−4

−2

0

2

4

6

Time(s)

θ 31 (

deg)

AnalyticalMaple

(e) Rotation of actuator 5

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5−24.5

−24

−23.5

−23

−22.5

−22

−21.5

−21

Time(s)

θ 32 (

deg)

AnalyticalMaple

(f) Rotation of actuator 6

Figure 3.7: Comparison of Inverse kinematics by analytical and MapleSim

By comparing these results, we can verify the inverse kinematics calcula-tion approach. These results show that the analytical model agrees with theMapleSim model, which verifies the model.

Dynamic model

The dynamic model was verified using the same trajectory sin(ωt) on theactuators, and the required torques on the actuators were calculated based onthe analytical dynamic model of section of 3.1.2 and MapleSim model. Theresults from these analyses are shown in Figure 3.8, where we can concludethat the results from the analytical model agree with the MapleSim model,which verifies the model.

Stiffness model

The stiffness model developed in section 3.1.4 was verified through simulationusing ANSYS and validated by physical experiments. In this verification andvalidation the Ares 6-DOF haptic device was used. A detailed description of


0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5−2.5

−2

−1.5

−1

−0.5

0

0.5

1

1.5

2

Time(s)

τ 11 (

Nm

)

AnalyticalMaple

(a) Torque on actuator 1

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5−16

−14

−12

−10

−8

−6

−4

−2

0

2

4

Time(s)τ 12

(N

m)

AnalyticalMaple

(b) Torque on actuator 2

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5−2

−1.5

−1

−0.5

0

0.5

1

1.5

2

Time(s)

τ 21 (

Nm

)

AnalyticalMaple

(c) Torque on actuator 3

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5−20

−15

−10

−5

0

5

10

Time(s)

τ 22 (

Nm

)

AnalyticalMaple

(d) Torque on actuator 4

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5−0.25

−0.2

−0.15

−0.1

−0.05

0

0.05

0.1

0.15

0.2

0.25

Time(s)

Tor

que

(Nm

)

AnalyticalMapleSim

(e) Torque on actuator 5

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 510

11

12

13

14

15

16

Time(s)T

orqu

e (N

m)

AnalyticalMapleSim

(f) Torque on actuator 6

Figure 3.8: Comparison of torque on actuators given by analytical model and MapleSim,respectively.

the validation is given in Paper C and some of the results are shown in Figure3.9.

0 1 2 3 4 5 6 7 80

0.05

0.1

0.15

0.2

0.25

0.3

Points in workspace

Def

lect

ions

in X

[mm

]

AnalyticalAnsysExperimental

(a)

0 1 2 3 4 5 6 7 80

0.05

0.1

0.15

0.2

0.25

0.3

Points in workspace

Def

lect

ions

in Y

[mm

]


(b)

0 1 2 3 4 5 6 7 8−0.02

−0.01

0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

Points in workspace

Def

lect

ions

in Z

[mm

]


(c)

Figure 3.9: Deflections of platform in (a) X-, (b) Y-, and (c) Z-directions, given by theanalytical model, ANSYS model, and experimental tests, respectively.

3.1.5 Robust optimum design

Transparency of a haptic device can be improved through design, which is op-timum and whose performance are less sensitive to variations in manufacturing


tolerances of optimum numerical value of design parameters. The haptic de-vice that has been used as a basis for developing this approach is the TAU6-DOF device and the optimization formulation in equation 3.6 is specific forthis device. However, the stepwise approach outlined below is general for anyhaptic device. The steps proposed to achieve a robust optimum design, asshown in Figure 3.10, are outlined as follows:

for i=1 to n

maxn n

max min,GI GI

Generate n numbers of random parameters within bounds of

design parameters

Objective function evaluation

( , , , , , )

calculate

GI VI GII GTRI GKI GMI GNFI

Start

Define bounds of design parameters

Generate random population

Monte Carlo Simulations

Store population g Rif pop pop

n nGI GI

Robustness criteria

+3 1

max

selection of

individual solutions

for i 1 to G

SelectionCrossoverMutation

maxG G

Evaluate fitness of each solution

Robust optimum solution

No

Yes

Yes

Monte Carlo Simulation

Normal distribution of design parameters

n nGI GI

Filter based on

robustness criteria

+3 1

,n nGI GI

Optimum tolerance design Using manufacturing cost

minimizaiton

Robust design optimization

Monte Carlo Simulation

Figure 3.10: Robust optimization approach.

• Normalize the performance indices by finding the maximum and min-imum bounds of the performance indices under the given constraints,and also find the bounds of the design parameters using Monte Carlosimulations.

• Perform RDO using a GA to find optimum robust values for the designparameters using robustness criteria on the performance indices. This isto ensure that the optimum design parameters are robust to tolerancevariations.


• Perform sensitivity analysis of the performance indices using Monte Carlosimulations based on the variations in the optimum numerical values ofthe design parameters obtained from the RDO in step 2. Perform designoptimization to find optimum tolerances that minimize manufacturingcost and satisfy the allowed variations in the performance indices.

The proposed RDO approach seeks to find numerical values of optimumdesign parameter by maximizing volume index (VI), global kinematic isotropyindex (GII), global stiffness index (GKI), and global natural frequency index(GNFI), while minimizing the global quasistatic force/torque requirement in-dex (GTRI) and dynamic dexterity index (MI). The design optimization isevaluated in a dexterous workspace (Paper C) along with the robustness cri-teria defined in equation 3.5, which tries to minimize the objective functionGDIR.

µGIn + 3σGIn ≤ 1 (3.5)

where µGIn and σGIn is the mean value and standard deviation of normalizedperformance indices (Paper C ). The robust design approach can be formulatedin the form of an optimization as given in equation 3.6:

minimizex

1

GDIR (x)

over x = [L1, R1, t1, L2, R2, bL3 , hL3 , Rp, θp, tp, θ32nom , ln]

subject to min(θii) ≤ θii ≤ max(θii)

min(φik) ≤ φik ≤ max(φik)

xlb ≤ xj ≤ xup

det(J) > 0

det(K) > 0

det(M) > 0

µGIn + 3σGIn ≤ 1

Nuvw =

{1 if Puvw ∈ W (X, Y, Z, α, β, ζ)

0 if Puvw /∈ W (X, Y, Z, α, β, ζ)

(3.6)

where variable x is a vector of design parameters, xlb and xup are thelower and upper bounds of the design parameters, respectively, while min(φik),min(θij), max(φik) and max(θij) are the minimum and maximum bounds ofthe active and passive joints, det(J) > 0; det(K) > 0 and det(M) > 0 aresingularity conditions in the workspace. The term J , K, and K defines theJacobian matrix, stiffness matrix and inertia matrix, while Puvw and Nuvw

defines the grid point in the workspace and binary flag, a grid point is areachable point when Nuvw = 1; otherwise not when Nuvw = 0. The variation


in the design parameters is assumed to be normally distributed and ±3σxjcovers the tolerance interval.

