simbeeotic : a simulator and testbed for micro- aerial vehicle swarm experiments

23
Simbeeotic: A Simulator and Testbed for Micro- Aerial Vehicle Swarm Experiments IPSN 2012 Bryan Kate, Jason Waterman, Karthik Dantu (Harvard Un iversity), Matt Welsh (Google) Study group 2012.04.23 Meng-Lin, Lu 1

Upload: keena

Post on 22-Feb-2016

54 views

Category:

Documents


0 download

DESCRIPTION

Simbeeotic : A Simulator and Testbed for Micro- Aerial Vehicle Swarm Experiments. Study group 2012.04.23 Meng -Lin, Lu. IPSN 2012 Bryan Kate, Jason Waterman, Karthik Dantu (Harvard University), Matt Welsh (Google). MAV Swarms Background. Features. Simulation. Extremely small. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Simbeeotic : A Simulator and  Testbed  for Micro- Aerial Vehicle Swarm Experiments

1

Simbeeotic: A Simulator and Testbed for Micro-AerialVehicle Swarm ExperimentsIPSN 2012Bryan Kate, Jason Waterman, Karthik Dantu

(Harvard University), Matt Welsh (Google)

Study group2012.04.23

Meng-Lin, Lu

Page 2: Simbeeotic : A Simulator and  Testbed  for Micro- Aerial Vehicle Swarm Experiments

2

MAV Swarms Background

Features

Extremely small

Large number

Simulation

Modeling the key aspects: actuation, sensing and communication

Staged deploymentto prototype testbedSimbeeotic

Page 3: Simbeeotic : A Simulator and  Testbed  for Micro- Aerial Vehicle Swarm Experiments

3

Outline

0Introduction0Simulator Design0Helicopter Testbed0Evaluation0Future Works0Conclusions

Page 4: Simbeeotic : A Simulator and  Testbed  for Micro- Aerial Vehicle Swarm Experiments

4

Introduction

0Simulation0 Rapid prototyping0 Emulation of future architectures0 Testing at scale

0Differences between static and dynamic WSN0 Radio is not primary energy consumer0 Duty cycling can’t work when sensors fly or move

0Several system had purposed before, but not satisfied their requirements and design decisions

Page 5: Simbeeotic : A Simulator and  Testbed  for Micro- Aerial Vehicle Swarm Experiments

5

Introduction

4 Contributions:0Scalability

0 Simulate thousands of MAVs in a single scenario0Completeness

0 Model as much of the problem domain as possible0Variable fidelity

0 Adjust for each purpose without losing accuracy0Staged development

0 Facilitate the development of software and hardware

Page 6: Simbeeotic : A Simulator and  Testbed  for Micro- Aerial Vehicle Swarm Experiments

6

Simulator Design

Simbeeotic:0Discrete event simulator

0 A simulation execution consists of one or more models that schedule events to occur at a future point in time

0 Virtual time – moved forward by an executive that get the next event and pass it to the intended recipent

0Written in Java programming language0 easily learned by neophytes0 large repository of high quality, open source libraries

0Repeatability0Ease of use

Page 7: Simbeeotic : A Simulator and  Testbed  for Micro- Aerial Vehicle Swarm Experiments

7

Simulator Design

Architecture

Bottom

Top

Heart

Bottom

Top

Page 8: Simbeeotic : A Simulator and  Testbed  for Micro- Aerial Vehicle Swarm Experiments

8

Simulator Design

Physics engine- JBullet0Rigid Bodies

0 Simple shapes, complex geometries0Dynamics Modeling

0 Integrating the forces and torques03D Continuous Collision Detection

0 Physical interactions between objects0Ray Tracing

0 Range finders and optical flow

Page 9: Simbeeotic : A Simulator and  Testbed  for Micro- Aerial Vehicle Swarm Experiments

9

Simulator DesignMAV domain models0Virtual world0Weather0Sensors – inertial (accelerometer, gyroscope, optical

flow), navigation (position, compass), environmental (camera, range, bump)

0RF communicationSoftware engineering tricks0Reflection0Runtime annotation processing0Parameterization: key-value pairs

Page 10: Simbeeotic : A Simulator and  Testbed  for Micro- Aerial Vehicle Swarm Experiments

10

Helicopter Testbed

0 Indoor MAV testbed0E-flite Blade mCX2 RC helicopter

0 Proprietary control board stabilizes flight (yaw axis only)0 Without other processors, sensors, or radios0 Not expensive, small V.S. toy

Remote control0Using Vicon motion capture system for remote control0 Input signal to the helicopter ‘s transmitter

0 yaw, pitch, roll, and throttle

Page 11: Simbeeotic : A Simulator and  Testbed  for Micro- Aerial Vehicle Swarm Experiments

11

Aircraft Yaw Motion

Page 12: Simbeeotic : A Simulator and  Testbed  for Micro- Aerial Vehicle Swarm Experiments

12

Aircraft Pitch Motion

Page 13: Simbeeotic : A Simulator and  Testbed  for Micro- Aerial Vehicle Swarm Experiments

13

Aircraft Roll Motion

Page 14: Simbeeotic : A Simulator and  Testbed  for Micro- Aerial Vehicle Swarm Experiments

14

Simbeeotic Integration

Ghost model

Page 15: Simbeeotic : A Simulator and  Testbed  for Micro- Aerial Vehicle Swarm Experiments

15

HWIL Discussion

Advantages0Fly real vehicles using virtual sensors0Transform laboratory space into an arbitrarily Env.0Test the limits of proposed hardware and software Disadvantages:0 Inaccuracy cauesd by Vicon motion capture system0Can’t fly outdoors0Heavy computing resources0Can’t process or sense on helicopter0Latency: processing, transmission, control

Page 16: Simbeeotic : A Simulator and  Testbed  for Micro- Aerial Vehicle Swarm Experiments

16

Evaluation

0Workload0 10Hz kinematic update rate0 1Hz compass sensor reading0 100 virtual seconds

0Environment Complexity

Page 17: Simbeeotic : A Simulator and  Testbed  for Micro- Aerial Vehicle Swarm Experiments

17

Evaluation

0Swarm Size

Page 18: Simbeeotic : A Simulator and  Testbed  for Micro- Aerial Vehicle Swarm Experiments

18

Evaluation0Model Complexity

0 Increase event execution time – event complexity, message explosion

Page 19: Simbeeotic : A Simulator and  Testbed  for Micro- Aerial Vehicle Swarm Experiments

19

Example Scenarios 1

0Coverage0 search a space for features of interest (e.g. flowers)

Page 20: Simbeeotic : A Simulator and  Testbed  for Micro- Aerial Vehicle Swarm Experiments

20

Example Scenarios 2

0Explores the possibility of using RF beacons

Page 21: Simbeeotic : A Simulator and  Testbed  for Micro- Aerial Vehicle Swarm Experiments

21

Future Works

0Scalability0 Physics engine is a bottleneck0 JBullet -> Bullet : JAVA->C++

0Fidelity0 Improving networking models for communication0 Expand HWIL capabilities to include real radios

0Autonomy0 Leverage ROS to control0 TOSSIM-like approach to simulate0 Kinect

Page 22: Simbeeotic : A Simulator and  Testbed  for Micro- Aerial Vehicle Swarm Experiments

22

Conclusions

0Provide a feasible way to simulate MAV swarms0Cool, and may be useful in simulation but seems

useless now in reality0Too complex to make whole system robust (network,

motion capture, robot control)