Dynamic Mission Planning for Multiple Mobile Robots

1
Dynamic Mission Planning for Multiple Mobile
Robots

• Barry Brumitt and Anthony Stentz
• 26 Oct, 1999
• AMRS-99 Class Presentation
• Brian Chemel

2
Overview

• Problem description
• Mission grammar
• System architecture GRAMMPS
• Brief digression D
• Results
• Analysis and limitations

3
Problem Description Environment

• Dynamic, complex environment
• Typical situation
• World state initially unknown
• Runtime observations incorporated into shared
  world model

4
Problem Description Robots

• Multiple, heterogeneous mobile robots
• Assumptions
• Robots have relatively open workspaces, so
  solution set is not too sparse
• Effective positioning, communication and
  perception are a given

5
Problem Description Goals

• Robot task move to specified locations, in
  specified order
• Reconnaissance example
• Waypoints to be scouted by team of robots
• Warehouse example
• Multiple pickup points, to be followed by
  multiple delivery points

6
Problem Description GRAMMPS Planner

• Military reconnaissance task
• Outdoor environment
• Multiple robot vehicles (Navlab HMMWVs, in
  practice)
• Multiple distributed goals, with sequencing
• Requires mission grammar to pass parameters to
  distributed planning system

7
Mission Grammar
Expression Meaning
Do A, then do B
A OR B
A AND B
Robot i
Goal j
Move robot r to goal g
8
Mission Grammar Example

• Move either robot 1 or robot 2 to goals 1, 2, 3,
  and 4. Then move both robots 1 and 2 to goal 5.

9
System Architecture Overview
Dynamic Planners (one per goal)

• Global (shared) dynamic planners
• Global (shared) mission planner
• Local (individual) plan execution

D
D
D
D
D
Mission Planner
Robot 1
Robot 2
Robot n

10
System Architecture Local Navigators
Dynamic Planners (one per goal)

• Input
• Path to assigned goal
• Perception information
• Output
• Steering commands

D
D
D
D
D
Mission Planner
Robot 1
Robot 2
Robot n

11
System Architecture Mission Planner
Dynamic Planners (one per goal)

• Input
• Estimated path costs for each (robot,goal) pair
• Output
• Mapping from robots to goals
• Algorithm
• TSP heuristic

D
D
D
D
D
Mission Planner
Robot 1
Robot 2
Robot n

12
System ArchitectureDynamic Planners
Dynamic Planners (one per goal)

• Input
• Current world state knowledge
• Output
• Path to each goal for each robot
• Planning algorithm
• D

D
D
D
D
D
Mission Planner
Robot 1
Robot 2
Robot n

13
Brief Digression The D Algorithm

• Modification of the A planning algorithm
• Provides efficient, optimal and complete path
  planning in unknown, partially known, and
  changing environments
• As new information about the environment is
  learned, cost information is propagated through
  state space
• Each time new information makes previous path
  calculations obsolete, a new path is calculated
• Original paper Stentz, ICRA 94

14
Example Simulation Initial Plan

• Mission statement

15
Example Simulation Final Plan

• Goals re-assigned on the fly
• Mission successfully completed

16
Highlights

• Demonstration implementation on Navlab HMMWVs
  allows real-time, team-based mission planning in
  dynamic environments
• System scales gracefully up to large numbers of
  robots and goals

17
Limitations

• Waypoint task structure is very limiting
• No discussion of how to modify TSP approach to
  allow heterogeneity
• One robot must act as leader of the entire team
• When D fails, system can lock up

18
Related Work

• Stentz, T. Optimal and Efficient Path Planning
  for Partially-Known Environments. ICRA 94
• Brumitt, B., Stentz, T. GRAMMPS A Generalized
  Mission Planner for Multiple Mobile Robots in
  Unstructured Environments. ICRA 98
• Brumitt, B., Hebert, M. Experiments in
  Autonomous Driving With Concurrent Goals and
  Multiple Vehicles. ICRA 98