3.2 Control design in the detailed design phase

The second step of the proposed design approach concerns activities duringthe detail design phase. In this phase, we focused on improving the force/-torque transparency by compensating for unwanted effects in the device, whichreduces transparency. We developed a model-based control strategy to com-pensate for the natural dynamics of the device, friction between joints, gravityof platform, and elastic deformation. For this strategy, we used the Ares 6-DOF haptic device as a test case for the development, testing and verificationof the control strategy. The objective of this strategy is to minimize the prob-lems with the existing approaches, and make it cost effective. The proposedapproach is based on closed-loop impedance control (CLIC) with model basedcompensation and force/torque feedback, where the force/torque at TCP ofthe platform is to be estimated using recursive least squares estimation withexponential forgetting [89]. The disturbance torque to compensate for theparasitic effects was estimated by their corresponding models, and added tothe torque applied by the feed forward of the reference force Fd and feedbackof estimated force Fest, where Kp is the gain of force feedback controller, andJ denotes the Jacobian matrix. The proposed control structure is shown inFigure 3.11.

estF

Device dynamics

mI

Recursive least square

estimation

q

l

l

q

FKM

q

Kp

FKM VE

Stiffness model

l

Enco

der

re

adin

g

1J

1gR

tK

J

gR

dF

Friction model( )f l

Filtered dynamics

( , , , )K l l q q p

estF

TJ

ContactFree space

u

Model-based compensators

Figure 3.11: Control design of 6-DOF haptic device.

3.2. CONTROL DESIGN IN THE DETAILED DESIGN PHASE 27

3.2.1 Friction estimation

In this thesis, different friction models were investigated to determine howwell these models are suited for performance simulation and control of a6-DOF haptic device. The studied models include the Dahl model, LuGremodel, Generalized Maxwell slip model (GMS), smooth Generalized Maxwellslip model (S-GMS), and Differential Algebraic Multi-state (DAM) frictionmodel. These models were evaluated both numerically and experimentallywith a 6-DOF Ares haptic device. To evaluate these models, we used criteriabased on fidelity to predict realistic friction phenomena, ease of implementa-tion, computational efficiency, and ease of estimating the model parameters.A detailed overview of evaluation is described in Paper D.

In this investigation, the S-GMS was identified as the best candidate forthe real-time friction compensation. A brief description of the selected modelis given below:

The analytical multi-state S-GMS model according to [95] is given as in(3.7);

dFi

dt= ki(l − ηi(Fi, l)

∣∣∣l∣∣∣ Fi

αig(l)) (3.7)

where ηi(Fi, l) is a smoothed transition function between presiding (ηi ' 0)and sliding regime (ηi ' 1), while ki is the stiffness of element i. The transition

function ηi(Fi, l) consists of two multiplicative terms given in equation (3.8),the transition function (3.9) and the reset function (3.10)

ηi(Fi, l) = ηAi(Fi, l)ηBi(Fi, l) (3.8)

ηAi(Fi, l) = 1− l

2tanh

[λ

(Fi

αig(l)+ ς

)]

+1

2tanh

[λ

(Fi

αig(l)− ς)]

(3.9)

ηBi(Fi, l) =1

2+

1

2tanh

[γ

(Fi

αig(l)

l

νs

)](3.10)

Where ς ∈ [0.9, 1]is a factor to avoid ringing in sliding regime and set thefrictional lag, the factor λ sets the transition sharpness of (3.9). The resetfunction (3.10) vanishes as the velocity reverse its direction. The factor γ sets

the transition sharpness of (3.10). The function g(l) describes the Stribeckeffect, which is defined as

g(l) = (fC + (fS − fC)e−( lνs

)2)


Where fS and fC are static and Coulomb friction force, νs the Stribeckvelocity, while

αi =Fi∑Ni=1 Fi

The resulting friction force is the summation of N number of stiffness forcesFi plus a viscous σv l term as given in equation (3.11).

τf (l) =N∑

i=1

Fi + σv l (3.11)

3.2.2 Force torque estimation

As an alternative to using a force/torque sensor, we developed an approach toestimate the force/torque feedback at the TCP of the end effector. For thispurpose, the filtered dynamic equation proposed by Van Damme et al. [89] wasused. A solution of the dynamic equation for the haptic device, which showsthe torques on actuators in the joint space, when an external force/torqueacting on the TCP of the end effector is given in 3.12 as,

H(θ)θ + C(θ, θ)θ +G(θ) = τ + JT (q)fe − τf (θ) (3.12)

here θ is the vector of joint variable, H(θ) is the inertia matrix, C(θ, θ)is the centrifugal matrix comprised of Centrifugal and Coriolis forces. G(θ)

is a vector that contains gravitational torque on the joints, τf (θ) is a vectorcontaining friction torques, and the τ is the vector representing the actuatortorques.

The dynamic equation defined in (3.12) was filtered with a low-pass filter

to remove the acceleration term θ, then a recursive least square estimationwas used to estimate the force/torque at TCP of the platform. A detaileddescription of the proposed approach is given in Paper E.

3.2.3 Transparency measurement

Transparency GT of the haptic device can be defined according to [96], as theratio of actual transmitted force to the user Ft , to the desired virtual forcegenerated in the virtual environment Fg as in equation 3.13.

GT =Ft

Fg

(3.13)

According to equation 3.13 for high transparency the ratio should be closeto unity. A comparison of transparency for uncompensated and compensatedinteraction in x, y and z-directions are shown in Figure 3.12. The results from

3.3. CHAPTER SUMMARY 29

this comparison show a significant transparency improvement of the proposedapproach.

(a) (b) (c)

Figure 3.12: Transparency for uncompensated and compensated forces while interactingwith a virtual object in (a) X-, (b) Y-, and (c) Z-directions.

3.3 Chapter summary

This chapter presents an overview of a model-based and simulation-drivendesign approach for design of 6-DOF haptic devices with a high transparencyin the force/torque feedback from the virtual environment. The first step of theapproach is applied during the conceptual design phase and focuses on RDOto find basic kinematic structure parameters and basic shape parameters. Thesecond step is applied during detail design phase, where the goal is to furtherimprove the transparency in the control design by compensating for unwantedeffects.

Chapter 4

Prototypes

Research dealing with design and development of physical products has a greatneed for prototypes to use as test cases in the virtual world and for experi-mental testing. In this work two physical 6-DOF haptic devices were used fordevelopment and testing of theories and methods. The first device, Ares, isbased on a Stewart platform and was initially developed and manufacturedduring an earlier project. The second device is named TAU because of thekinematic structure (TAU) that it is based on and has been developed andmanufactured during this project.

4.1 The Ares haptic device

A physical prototype and schematic of the 6-DOF Ares haptic device is shownin Figure 4.2. The schematic structure consists of a fixed base, a movingplatform and six identical kinematic chains connecting the platform to thebase. Each kinematic chain consists of an active actuator fixed to the base toactuate a linear guideway of length L1 using cable transmissions to providelinear motion to each linear guideway, a spherical joint, a constant lengthproximal link of length L2, and a universal joint. The joint attachment pointpairs are symmetrically separated by 120◦, and lie on a circle, both on the baseand platform. The platform attachment points are rotated 60◦ clockwise fromthe base attachment points. The 6-PSU (active prismatic, passive sphericaland universal) joint’s configuration is used to get 6-DOF motion at TCP.

4.1.1 Model development and analysis based on the Ares haptic device

The Ares haptic device was developed during an earlier research project (seefor example Khan et al. [3]). In this work, this device was used for thedevelopment of the stiffness modeling methodology, where the verification ofthe approach was made with measurements on this device. It has also beenutilized for the evaluation of friction models as discussed in section 3.2.1,as well as for the development and verification of the proposed model-based

31

32 CHAPTER 4. PROTOTYPES

(a)

y

{A}x

z

{U} xA

yA

zA

xB

yB

zB

{B}

xC

yC

zC

{C}

2β

{P} xP

yP

zP

2α

Spherical jo int

Base

1L

Platform

Universal joint

2L

Linear guideway

Pr oximal link

(b)

Figure 4.1: (a) Physical prototype, and (b) schematic of 6-DOF haptic device.

strategy in Paper E. It has also been used in the formulation of the model-based and simulation driven design approach for haptic devices.

