Simulating Multi-Robot Exploration Using ROS and MORSE Zhi Yan, Luc - - PowerPoint PPT Presentation

▶

simulating multi robot exploration using ros and morse

Simulating Multi-Robot Exploration Using ROS and MORSE Zhi Yan, Luc - - PowerPoint PPT Presentation

Oct 23, 2022 318 likes •564 views

Introduction Our infrastructure Experiments Conclusion and future work Appendix Simulating Multi-Robot Exploration Using ROS and MORSE Zhi Yan, Luc Fabresse, Jannik Laval and Noury Bouraqadi firstname.lastname@mines-douai.fr Institut

slide-1

SLIDE 1

Introduction Our infrastructure Experiments Conclusion and future work Appendix

Simulating Multi-Robot Exploration Using ROS and MORSE

Zhi Yan, Luc Fabresse, Jannik Laval and Noury Bouraqadi firstname.lastname@mines-douai.fr

Institut Mines-Telecom, Mines Douai http://car.mines-douai.fr

CAR 2014, June 23, 2014

1 / 16

slide-2

SLIDE 2

Introduction Our infrastructure Experiments Conclusion and future work Appendix

Plan

1

Introduction

2

Our infrastructure

3

Experiments

4

Conclusion and future work

2 / 16

slide-3

SLIDE 3

Introduction Our infrastructure Experiments Conclusion and future work Appendix Multi-robot exploration Robotics simulation

Multi-robot exploration

Research background Exploring an unknown environment in cooperation Building a map of this environment Application areas Search and rescue in earthquake Fire searching inside building Mineral exploration Mine clearance

3 / 16

slide-4

SLIDE 4

Introduction Our infrastructure Experiments Conclusion and future work Appendix Multi-robot exploration Robotics simulation

Multi-robot exploration

Research background Exploring an unknown environment in cooperation Building a map of this environment Application areas Search and rescue in earthquake Fire searching inside building Mineral exploration Mine clearance Question Which coordination strategy?

3 / 16

slide-5

SLIDE 5

Introduction Our infrastructure Experiments Conclusion and future work Appendix Multi-robot exploration Robotics simulation

Multi-robot exploration

Research background Exploring an unknown environment in cooperation Building a map of this environment Application areas Search and rescue in earthquake Fire searching inside building Mineral exploration Mine clearance Question Which coordination strategy? Answer? Need to compare!

3 / 16

slide-6

SLIDE 6

Introduction Our infrastructure Experiments Conclusion and future work Appendix Multi-robot exploration Robotics simulation

How to compare?

Problem Development and validation of coordination strategies with actual robots: long time! Debugging and testing: quite complex!

4 / 16

slide-7

SLIDE 7

Introduction Our infrastructure Experiments Conclusion and future work Appendix Multi-robot exploration Robotics simulation

How to compare?

Problem Development and validation of coordination strategies with actual robots: long time! Debugging and testing: quite complex! Solution Simulation before deploying to actual robots! Which simulator?

BOSS: a discrete multi-roBOt Simulator in Smalltalk Stage: a 2D multi-robot simulator

BOSS DEMO WITH 5 ROBOTS STAGE DEMO WITH 4 ROBOTS 4 / 16

slide-8

SLIDE 8

Introduction Our infrastructure Experiments Conclusion and future work Appendix Multi-robot exploration Robotics simulation

Robotics simulation

Challenges Simulations need to be as realistic as possible Difference between simulated and real robots should be minimized

5 / 16

slide-9

SLIDE 9

Introduction Our infrastructure Experiments Conclusion and future work Appendix Multi-robot exploration Robotics simulation

Robotics simulation

Our goals Build a realistic test bed for evaluating different coordination algorithms in different conditions Develop performance benchmarks for quantitative analysing and comparing different algorithms

6 / 16

slide-10

SLIDE 10

Introduction Our infrastructure Experiments Conclusion and future work Appendix Multi-robot exploration Robotics simulation

Robotics simulation

Our goals Build a realistic test bed for evaluating different coordination algorithms in different conditions Develop performance benchmarks for quantitative analysing and comparing different algorithms Our propose MORSE: a 3D simulator with realistic physics engine ROS: a de facto standard middleware Cluster: a high performance distributed computing

6 / 16

slide-11

SLIDE 11

Introduction Our infrastructure Experiments Conclusion and future work Appendix System overview

System overview

Multi-robot mapping: robot exchanges the explored map with its teammates Multi-robot motion planning: robot moves towards the nearest frontier

7 / 16

slide-12

SLIDE 12

Introduction Our infrastructure Experiments Conclusion and future work Appendix System overview

System overview

Multi-robot mapping: robot exchanges the explored map with its teammates Multi-robot motion planning: robot moves towards the nearest frontier Time metric: total time required to complete an exploration mission Cost metric: sum of energy consumed by all robots in the team

