decentralized prioritized planning in large multirobot teams

27
Decentralized prioritized planning in large multirobot teams Prasanna Velagapudi Paul Scerri Katia Sycara Carnegie Mellon University, Robotics Institute IROS 2010

Upload: asher

Post on 24-Feb-2016

51 views

Category:

Documents


0 download

DESCRIPTION

Decentralized prioritized planning in large multirobot teams. Prasanna Velagapudi Paul Scerri Katia Sycara Carnegie Mellon University, Robotics Institute. Motivation. Disaster response, Convoy planning 100s of robots coordinating to plan P lanning is offline - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Decentralized prioritized planning in large  multirobot  teams

IROS 2010

Decentralized prioritized planning in large multirobot teams

Prasanna VelagapudiPaul Scerri

Katia Sycara

Carnegie Mellon University, Robotics Institute

Page 2: Decentralized prioritized planning in large  multirobot  teams

IROS 2010 2

Motivation

• Disaster response, Convoy planning

• 100s of robots coordinating to plan

• Planning is offline• Computing is

distributed across robots

Page 3: Decentralized prioritized planning in large  multirobot  teams

IROS 2010 3

Multiagent Path Planning

Start

Goal

Page 4: Decentralized prioritized planning in large  multirobot  teams

IROS 2010 4

Large-Scale Path Planning

Page 5: Decentralized prioritized planning in large  multirobot  teams

IROS 2010 5

Large-Scale Path Planning

Page 6: Decentralized prioritized planning in large  multirobot  teams

IROS 2010 6

Large-Scale Path Planning

Page 7: Decentralized prioritized planning in large  multirobot  teams

Multiagent Path Planning

• Many, many approaches: offline fewer robots• Take a simple, decoupled approach, prioritized

planning– [Erdman 1987], [van den Berg 2005]

• Try parallelization + scale up, see what happens– Large teams, fast convergence, low communication

• Similar to some reactive/online approaches– [Chun 1999], [Clark 2003], [Chiddawar 2009]

Page 8: Decentralized prioritized planning in large  multirobot  teams

IROS 2010 8

Prioritized Planning

• Assign priorities to agents based on path length

[Erdman, et al 1987; van den Berg, et al 2005]

Page 9: Decentralized prioritized planning in large  multirobot  teams

IROS 2010 9

Prioritized Planning

• Plan from highest priority to lowest priority• Use previous agents as dynamic obstacles

[Erdman, et al 1987; van den Berg, et al 2005]

Effective, but requiresn sequential planning steps

Page 10: Decentralized prioritized planning in large  multirobot  teams

IROS 2010 10

Can we do better?

• Each agent has local computing anyway

• Let agents try to plan instead of doing nothing– Maybe we’ll need to re-plan– If we don’t re-plan, we have saved time

• Hypothesis: Agents only actually collide with few other agents, so sequential iterations << n

Page 11: Decentralized prioritized planning in large  multirobot  teams

IROS 2010 11

Distributed Prioritized Planning

Parallelizable& Equivalent

Page 12: Decentralized prioritized planning in large  multirobot  teams

IROS 2010 12

Distributed Prioritized Planning

• At each robot:1. Compute initial path2. Determine local priority3. Broadcast path to team4. Listen for other teammates paths5. If a higher priority path is received, add as an obstacle in

space-time6. Compute new collision-free path7. Go to step 3.

Equivalent, but n2 messages!

Page 13: Decentralized prioritized planning in large  multirobot  teams

Reduced DPP

• DPP requires broadcasting messages to every teammate every time agents replan

• Reduce this with two assumptions– If you didn’t hear from someone, they didn’t change their

plan– If someone is higher priority, they don’t care what you do,

so don’t send them anything

Better, but still O(n2) messages

Page 14: Decentralized prioritized planning in large  multirobot  teams

Can we send even less?

• Birthday Paradox– If everybody in a room compares birthdays, chances of two

people having the same birthday grows quickly as number of people grows

• Collision communications– If everybody in the team compares a few other agents’

paths, the chance of detecting a collision between anybody grows quickly as number of paths compared increases

– Each agent is doing a small O(n2) check

IROS 2010 14

Page 15: Decentralized prioritized planning in large  multirobot  teams

IROS 2010 15

Can we send even less?

• Choose num_paths_sent = k * sqrt(n)

Page 16: Decentralized prioritized planning in large  multirobot  teams

IROS 2010 16

Sparse DPP

• Goal: reduce # of messages even more than RDPP O(n*sqrt(n))1. Each robot sends path to k*sqrt(n) random neighbors2. Each robot checks for conflicts between every

combination of paths it receives, then notifies conflicting robots

3. Lower priority robots in the collision re-plan

Page 17: Decentralized prioritized planning in large  multirobot  teams

Experimental Results

• Scaling Dataset– # robots varied: {40, 60, 80, 120, 160, 240}– Density of map constant: 8 cells per robot

• Density Dataset– # robots constant: 240– Density of map varied: {32, 24, 16, 12, 8} cells per robot

• Cellular automata to generate 15 random maps• Maps solved with centralized prioritized planning• DPP variants capped at 20 iterations• Local planner: A*

IROS 2010 17

Page 18: Decentralized prioritized planning in large  multirobot  teams

IROS 2010 18

Same near-optimal solutions as PP

Varying Team Size Varying Density

Page 19: Decentralized prioritized planning in large  multirobot  teams

IROS 2010 19

Fewer sequential iterations (Iteration limit = 20)

Varying Team Size Varying Density

Page 20: Decentralized prioritized planning in large  multirobot  teams

IROS 2010 20

Sparse DPP fails to converge (Complete, Reduced DPP always converged)

Varying Team Size Varying Density

Page 21: Decentralized prioritized planning in large  multirobot  teams

IROS 2010 21

Reduced DPP reduces communication

Varying Team Size Varying Density

Complete Communication

Page 22: Decentralized prioritized planning in large  multirobot  teams

IROS 2010 22

DPP takes… longer?

Varying Team Size Varying Density

Page 23: Decentralized prioritized planning in large  multirobot  teams

IROS 2010 23

Distribution of Planning Times

Page 24: Decentralized prioritized planning in large  multirobot  teams

IROS 2010 24

• Prioritized Planning

• DPP

Replanning for the Worst Agent

ABCD

ABCD

Longest planning agents might replan multiple times

Individual agent planning times varied by >2 orders of magnitude

Potential solution:Incremental Planning

Page 25: Decentralized prioritized planning in large  multirobot  teams

IROS 2010 25

Summary of Results

• DPP gets same quality solutions as centralized• Reduced DPP is efficient

– Many fewer sequential steps, messages– Longer wall-clock time (due to uneven planning times)

• Sparse DPP does surprisingly poorly overall – Detecting collisions alone (reactive) leads to slower

convergence, more re-planning– Better to exchange relevant paths (proactive)– In Reduced DPP, agents preemptively discover conflicts

before collisions occur

Page 26: Decentralized prioritized planning in large  multirobot  teams

IROS 2010 26

Conclusions

• DPP shows promise for larger problems with distributed computing– Far fewer sequential planning iterations– Incremental planning should reduce execution time

• However, there are some caveats– Sensitive to collision detection– If distribution of planning times varies, can be slow

Page 27: Decentralized prioritized planning in large  multirobot  teams

IROS 2010 27

Future Work

• Generalizing framework for distributed planning through iterative message exchange

• Asynchronous collision-detection, re-planning• Reducing necessary communication• Planning under uncertainty• Scaling to larger team sizes