4.2 The TAU haptic device

The TAU haptic device was designed and implemented within the timeframeof this thesis work. Based on a deterministic optimization of the kinematicstructure in the early parts of the project, a mechanical prototype systemwas designed and developed and physically realized during a capstone courseproject in machine design at Kungliga Tekniska Hogskolan (KTH) . This struc-ture was slightly modified later in the project when the RDO approach [97]was developed and applied to find optimum numerical values of design param-eters of the kinematic structure of the 6-DOF TAU haptic device, as shownin Figure 4.2. Based on these slightly adjusted optimum numerical values ofdesign parameters, the prototype was changed accordingly. A brief overviewof the prototype is given in this paper.

The TAU schematic consists of a fixed I column, made from steel, a movingplatform, and three kinematic chains i = 1, 2, 3, where the two chains i =1, 2 are symmetrical, while the third chain i = 3 is asymmetrical. Thesethree chains have two active revolute joints with angles θij, where i = 1, 2.Furthermore, the two symmetrical chains have two passive universal jointswith angles φik where k = 4, while the third chain has a passive revolute jointwith angle φ31. The three kinematic chains connect the base coordinate N tothe moving platform.

The size of the prototype is 250×250×300 mm. Six Maxon [98] DC motors(model-339152, RE 25 mm, Graphite Brushes, 20 watt) coupled with a HEDL5540 encoder with a resolution of 500 counts per revolution along with 4-Q-DC Servo Amplifier were selected. Three of the motors with rotation angles

4.2. THE TAU HAPTIC DEVICE 33

(a)

31

XN N

1.5 d

1.5 d L1

L2

1112

2221

32

1a

1b

1c

1d

1e

2e

2c

2d

2b2a

3c

3b3a

12

11

1314

21

2223

24

31

YN

ZN

P

XP

YP

ZP3 1 5L d .

1L

2L

1.5 d

3L

(b)

Figure 4.2: (a) Physical prototype, and (b) schematic of 6-DOF TAU haptic device.

[θ11, θ21, θ31] are fixed to the I column, while the motors with [θ12, θ22, θ32]are attached to [θ11, θ21, θ31] through a kinematic link close to the I column,thus reducing the moving inertia of the device. The active and proximallinks with length L1 and L2, respectively, are made of carbon fiber reinforcedpolymer, while the platform is made from ABS plastic. Therefore, the overallmoving weight and inertia of the device are small.

4.2.1 Transmission system

A cable transmission mechanism with a pulley was used for the gear ratio,as shown in Figure 4.3, which has a small contact area as compared to thetransmission with a linear guideway in a 6-PSU joint-based configuration likea Stewart platform [99]. A similar cable transmission mechanism is also usedin the Phantom haptic device [100] to make it back-drivable. This character-istic improves the transparency of the overall haptic system and simplifies thecontrol of the device.

4.2.2 Spherical joint design

The proposed parallel mechanism has five spherical joints. However, a con-ventional ball-and-socket type spherical joint has rotational limitations andlarge friction. Therefore, it is necessary to design a spherical joint that allowsfor the largest possible rotations with less friction. A combination of universaland revolute joints was used to provide rotations in 3-DOF.


Figure 4.3: Cable based capstan transmission mechanism

4.2.3 Model development and analysis based on the TAU haptic device

The TAU haptic device was developed during this project, wherein the knowl-edge of kinematic modeling and optimization of the Ares device was furtherdeveloped and applied during the design of this device. In this thesis work,the device was used for developing kinematic, dynamic and stiffness modelsthat were used during the development of the RDO approach described insection 3.1. It has also been used in the formulation of the model-based andsimulation-driven design approach for haptic devices.

4.2.4 Performance evaluation of the TAU haptic device

An evaluation of the main characteristics of the TAU haptic device has beenmade in order to investigate how well the stated requirements in Chapter 1have been fulfilled.

Workspace of a 6-DOF TAU haptic device

The physical prototype of a 6-DOF TAU haptic device was developed based onthe optimum numerical values of the design parameters of [97]. The reachableworkspace based on these design parameters is shown in Figure 4.4 and is freefrom singularities within that space.

Force/torque capability

In order to measure the force and torque capability of the developed prototype,we used the force/torque estimation technique described in Chapter 3. Theestimated maximum forces/torques capabilities at TCP are shown in Table

4.2. THE TAU HAPTIC DEVICE 35

Enco

der

rea

din

g

A/D

Cu

rren

t M

easu

rem

ent

Po

siti

on

co

ntr

ol

Sim

ulin

k

dSp

ace

Cu

rren

t am

plif

ier

dri

ver

YN

ZN

12

11

21

22

31

32

XN

Motor 6

Motor 5

Motor 4

Motor 3

Motor 2

Motor 1

Figure 4.4: Reachable workspace of 6-DOF TAU.

4.1. The peak forces and torques capabilities of the device, measured at eachcorner of the cube within the workspace shown in Figure 4.4, are fulfilling thestated force and torque requirements in Chapter 1.

Table 4.1: Force/torque capability of a 6-DOF TAU haptic device.

Points Fx [N] Fy [N] Fz [N] Tx [Nm] Ty [Nm] Tz [Nm]1 53.3 54.2 58.4 1.23 1.12 1.222 52.2 50.5 58.3 1.2 1.25 1.193 56.8 56.2 61.2 1.4 1.34 1.394 55.4 53.8 60.0 1.29 1.21 1.115 52.5 49.1 54.6 1.18 1.24 1.096 53.0 50.2 53.5 1.13 1.12 1.237 52.3 53.9 52.8 1.04 1.21 1.158 50.2 55.1 53.7 1.23 1.22 1.199 49.98 51.7 52.9 1.13 1.31 1.4

Stiffness capability

In order to verify the stiffness of the developed device, the TCP of 6-DOFTAU haptic device was positioned at the center (point 0, nominal position)and at eight corner points, 1-8, of a cube of 50 × 50 × 50 mm within thereachable workspace of the haptic device, as shown in Figure 4.5. The inversekinematic model was used to calculate the active joint angles for the requiredposition of TCP. At each position the deflection was measured for some givenloads and an average stiffness was calculated.

4.2.5 Performance characteristics

The performance characteristics of the developed TAU 6-DOF haptic deviceare summarized in Table 4.2.


Figure 4.5: Stiffness measurement of a 6-DOF TAU haptic device.

Table 4.2: Performance characteristics of a 6-DOF TAU haptic device and mapping betweenrequirements.

Characteristics Required values PrototypeDOF 6 6

Device footprint [mm] 250× 25× 300 250× 25× 300

Workspace

Translation [mm] 50× 50× 50 70× 80× 100

RotationRoll ±40◦ ±45◦

Pitch ±40◦ ±45◦

Yaw ±40◦ ±45◦

Minimum stiffness [N/mm]50 (x) 60 (x)50 (y) 70 (y)50 (z) 55 (z)

Maximum force [N]- 62 (x)- 57 (y)- 48 (z)

Continuous force [N] - 35Sampling time [ms] - 1

In summary, the stated requirements given in Chapter 1 on a haptic devicefor surgical applications with interactions with hard tissue have been fulfilledby the developed TAU device.