7 / 16

slide-13

SLIDE 13

Introduction Our infrastructure Experiments Conclusion and future work Appendix Setup Results

Robot setup

gmapping: laser-based SLAM (Grisetti et al., 2007) explore: frontier-based exploration (Yamauchi, 1997) map fusion: multiple maps merging (developed by our team) move base: mobile robot navigation (using Dijkstra pathfinding algorithm)

8 / 16

slide-14

SLIDE 14

Introduction Our infrastructure Experiments Conclusion and future work Appendix Setup Results

Robot setup

gmapping: laser-based SLAM (Grisetti et al., 2007) explore: frontier-based exploration (Yamauchi, 1997) map fusion: multiple maps merging (developed by our team) move base: mobile robot navigation (using Dijkstra pathfinding algorithm)

8 / 16

slide-15

SLIDE 15

Introduction Our infrastructure Experiments Conclusion and future work Appendix Setup Results

Robot setup

gmapping: laser-based SLAM (Grisetti et al., 2007) explore: frontier-based exploration (Yamauchi, 1997) map fusion: multiple maps merging (developed by our team) move base: mobile robot navigation (using Dijkstra pathfinding algorithm)

8 / 16

slide-16

SLIDE 16

Introduction Our infrastructure Experiments Conclusion and future work Appendix Setup Results

Test bed setup

Robot 1

roscore MORSE

ROS middleware

Robot 2 Robot n Computer cluster 70 computing nodes: 8 to 12 processors and 16Go to 48Go RAM 1 master node: scheduler

9 / 16

slide-17

SLIDE 17

Introduction Our infrastructure Experiments Conclusion and future work Appendix Setup Results

Experimental setup

multiple dead-ends

Fixed parameters Robot characteristics: a homogeneous team of Pioneer 3-DX robots equipped with a SICK LMS500 laser scanner Terrain properties: a maze-like space with 80 meters long and 80 meters wide Communication range: 200 meters Coordination strategy: collaborative mapping, robots exchange their map once every 5 seconds

MORSE DEMO WITH 8 ROBOTS 10 / 16

slide-18

SLIDE 18

Introduction Our infrastructure Experiments Conclusion and future work Appendix Setup Results

Experimental setup

Experiments A Blind positionning Experiments B 1 robot per entry point Experiments C 2 robots per entry points

Varied parameters Exploration team size: from 1 to 14 robots Initial positions of robots: experiment A, B and C

11 / 16

slide-19

SLIDE 19

Introduction Our infrastructure Experiments Conclusion and future work Appendix Setup Results

Results

200

400 600 800 1000 1200 1400 Number of robots Time 1 2 3 4 5 6 7 8 9 10 11 12 13 14

experiment A: blind

experiment B: 1 robot per entry point experiment C: 2 robots per entry point

500

1000 1500 2000 2500 Number of robots Cost 1 2 3 4 5 6 7 8 9 10 11 12 13 14

experiment A: blind

experiment B: 1 robot per entry point experiment C: 2 robots per entry point

12 / 16

slide-20

SLIDE 20

Introduction Our infrastructure Experiments Conclusion and future work Appendix

Conclusion and future work

Contribution A realistic test bed for evaluating different coordination strategies for multi-robot exploration Preliminary experiments with time and cost evaluation metrics Future work Different representative environments More robots → homogeneous and heterogeneous teams Odometry noise → more efficient map fusion algorithms Communication range → more efficient coordination strategies

13 / 16

slide-21

SLIDE 21

Introduction Our infrastructure Experiments Conclusion and future work Appendix

Simulating Multi-Robot Exploration Using ROS and MORSE

Zhi Yan, Luc Fabresse, Jannik Laval and Noury Bouraqadi firstname.lastname@mines-douai.fr

Institut Mines-Telecom, Mines Douai http://car.mines-douai.fr

CAR 2014, June 23, 2014

14 / 16

slide-22

SLIDE 22

Introduction Our infrastructure Experiments Conclusion and future work Appendix Multi-robot communication Multi-robot mapping

Multi-robot communication

Algorithm 1 Communication Connection for roboti

1: Querying all published ROS topics 2: Subscribing to robot pose topics 3: if ∃robotj ∈ exploration team : distBetween(robotj −roboti) <

max comm distance then

4:

Establishing connection with robotj

5: end if

15 / 16

slide-23

SLIDE 23

Introduction Our infrastructure Experiments Conclusion and future work Appendix Multi-robot communication Multi-robot mapping

Multi-robot mapping

Algorithm 2 Map Fusion for roboti

1: δ ← (roboti.init pose −robotj.init pose)×map scale 2: roboti.fused map ← roboti.map 3: for all grid in roboti.fused map do 4:

if grid = NO INFORMATION then

5:

grid ← robotj.mapgrid.pose+δ

6:

end if

7: end for

16 / 16