Multi-Robot Mapping

1
Multi-Robot Mapping

• An Investigation into the Substitutability of
  Sensing and Communication

2
Problem Definition

• Given a team of robots at random, unknown
  starting locations in a known-sized environment,
  how much time is required until at least one
  robot has a complete map of the environment?
• To what degree are communication and sensing
  substitutable in the mapping task?
• How to combine maps with low communication?
• How to ensure cooperative robot behavior?

3
Background

• N. Roy G. Dudek Collaborative Exploration and
  Rendezvous Algorithms, Performance Bounds and
  Observations. Autonomous Robots, 11(2) 117-136,
  2001.
• Simple version
• Travel through environment
• Identify good rendezvous locations
• At time-out, choose best location
• Go to location and share map with other robots

4
Rendezvous

• How to identify landmarks (rendezvous points)?
• What if landmark sets not equal? Best landmark
  not known to all robots?
• What if robots fail to rendezvous
  (asynchronicity/travel time, missing landmarks)?

5
Collaborative Exploration

• Burgard et. al. Collaborative Multi-Robot
  Exploration, IEEE Conference on Robotics and
  Automation, 2000
• How to keep robots from all going to the same
  large unexplored area?
• Trade off Utility of exploring area with Cost of
  movement to area
• Decrement utility for others if a robot chooses a
  destination (i.e., assume the exploration has
  already happened)

6
(No Transcript)
7
Project Plan

• Develop single-robot behavior based controller
  for mapping (potential field)
• Implement existing MRS mapping solution
• Extend single-robot solution to MRS by adding
  communication as another sensor
• Evaluate MRS mapping WRT Sensing v.s.
  Communication
• Report results

8
The Details

• 360O sensing
• Broadcast communication (range limited)
• Fixed size robot team (3 or 5 robots)
• 3 fixed-size environment
• Clear/Few obstacles
• Moderate obstacles
• Cluttered environment
• Perfect odometry

9
The Details

• Keep Communication coverage area Sensor
  coverage area constant
• Broadcast location of detected obstacles WRT
  local coordinate frame
• Unexplored areas act as attractors
• 7 configurations x 3 environments x 10 trials
  210 experiments
• 9 config. x 2 team sizes x 5 env. X 100 trials
  9000 experiments

10
Issues and Expectations

• Is communication used to coordinate exploration?
• How much information to exchange, and when?
• Low comm good for clear
• Low sensing good for cluttered
• In most cases, non-extreme configurations better

11
Questions?