4.3. CHAPTER SUMMARY 37

4.3 Chapter summary

This chapter presents the physically realized haptic devices that have beenutilized in this thesis work. Firstly, the Ares 6-DOF haptic device is brieflydescribed along with the developed models and research activities. Thereafter,the TAU 6-DOF haptic device, which is realized based on the research resultsin this project, is described. The developed models and proposed methodsbased on this device are given, and finally a performance evaluation of thecharacteristics of the TAU haptic device is given.

Chapter 5

Summary of appended papers

This chapter gives a brief review of the appended papers.

5.1 Paper A: a model-based and simulation-driven methodologyfor design of haptic devices

In this paper a systematic step-by-step model-based and simulation-drivendesign methodology has been developed for high precision and reliable hapticdevices. Design of these devices is based on functional and performance re-quirements, such devices as high stiffness, low Inertia, large workspace; highmanipulability, low friction, high transparency, as well as they must satisfycost constraints. These requirements are conflicting which ask for a systematicusage of kinematic, dynamic, stiffness, friction and control models.

The methodology is based on V-model along with a stage-gate model, andevaluated through a test case where a haptic device, based on a Stewart plat-form, is designed and realized as a surgical simulator for training on hardtissues including bone and teeth. The objective of the presented approachis to minimize technical risks by dividing the product development processfrom abstract ideas to realization of the physical product into a sequence ofdecomposition stages with go/no-go/wait decision at each stage.

In the model-based and simulation-driven design approach, the design startsfrom an abstract top-level model, which is extended via stepwise refinementsthrough a design space exploration phase into a complete realization of thesystem. The transparency performance of the device was improved with afriction compensation strategy. We can conclude both from simulations andresults from physical experiments that the performance of the designed hapticdevice satisfies the stated requirements. Experiences from this case indicatethat the methodology is capable of supporting effective and efficient develop-ment of high performing haptic devices.

39

40 CHAPTER 5. SUMMARY OF APPENDED PAPERS

5.2 Paper B: a stiffness modeling methodology forsimulation-driven design of haptic devices

This work proposes a new methodology for creating an analytical and com-pact model for quasi-static stiffness analysis of a haptic device, which considersthe stiffness of an actuation system, flexible links and passive joints. For themodeling of passive joints, a Hertzian contact model is introduced for bothspherical and universal joints, and a simply supported beam model for univer-sal joints. The verification and validation process is presented as a systematicguideline to evaluate the stiffness parameters using both parametric FEMmodeling and physical experiments. Pre-loading has been used to considerthe clearances and possible assembling errors during manufacturing. A hapticdevice based on a Stewart platform is used as an example to illustrate themodeling and validation methodology.

The approach used in this methodology is to start with a simplified an-alytical model, which is verified by a simplified FEM model in an iterativeway until these two models give the same results. Thereafter, a simplifiedphysical experiment is made to validate the analytical model. If the differ-ence between the analytical and experimental results is not acceptable, thatis it does not validate the analytical model, a more detailed analytical modelmust be developed. For the detailed analytical model the passive joints andactuation system are also included. This is then verified with a detailed FEMmodel with a corresponding detailing level in the same way as for the simpli-fied model. Thereafter, this detailed analytical model is validated by meansof a detailed physical experiment. After validating the proposed model, asensitivity analysis is performed to map the variation of static stiffness inthe workspace. In these maps, the stiffness is visualized as a function of thegeneralized coordinates of the workspace.

5.3 Paper C: an optimization approach towards a robust designof 6-DOF haptic devices

In this paper a robust design optimization approach is proposed for the designof haptic devices. The objective is to find optimal numerical values for a setof design parameters of haptic devices, which enable to its performance lesssensitive to parameters variations and allows active trade-off decisions betweendevice-performance variations and parameter tolerances.

The presented approach is verified through a test case to design a 6-DOFhaptic device based on TAU configuration to maximize its kinematic, dy-namic, and kinetostatic performance indices of a 6-DOF haptic device whileminimizing its sensitivity to variations in manufacturing tolerances. Theseindices were defined through kinematic, dynamic, and stiffness models. Dur-ing design evaluation global indices were defined such as workspace volume,

5.4. PAPER D: EVALUATION OF FRICTION MODELS FOR HAPTIC DEVICES 41

quasi-static torque requirements of the actuators, kinematic isotropy, dynamicisotropy, stiffness isotropy, and natural frequency of the device.

A new approach used to normalize the performance indices to find theminimum and maximum performance indices within the bounds of the designparameters using Monte Carlo Simulations (MCS). For design optimization,a new hybrid optimization approach was used, that combines GA and MCSincluding a robustness criterion for the 6-sigma principle. The effectivenessof the proposed approach is verified through a comparative analysis with adeterministic design optimization, it has been shown that the proposed ap-proach can find the numerical values of the design parameters that are bothoptimal and robust (i.e., less sensitive to variation and thus to uncertainties inthe design parameters). Further on a set of optimum tolerances is determinedthat minimizes manufacturing cost and also satisfies the allowed variations inthe performance indices. The presented approach can thus enable the designerto evaluate trade-offs between allowed performance variations and tolerancecost.

5.4 Paper D: evaluation of friction models for haptic devices

In this paper the classical old models like Dahl, LuGre and more recent ad-vanced models like Generalized Maxwell slip model (GMS), the smooth Gen-eralized Maxwell slip model (SGMS) and the Differential Algebraic Multistate(DAM) friction models have been investigated to determine how these modelsare suited for performance simulation and control of a 6-DOF haptic device.

To evaluate the friction models, we use criterion based on: fidelity to pre-dict realistic friction phenomena, easiness to implementation and easiness toestimate model parameters for online parameter estimation. These modelsare evaluated both numerically and experimentally with an existing 6-DOFhaptic device that is based on a Stewart platform.

In order to evaluate how well these models can compensate for friction, amodel-based feedback friction compensation strategy along with a PID con-troller was used for position tracking accuracy. The accuracies of the fric-tion compensation models were examined separately for both low-velocity andhigh-velocity motions of the system. To evaluate these models, we used criteriabased on fidelity to predict realistic friction phenomena, ease of implementa-tion, computational efficiency and ease of estimating the model parameters.Experimental results show that friction compensated with GMS, S-GMS andDAM models give better accuracy in terms of standard deviation, Root MeanSquare Error, and maximum error between a reference and measured trajec-tory. Based on the criteria of fidelity, ease of implementation and ease ofestimating model parameters, the S-GMS model, which represents a smoothtransition between sliding and pre-sliding regimes through an analytical set ofdifferential equations, is suggested.

42 CHAPTER 5. SUMMARY OF APPENDED PAPERS

5.5 Paper E: model-based control strategy for 6-DOF hapticdevices

In this work, a new model-based control strategy is proposed for improvingtransparency and fidelity of force/torque feedback from a 6-DOF haptic deviceused for hard tissues interactions. The presented approach is evaluated ex-perimentally with an existing 6-DOF haptic device that is based on a Stewartplatform. The model-based design is driven through dynamic, stiffness andfriction models in order to compensate inertia, gravitational forces, elastic de-formations, and friction effects in the active joints. Parameters of the 6-DOFhaptic device were identified using a new approach. The effectiveness of theapproach is verified in free space and while interacting with the virtual object.The comparison is made between force/torque generated and force/torque es-timated using sensor-less force/torque estimation using recursive least squaresestimation with exponential forgetting. The experimental results show thatthe proposed approach can increase transparency of the haptic devices, therebyincreasing the realism of the simulation.

