The design and making of a humanoid robotic hand

presented by

Tian Li

Research associate

Supervisor’s Name: Prof. Nadia Magnenat Thalmann，Prof. Daniel Thalmann & Prof. Jianmin Zheng

Project 2: Mixed Society of People, Virtual Humans, and Social Robots in Telepresence

BeingTogether Centre

Date of Presentation : 10 October 2017

BTC Task 2.4: Realistic Object

Manipulation by a Social Robot in a

Social Context• Integration in a social robot of an articulated hand for

grasping different types of objects.

• Vision-based detection, recognition and localisation of different

types of objects

• Determination of a strategy for reaching and grasping the object

in a human-like way

2

Target

• Static - Looks like human hand

Solution: Use silicon artificial skin

• Dynamic - Moves like human hand

Solution: design and making of a

humanoid robotic hand

Criteria: Dimensions, Degree of

freedom(DOF), Joints structure and

motion range, move accuracy and

speed.

• Other requirement – light weight, low

cost, easy to make, easy to control

Related works

[1] Shadow hand: https://www.shadowrobot.com/products/dexterous-hand

[2] Zhe Xu and Emanuel Todorov, “Design of a highly biomimetic anthropomorphic robotic hand towards

artificial limb regeneration” 2016 IEEE International Conference on Robotics and Automation (ICRA)

[3] Bebionic hand: http://bebionic.com/the_hand

40 actuators(20 DOFs) Highly biomimetic Robust and compact

Shadow Hand[1] Xu Zhe hand[2] Bebionic hand[3]

How does hand work – mechanism

http://nimblevr.com/latest/doc/handModel.html

Metacarpal-phalangeal joint (MCP) Proximal interphalangeal joint (PIP)

Distal interphalangeal joint (DIP) Carpal-metacarpal joint (CMC)

Metacarpal-phalangeal joint (MCP) Interphalangeal joint (IP)

15 movable joints, 20DOFs, 27 bones

Mechanism design - interlock

https://en.wikipedia.org/wiki/File:Joint.svg

How does hand work – actuation “Fingers do not contain muscles (other than arrector pili). The muscles that

move the finger joints are in the palm and forearm. The long tendons that

deliver motion from the forearm muscles may be observed to move under the

skin at the wrist and on the back of the hand.” (From Finger - Wikipedia)

https://answersingenesis.org/human-body/our-index-finger/

Actuation design – cable-driven

https://answersingenesis.org/human-body/our-index-finger/

• Total DOFs reduced from 20 to 6• One cable control 3 joints• Thumb have 2 cables

Nadine hand conclusion • Humanlike mechanical design – Dimensions, joints, motion

factors

• Simple actuation method - Cable driven, six actuators

• 3D printable – Low cost, fast, light, motion range control

• Single model with interlock joint – No assemble

required

• Artificial skin compatible

Nadine hand - originality

[1]Xu, Zhe, Vikash Kumar, and Emanuel

Todorov. "A low-cost and modular, 20-DOF

anthropomorphic robotic hand: design,

actuation and modeling." Humanoid Robots

(Humanoids), 2013 13th IEEE-RAS

International Conference on. IEEE, 2013.

[2] Zhe Xu and Emanuel Todorov, “Design of a highly

biomimetic anthropomorphic robotic hand towards

artificial limb regeneration” 2016 IEEE International

Conference on Robotics and Automation (ICRA)

[3] http://inmoov.fr/

Single

interlocked

model

Servo bed design

• HS-5070MH 23.6 x 11.6 x 29mm, 12.5g, 3.8 kg.cm

• HK15298B 117 x 20 x 80mm 66g, 20 kg.cm

Magnenat Thalmann, N., Tian, L., & Yao, F. (2017). Nadine: A Social Robot that Can Localize

Objects and Grasp Them in a Human Way Frontiers in Electronic Technologies (pp. 1-23):

Springer.

ComparisonsHand Developer Mass(g) Size (length x

width x

thickness, mm)

Number

of

Joints

DOF Number

of

Actuators

Actuatio

n

Method

Material

Nadine

hand

NTU IMI 205 147 x 70 x

13(without

artificial skin)

15 6 6 DC

Motor -

Tendons

3D print ABS

Shadow

Hand

SHADOW

ROBOT

COMPANY

>2000 200 x 98 x 27 24 20 40 pneumati

c motor -

Tendons

Metal + plastic

Xu Zhe’s

hand

Xu Zhe 942 No data

(Human size)

15 10 10 DC

Motor -

Tendons

3D print ABS +

Laser-cut tissues

(rubber)

Bebionic

(2014)

RSL

steeper

>550 ~200 x 92 x 50 11 6 5 DC

Motor -

Lead

Screw

Metal

Grasp test (1)

[1]Cutkosky, M.R., On Grasp Choice, Grasp Models, and the Design of Hands for Manufacturing Tasks.

Ieee Transactions on Robotics and Automation, 1989. 5(3): p. 269-279.

[2]Li Tian, Nadia Magnenat Thalmann, Daniel Thalmann, Jianmin Zheng , The making of a humanoid

hand: problems and solutions, submitted to Front. Robot. AI at 18 May

Grasp test (2)

[1]Cutkosky, M.R., On Grasp Choice, Grasp Models, and the Design of Hands for Manufacturing Tasks.

Ieee Transactions on Robotics and Automation, 1989. 5(3): p. 269-279.

[2]Li Tian, Nadia Magnenat Thalmann, Daniel Thalmann, Jianmin Zheng , The making of a humanoid

hand: problems and solutions, submitted to Front. Robot. AI at 18 May

The test is based on Cutkosky’s hand taxonomy. It shows our robotic hand

can grasp in all 16 grasp gestures.

Grasp test demos

Thank you!

Q&A