Chapter 6

Discussion, conclusions and futurework

This chapter give a brief discussion, conclusions and contributions of this workalong with an outline of promising future research directions enabled by thiswork.

6.1 Discussion

The training of a surgeon is a complex and multi-dimensional process partic-ularly in the case of hard tissues like bone and dental procedures. Simulatorswill create new training opportunities for surgical procedures, which are im-possible to train with traditional methods. Moving training for surgical proce-dures from the operating room to a simulator will offer considerable economicadvantages. In this context, a haptic device is one of the key components toenable realistic force/torque feedback in a virtual environment. Developmentof a haptic device for this type of application is a challenging task because ofthe multi-criteria design consideration and a successful design requires that itis able to provide a sufficient level of realistic feedback.

The focus of this thesis work has been to provide a realistic force/torquefeedback from the virtual environment via a haptic device. This has includeddevelopment of methods on a high level such as the model-based and simu-lation driven design approach as well on a more specific level such as frictionestimation and compensation.

During this work, which is dealing with design and development of physicalproducts it has been a great advantage to have prototypes to use as test casesfor model development as well as for experimental testing. In this work twophysical 6-DOF haptic devices have been used for development and testing oftheories and methods. Having access to two prototypes, being developed fromthe same set of requirements has also been used to show and claim that theresults developed in this work can be generalized.

43

44 CHAPTER 6. DISCUSSION, CONCLUSIONS AND FUTURE WORK

6.2 Conclusions

The overall research objective in this thesis is to find a methodology that cansupport the development of haptic devices with realistic force/torque feedbackfrom the virtual environment. This objective is addressed through the devel-opment of a model-based and simulation-driven methodology for the design ofhaptic devices (Paper A). This overall objective was divided into more specificresearch questions, which are answered below.

• Q1: How to accurately estimate stiffness of a 6-DOF haptic device throughan analytical model, which is compact and can be used in real-time con-trol?

A stiffness modeling methodology has been developed to create an ana-lytically compact and computationally efficient (Paper C) stiffness modelof a 6-DOF haptic device. This model is obtained in a step-by-step mod-eling manner starting with a simplified model considering the stiffness ofthe compliant element within the system. It also considers the stiffnessof the actuation system, linear guideways, proximal links, and passivejoints. The force acting at the TCP of the platform is decomposed intoindividual link forces, thereafter individual link deflections are computedfrom the link stiffness properties. Finally, all these displacements aretransformed and added to obtain the final global compliance matrix.

• Q2: How to find a robust optimum design of haptic devices, whose per-formances are less sensitive to variations in manufacturing tolerances?

A hybrid design optimization approach has been developed, which com-bines a genetic algorithm (GA) and Monte Carlo simulation (MCS), andcan find the numerical values of the design parameters that are both op-timal and robust (i.e., less sensitive to variation and thus to uncertaintiesin the design parameters) (Paper B).

• Q3: How to estimate friction in haptic devices that can mimic realisticbehavior and be used in real-time control?

Different friction models like classical old models such as Dahl, LuGreand more recently advanced models like Generalized Maxwell slip model(GMS), the smooth Generalized Maxwell slip model (S-GMS) and theDifferential Algebraic Multistate (DAM) friction models have been in-vestigated to determine if these models are suited for performance simu-lation and control of a 6-DOF haptic device (Paper D). These models areevaluated both numerically and experimentally with an existing 6-DOFhaptic device that is based on a Stewart platform. Experimental resultsshow that friction compensated with GMS, S-GMS and DAM modelsgive better accuracy in terms of standard deviation, Root Mean SquaredError, and maximum error between a reference and measured trajectory.

6.3. CONTRIBUTIONS 45

Based on the criteria of fidelity, ease of implementation and ease of esti-mating model parameters, the S-GMS model, which represents a smoothtransition between sliding and pre-sliding regimes through an analyticalset of differential equations, is suggested.

• Q4: How to evaluate force/torque capability of a haptic device without aforce/torque sensor?

The developed approach evaluates the force/torque capability of the de-vice by estimating a vector of force/torque of six degrees in x-, y- andz-directions using a linearized form of a dynamic equation along withleast squares estimation with exponential forgetting.

• Q5: How to develop a control strategy that can be used to compensate forunwanted effects and maximize the transparency of a haptic device?

A model-based control strategy has been developed that compensatesfor the natural dynamics of the device, friction between joints, gravityof platform, and elastic deformation (Paper E). The developed approachis based on a closed-loop impedance control (CLIC) with model-basedcompensation and an estimated force/torque feedback.

6.3 Contributions

This thesis makes the following key contributions to the design of haptic de-vices used for interaction with hard tissues.

• A model-based and simulation-driven design methodology is developed,where one starts from an abstract, top-level model that is extended viastepwise refinements and design space exploration into a detailed andintegrated systems model that can be physically realized.

• A methodology for the development of an analytical stiffness model of a6-DOF haptic device, which is compact, can capture realistic stiffness andis computationally efficient to be used in real-time stiffness compensationof elastic deformation.

• An optimization approach is presented towards a robust design of 6-DOFhaptic devices that maximize the kinematic, dynamic, and kineto-staticperformances of a 6-DOF haptic device, while minimizing its sensitivityto variations in manufacturing tolerances.

• A cost-effective approach is presented for force/torque feedback controlusing force/torque estimated through a recursive least squares estimation.

• A model-based control strategy has been proposed based on a CLIC withmodel-based compensation in order to improve the transparency and fi-delity of force/torque feedback of a 6-DOF haptic device.

46 CHAPTER 6. DISCUSSION, CONCLUSIONS AND FUTURE WORK

6.4 Future work

Future research can be directed into a number of interesting areas, which canextend this thesis work, for example:

• Stability of control system.

• Evaluation of velocity and acceleration capability of the TAU haptic de-vice.

• Face validation study of 6-DOF TAU haptic device.

• Backlash compensation in joints.

• Compensation for passive joints friction.

• Evaluation of the proposed control strategy in Sim-Mechanics.

• Verification and validation of the proposed control strategy through aforce/torque sensor.

References

[1] M. A. Srinivasan and C. Basdogan, “Haptics in virtual environments:Taxonomy, research status, and challenges,” Computers & Graphics,vol. 21, no. 4, pp. 393–404, 1997.

[2] H. Flemmer, “Control design and performance analysis of force reflectiveteleoperators-a passivity based approach,” PhD Thesis, KTH, Stock-holm, 2004.

[3] K. Suleman, “Design and optimization of parallel haptic devices Designmethodology and experimental evaluation,” PhD Thesis, KTH, Stock-holm, 2012.

[4] M. G. Eriksson, “Haptic milling simulation in six degrees-of-freedomwith application to surgery in stiff tissue,” PhD Thesis, KTH, Stock-holm, 2012.

[5] M. Samadimaleh, “Design and implementation of control system for 6-dof haptic device,” Master of Science Thesis, KTH, Stockholm, Sweden,2011.

[6] C. M. Salisbury, “Haptic hardware: Evaluation, design and placement,”PhD Thesis, Stanford University, 2010.

[7] C. Luciano, P. Banerjee, and T. DeFanti, “Haptics-based virtual realityperiodontal training simulator,” Virtual reality, vol. 13, no. 2, pp. 69–85,2009.

[8] Digital dentures: A complete game changer for local dental labs.[Online]. Available: http://www.capedental.com/dental-blog/2011/11/digital-dentures-a-complete-game-changer-for-local-dental-labs/

[9] L.-F. Lee, “Analysis and design optimization of in-parallel haptic de-vices,” PhD Thesis, University at Buffalo, Buffalo, 2010.

[10] S. Khan, K. Andersson, and J. Wikander, “A design approach for a new6-dof haptic device based on parallel kinematics,” in ICM 2009. IEEEInternational Conference on Mechatronics, 2009, april 2009, pp. 1 –6.

47

http://www.capedental.com/dental-blog/2011/11/digital-dentures-a-complete-game-changer-for-local-dental-labs/

http://www.capedental.com/dental-blog/2011/11/digital-dentures-a-complete-game-changer-for-local-dental-labs/

48 REFERENCES

[11] G. Gogu, “Fully-isotropic parallel mechanisms-an innovative concept forhaptic devices,” in Product Engineering. Springer, 2008, pp. 169–194.

[12] E. Samur, Springer Series on Touch and Haptic Systems: PerformanceMetrics for Haptic Interfaces. Springer, 2012.

[13] T. H. Massie and J. K. Salisbury, “The phantom haptic interface: Adevice for probing virtual objects,” in Proceedings of the ASME winterannual meeting, symposium on haptic interfaces for virtual environmentand teleoperator systems, vol. 55, no. 1. IOS Press, 1994, pp. 295–300.

[14] Force dimension. [Online]. Available: http://www.forcedimension.com

[15] S. Martin and N. Hillier, “Characterisation of the novint falcon hapticdevice for application as a robot manipulator,” in Australasian Confer-ence on Robotics and Automation (ACRA), 2009, pp. 291–292.

[16] R. Q. Van der Linde, P. Lammertse, E. Frederiksen, and B. Ruiter,“The hapticmaster, a new high-performance haptic interface,” in Proc.Eurohaptics, 2002, pp. 1–5.

[17] Haption. [Online]. Available: http://www.haption.com

[18] Mpb technologies. [Online]. Available: http://www.mpb-technologies.ca/

[19] Quanser. [Online]. Available: http://www.quanser.com/

[20] Mentice sa. [Online]. Available: http://www.mentice.com/

[21] Cae healthcare: Endoscopyvr. [Online]. Available: http://www.cae.com

[22] J. H. Lee, K. S. Eom, and I. I. Suh, “Design of a new 6-dof parallel hapticdevice,” in Robotics and Automation, 2001. Proceedings 2001 ICRA.IEEE International Conference on, vol. 1. IEEE, 2001, pp. 886–891.

[23] H. K. Fathy, J. A. Reyer, P. Y. Papalambros, and A. Ulsov, “On thecoupling between the plant and controller optimization problems,” inProceedings of American Control Conference, 2001, vol. 3. IEEE, 2001,pp. 1864–1869.

[24] S. Khan and K. Andersson, “A design methodology for haptic devices,”in Proceedings of the 18th International Conference on Engineering De-sign (ICED11), Vol. 4, 2011, pp. 288–298.

[25] S. Khan, K. Andersson, and J. Wikander, “Investigation of parallel kine-matic mechanism structures for haptic devices,” in 2nd Nordic Confer-ence on Product Lifecycle Management–NordPLM’09, Gothenburg Jan-uary 2009., 2009.

http://www.forcedimension.com

http://www.haption.com

http://www.mpb-technologies.ca/

http://www.mpb-technologies.ca/

http://www.quanser.com/

http://www.mentice.com/

http://www.cae.com

REFERENCES 49

[26] VDI, “Design methodology for mechatronic systems (VDI 2206),” VDI,Tech. Rep., 2004.

[27] R. G. Cooper, Winning at new products: accelerat-ing the process from idea to launch. Basic Books, 2001,vol. 2nd ed. [Online]. Available: http://www.amazon.com/Winning-New-Products-Accelerating-Process/dp/0738204633

[28] R. G. Cooper and S. J. Edgett, “Best practices in the idea-to-launch process and its governance,” ResearchTechnology Man-agement, vol. 55, no. 2, pp. 43–54, 2012. [Online]. Avail-able: http://openurl.ingenta.com/content/xref?genre=article&issn=0895-6308&volume=55&issue=2&spage=43

[29] Y. Lou, G. Liu, and Z. Li, “Randomized optimal design of parallel ma-nipulators,” Automation Science and Engineering, IEEE Transactionson, vol. 5, no. 2, pp. 223–233, 2008.

[30] P. Zhang, Z. Yao, and Z. Du, “Global Performance Index System forKinematic Optimization of Robotic Mechanism,” Journal of MechanicalDesign, vol. 136, no. 3, p. 31001, Dec. 2013. [Online]. Available:http://dx.doi.org/10.1115/1.4026031

[31] X.-J. Liu, Z.-L. Jin, and F. Gao, “Optimum design of 3-dof sphericalparallel manipulators with respect to the conditioning and stiffness in-dices,” Mechanism and Machine Theory, vol. 35, no. 9, pp. 1257–1267,2000.

[32] Y. Su, B. Duan, and C. Zheng, “Genetic design of kinematically op-timal fine tuning stewart platform for large spherical radio telescope,”Mechatronics, vol. 11, no. 7, pp. 821–835, 2001.

[33] Y. Zhang and Y. Yao, “Kinematic optimal design of 6-ups parallel ma-nipulator,” in Mechatronics and Automation, Proceedings of the 2006IEEE International Conference on. IEEE, 2006, pp. 2341–2345.

[34] E. Ottaviano and M. Ceccarelli, “Optimal design of capaman (cassinoparallel manipulator) with a specified orientation workspace,” Robotica,vol. 20, no. 02, pp. 159–166, 2002.

[35] R. Boudreau and C. M. Gosselin, “La synthese d’une plate-forme degough-stewart pour un espace atteignable prescrit,” Mechanism and ma-chine theory, vol. 36, no. 3, pp. 327–342, 2001.

[36] K. Miller, “Optimal design and modeling of spatial parallel manipula-tors,” The International Journal of Robotics Research, vol. 23, no. 2, pp.127–140, 2004.

http://www.amazon.com/Winning-New-Products-Accelerating-Process/dp/0738204633

http://www.amazon.com/Winning-New-Products-Accelerating-Process/dp/0738204633

http://openurl.ingenta.com/content/xref?genre=article&issn=0895-6308&volume=55&issue=2&spage=43

http://openurl.ingenta.com/content/xref?genre=article&issn=0895-6308&volume=55&issue=2&spage=43

http://dx.doi.org/10.1115/1.4026031

50 REFERENCES

[37] F. Hao and J.-P. Merlet, “Multi-criteria optimal design of parallelmanipulators based on interval analysis,” Mechanism and MachineTheory, vol. 40, no. 2, pp. 157 – 171, 2005. [Online]. Available:http://www.sciencedirect.com/science/article/pii/S0094114X04001211

[38] X.-J. Liu, J. Wang, K.-K. Oh, and J. Kim, “A new approach to thedesign of a delta robot with a desired workspace,” Journal of Intelligentand Robotic Systems, vol. 39, no. 2, pp. 209–225, 2004.

[39] Zbigniew Michalewicz, Genetic algorithms + data structures = Evolutionprograms. Springer, 1996.

[40] S.-D. Stan, V. Maties, R. Balan, C. Rusu, and S. Besoiu, “Optimallink design of a six degree of freedom micro parallel robot based onworkspace analysis,” in Advanced Motion Control, 2008. AMC ’08. 10thIEEE International Workshop on, march 2008, pp. 637 –642.

[41] Y.-K. Hwang, J.-W. Yoon, and J.-H. Ryu, “The optimum design of a6-dof parallel manipulator with large orientation workspace,” in SICE-ICASE, 2006. International Joint Conference, oct. 2006, pp. 1255 –1259.

[42] R. Ur-Rehman, S. Caro, D. Chablat, and P. Wenger, “Multi-objectivepath placement optimization of parallel kinematics machines based onenergy consumption, shaking forces and maximum actuator torques: Ap-plication to the orthoglide,” Mechanism and Machine Theory, vol. 45,no. 8, pp. 1125–1141, 2010.

[43] S. Khan, K. Andersson, and J. Wikander, “Optimal design of a 6-dofhaptic device,” in Mechatronics (ICM), 2011 IEEE International Con-ference on. IEEE, 2011, pp. 713–718.

[44] C. Zang, M. Friswell, and J. Mottershead, “A review of robust opti-mal design and its application in dynamics,” Computers & Structures,vol. 83, no. 4, pp. 315–326, 2005.

[45] A. Parkinson, C. Sorensen, and N. Pourhassan, “A general approach forrobust optimal design,” Journal of Mechanical Design, vol. 115, p. 74,1993.

[46] X. Du, P. K. Venigella, and D. Liu, “Robust mechanism synthesis withrandom and interval variables,” Mechanism and Machine theory, vol. 44,no. 7, pp. 1321–1337, 2009.

[47] S. Caro, F. Bennis, and P. Wenger, “Tolerance synthesis of mechanisms:a robust design approach,” Journal of Mechanical Design, vol. 127, p. 86,2005.

http://www.sciencedirect.com/science/article/pii/S0094114X04001211

REFERENCES 51

[48] H.-G. Beyer and B. Sendhoff, “Robust optimization–a comprehensivesurvey,” Computer methods in applied mechanics and engineering, vol.196, no. 33, pp. 3190–3218, 2007.

[49] C. C. Zhang and H.-P. B. Wang, “Robust design of assembly and ma-chining tolerance allocations,” IIE transactions, vol. 30, no. 1, pp. 17–29,1997.

[50] M. Li, S. Azarm, and V. Aute, “A multi-objective genetic algorithm forrobust design optimization,” in Proceedings of the 2005 conference onGenetic and evolutionary computation. ACM, 2005, pp. 771–778.

[51] L.-F. Lee, M. S. Narayanan, F. Mendel, V. N. Krovi, and P. Karam,“Kinematics analysis of in-parallel 5 dof haptic device,” in AdvancedIntelligent Mechatronics (AIM), 2010 IEEE/ASME International Con-ference on. IEEE, 2010, pp. 237–241.

[52] C. Cruz-Valverde, O. A. Dominguez-Ramirez, E. R. Ponce-de Leon-Sanchez, I. Trejo-Mota, and G. Sepulveda-Cervantes, “Kinematic anddynamic modeling of the phantom premium 1.0 haptic device: Experi-mental validation,” in Electronics, Robotics and Automotive MechanicsConference (CERMA), 2010. IEEE, 2010, pp. 494–501.

[53] H. Cui, Z. Zhu, Z. Gan, and T. Brogardh, “Kinematic analysis anderror modeling of tau parallel robot,” Robotics and Computer-IntegratedManufacturing, vol. 21, no. 6, pp. 497–505, 2005.

[54] Z. Zhu, J. Li, Z. Gan, and H. Zhang, “Kinematic and dynamic modellingfor real-time control of tau parallel robot,” Mechanism and machinetheory, vol. 40, no. 9, pp. 1051–1067, 2005.

[55] B. Dasgupta and T. Mruthyunjaya, “A newton-euler formulation for theinverse dynamics of the stewart platform manipulator,” Mechanism andMachine Theory, vol. 33, no. 8, pp. 1135–1152, 1998.

[56] H. Abdellatif and B. Heimann, “Computational efficient inverse dynam-ics of 6-dof fully parallel manipulators by using the lagrangian formal-ism,” Mechanism and Machine Theory, vol. 44, no. 1, pp. 192–207, 2009.

[57] L.-W. Tsai, “Solving the inverse dynamics of a stewart-gough manipula-tor by the principle of virtual work,” Journal of Mechanical design, vol.122, p. 3, 2000.

[58] A. Mendes Lopes and F. Almeida, “The generalized momentum ap-proach to the dynamic modeling of a 6-dof parallel manipulator,” Multi-body System Dynamics, vol. 21, no. 2, pp. 123–146, 2009.

52 REFERENCES

[59] C. C. Nguyen and F. J. Pooran, “Dynamic analysis of a 6 dof ckcm robotend-effector for dual-arm telerobot systems,” Robotics and AutonomousSystems, vol. 5, no. 4, pp. 377–394, 1989.

[60] M. Ahmadi, M. Dehghani, M. Eghtesad, and A. R. Khayatian, “Inversedynamics of hexa parallel robot using lagrangian dynamics formulation,”in Intelligent Engineering Systems, 2008. INES 2008. International Con-ference on. IEEE, 2008, pp. 145–149.

[61] A. Pashkevich, A. Klimchik, and D. Chablat, “Nonlinear effects in stiff-ness modeling of robotic manipulators,” arXiv preprint arXiv:0912.1420,2009.

[62] C. Gosselin, “Stiffness mapping for parallel manipulators,” IEEE Trans-actions on Robotics and Automation,, vol. 6, pp. 321–342, jun 1990.

[63] C. Gosselin and D. Zhang, “Stiffness analysis of parallel mechanismsusing a lumped model,” Int. J. of Robotics and Automation,, vol. 17,pp. 17–27, April 2002.

[64] C. M. D. Zhang, F. Xi and S. Lang, “Analysis of parallel kinematicmachine with kinetostatic modeling method,” Robotics and Computer-Integrated Manufacturing,, vol. 20, pp. 151– 165, April 2004.

[65] D. Zhang and C. M. Gosselin, “Kinetostatic analysis and design opti-mization of the tricept machine tool family,” Journal of manufacturingscience and engineering, vol. 124, no. 3, pp. 725–733, 2002.

[66] Y. Wang, T. Huang, X. Zhao, J. Mei, D. Chetwynd, and S. Hu, “Fi-nite element analysis and comparison of two hybrid robots-the triceptand the trivariant,” in Intelligent Robots and Systems, 2006 IEEE/RSJInternational Conference on, oct. 2006, pp. 490 –495.

[67] Y. Yun and Y. Li, “Comparison of two kinds of large displacementprecision parallel mechanisms for micro/nano positioning applications,”in Robotics, Automation and Mechatronics, 2008 IEEE Conference on,sept. 2008, pp. 284 –289.

[68] R. S. Goncalves and J. C. M. Carvalho, “Stiffness analysis of parallelmanipulator using matrix structural analysis,” in Proceedings of EU-COMES 08. Springer, 2009, pp. 255–262.

[69] D. Deblaise, X. Hernot, and P. Maurine, “A systematic analyticalmethod for pkm stiffness matrix calculation,” in Robotics and Automa-tion, 2006. ICRA 2006. Proceedings 2006 IEEE International Confer-ence on, may 2006, pp. 4213 –4219.

[70] Przemieniecki, J.S, Theory of Matrix Structural Analysis. New York:Dover Publications, Inc, 1985.

REFERENCES 53

[71] W. Dong, Z. Du, and L. Sun, “Stiffness influence atlases of a novel flexurehinge-based parallel mechanism with large workspace,” in IntelligentRobots and Systems, 2005. (IROS 2005). 2005 IEEE/RSJ InternationalConference on, aug. 2005, pp. 856 – 861.

[72] M. Uchiyama, Y. Tsumaki, and W.-K. Yoon, “Design of a compact6-dof haptic device to use parallel mechanisms,” in Robotics Research.Springer, 2007, pp. 145–162.

[73] C. Pinto, J. Corral, O. Altuzarra, and A. Hernandez, “A methodologyfor static stiffness mapping in lower mobility parallel manipulators withdecoupled motions,” Robotica, vol. 28, no. 5, pp. 719–735, 2010.

[74] M. Ceccarelli and G. Carbone, “Numerical and experimental analysisof the stiffness performances of parallel manipulators,” in 2nd Inter-national Colloquium of the Collaborative Research Centre 562,Braun-schweig, 2005.

[75] ANSI/RIA R15.05-1-1990 (R1999), Evaluation of point-topoint andstatic performance characteristics of industrial robots and robot systems,1999.

[76] ISO 9283:1998, Manipulating industrial robots-performance criteria andrelated test methods, 2003.

[77] C. R. Carignan and K. R. Cleary, “Closed-loop force control for hapticsimulation of virtual environments,” Haptics-e, vol. 1, no. 2, pp. 1–14,2000.

[78] K. Wen, D. Necsulescu, and J. Sasiadek, “Haptic force control based onimpedance/admittance control aided by visual feedback,” MultimediaTools and Applications, vol. 37, no. 1, pp. 39–52, 2008.

[79] O. Baser, E. I. Konukseven, and H. Gurocak, “Transparency improve-ment in haptic devices with a torque compensator using motor cur-rent,” in Haptics: Perception, Devices, Mobility, and Communication.Springer, 2012, pp. 37–46.

[80] A. Frisoli, E. Sotgiu, C. Avizzano, D. Checcacci, and M. Bergamasco,“Force-based impedance control of a haptic master system for teleoper-ation,” Sensor Review, vol. 24, no. 1, pp. 42–50, 2004.

[81] M. Ueberle and M. Buss, “Design, control, and evaluation of a new 6dof haptic device,” in Intelligent Robots and Systems, 2002. IEEE/RSJInternational Conference on, vol. 3. IEEE, 2002, pp. 2949–2954.

[82] M. Honegger, A. Codourey, and E. Burdet, “Adaptive control of thehexaglide, a 6 dof parallel manipulator,” in Robotics and Automation,

54 REFERENCES

1997. Proceedings., 1997 IEEE International Conference on, vol. 1.IEEE, 1997, pp. 543–548.

[83] A. Ahmad, K. Andersson, U. Sellgren, and M. Boegli, “Evaluation offriction models for haptic devices,” in In proceeding of: DSCC 2013ASME Dynamic Systems and Control Conference October 21-23, 2013,Stanford university, Munger Centre, Palo Alto, CA, USA.

[84] M. Boegli, T. Laet, J. De Schutter, and J. Swevers, “A smoothed gmsfriction model suited for gradient-based friction state and parameter es-timation,” Mechatronics, IEEE/ASME Transactions on, vol. PP, no. 99,pp. 1–10, 2013.

[85] J. Wang, H. Zhang, and T. Fuhlbrigge, “Improving machining accuracywith robot deformation compensation,” in Intelligent Robots and Sys-tems, 2009. IROS 2009. IEEE/RSJ International Conference on. IEEE,2009, pp. 3826–3831.

[86] M. J. Puerto, E. Sanchez, and J. J. Gil, “Control strategies applied tokinesthetic haptic devices,” in Robotic Intelligence in InformationallyStructured Space, 2009. RIISS’09. IEEE Workshop on. IEEE, 2009,pp. 137–144.

[87] R. J. Adams and B. Hannaford, “Control law design for haptic interfacesto virtual reality,” Control Systems Technology, IEEE Transactions on,vol. 10, no. 1, pp. 3–13, 2002.

[88] J. J. Gil and E. Sanchez, “Control algorithms for haptic interaction andmodifying the dynamical behavior of the interface,” in 2nd internationalconference on Enactive Interfaces, 2005, pp. 17–18.

[89] M. Van Damme, P. Beyl, B. Vanderborght, V. Grosu, R. Van Ham,I. Vanderniepen, A. Matthys, and D. Lefeber, “Estimating robot end-effector force from noisy actuator torque measurements,” in Robotics andAutomation (ICRA), 2011 IEEE International Conference on. IEEE,2011, pp. 1108–1113.

[90] U. Sellgren, “Simulation-driven design: motives, means, and opportuni-ties,” PhD Thesis, KTH, Stockholm, 1999.

[91] J. F. Broenink, Y. Ni, and M. A. Groothuis, “On model-driven design ofrobot software using co-simulation,” SIMPAR, Workshop on SimulationTechnologies in the Robot Development Process, 2010.

[92] Rivin, E.I, Stiffness and Damping in Mechanical Design. New York:Marcel Dekker Inc, 1999.

[93] Waterloo maple inc. (maplesoft). [Online]. Available: www.maplesoft.com

www.maplesoft.com

www.maplesoft.com

REFERENCES 55

[94] Ansys, inc. [Online]. Available: http://www.ansys.com

[95] M. Boegli, T. De Laet, J. De Schutter, and J. Swevers, “A smoothedgms friction model suited for gradient-based friction state estimation,”in American Control Conference (ACC), 2012, June, pp. 2627–2632.

[96] S. T. Mclunkin, “Transparency improvement for haptic interfaces,” PhDThesis, Rice University, Houstan, Texas, 1999.

[97] A. Ahmad, K. Andersson, and U. Sellgren, “An optimization approachtowards a robust design of 6-DOF haptic devices,” submitted to Journalof Mechanical Design, Jan. 2014.

[98] Maxon motor ag. [Online]. Available: http://maxonmotor.com

[99] B. R. Hopkins and R. L. Williams Ii, “Kinematics, design and control ofthe 6-psu platform,” Industrial Robot: An International Journal, vol. 29,no. 5, pp. 443–451, 2002.

[100] Geomagic. [Online]. Available: http://www.geomagic.com

http://www.ansys.com

http://maxonmotor.com

http://www.geomagic.com

A model-based and simulation-driven methodology for design of haptic devices

Aftab Ahmad, Kjell Andersson and Ulf Sellgren

Journal of Mechatronics, Jan 2014.

A

A stiffness modeling methodology for simulation-driven design of haptic devices

Aftab Ahmad, Kjell Andersson, Ulf Sellgren and Suleman Khan

Journal of Engineering with Computers, Dec, 2012.

B

An optimization approach towards a robust design of 6-DOF haptic devices

Aftab Ahmad, Kjell Andersson and Ulf Sellgren

Submitted to Journal of Mechanical Design, Jan 2014.

C

Evaluation of friction models for haptic devices

Aftab Ahmad, Kjell Andersson, Ulf Sellgren, and Max Boegli

In proceeding of DSCC 2013 ASME Dynamic Systems and Control Conference

October 21-23, 2013, Stanford university, Munger Centre, Palo Alto, CA, USA.

D

Model-based control strategy for 6-DOF haptic devices

Aftab Ahmad, Kjell Andersson, and Ulf Sellgren

To be submitted to IEEE Transaction on Robotics.

E

e ective development of haptic devices using a model-based and...

Documents