Autonomous Data Exchange in MultiRobot Collectives

1
Autonomous Data Exchange in Multi-Robot
Collectives

• Dr. William M. Spears
• Dr. Diana F. Spears
• Dr. Jerry Hamann
• University of Wyoming

2
Description/Objective

• AFRL has identified a need to develop the ability
  for robots to interact with each other with
  little or no human intervention for data exchange
  (maps, obstacles, enemy location, sensor
  information).
• If robots know their locations relative to their
  neighbors, they can exchange sensor information
  in a spatially accurate, cooperative manner.
  Novel localization technology, coupled with
  algorithms for distributed control, data
  exchange, and data fusion will result in
  coordinated behavior towards a common goal.
  Applications include surveillance, search,
  distributed sensing grids, mapping, and mine
  clearing.

3
Key Technology Invention

• Novel UW robot localization technology serves to
  unify control, positioning, and data exchange.
  GPS knowledge or human intervention not
  necessary, although they can be integrated if
  desired.
• This technology is based on a concept called
  trilateration.

4
Novel Localization Technique
Acoustic Transducer
Robot 2
RF
Robot 1
d1
d2
d3
Each robot carries one RF and three acoustic
transducers.
Robot 1 emits RF and acoustic pulse. When 2
receives RF pulse, it starts 2s clock. When the
acoustic pulse is received, the time of flight
information is converted to 3 distances
measurements. A simple formula,
easily implemented in hardware, computes the
range and bearing of robot 1 relative to robot
2s coordinate system.
5
Multi-Robot Localization and Data Fusion
Y axis
Robot 1
Robot 3
X axis
Robot 2
Each robot carries one RF and three acoustic
transducers for localization and data exchange.
For example, robot 1 can see 2, but not 3. Robot
2 can see 1 and 3. If each robot has a unique
identifier, linear transformations fuse
localization with other sensor data.
6
Maxelbot Version 1.0
Top down and sideways view of the Version 1.0
Maxelbot. The performance is good, but the
platform is not robust.
7
Maxelbot Version 2.0
Porting to the MMP5 chassis (wheels/motors/batteri
es/body). Far more robust will probably be able
to go outdoors!
8
Work Flow
Software Development
Hardware Additions
Demonstrations
9
Time Table
10
Hardware

• Trilateration Hardware
• Digital Compass
• Shaft Encoders
• Obstacle Avoidance Module
• High Speed RF Communication
• Digital Thermometer

11
Trilateration Hardware

• This module allows each robot to compute the
  range and bearing to the other robots. This is
  essential for localization. It does not require
  global beacons, GPS, etc.
• We expect to have this module installed on 5 - 7
  MMP5 platforms.

12
Digital Compass

• This allows each robot to be aware of a global
  coordinate system. Useful if the robots are
  communicating information back to a human.
• We expect to have one digital compass per MMP5
  platform.

13
Shaft Encoders

• These act like odometers in your car they
  provide accurate measurements of the distance the
  drive train in the MMP5 platform has traveled.
  Allows for more accurate movement.
• We expect to have 2 shaft encoders per MMP5
  robot, one for each drive train.

14
Obstacle Avoidance Module

• Built from Sharp IR sensors, that measure
  distance to an object.
• We expect to use 8 Sharp IR sensors per robot, in
  differing configurations.

15
High Speed RF

• This module will allow faster inter-robot
  communication as well as (potentially)
  communication with a human observer.
• Will enable various degrees of Human-Computer
  Interface (HCI).

16
Digital Thermometer

• Provides useful environmental information.
• Allows us to test search algorithms, to try to
  find hot spots in the environment.

17
Software

• Control Algorithms
• Run-Time Checking
• Search Algorithms
• Built-In Test

18
Control Algorithms

• We will make use of the latest in swarm
  robotics control algorithms, including
  behavior-based algorithms and Artificial Physics
  (invented by W. M. Spears)
• The latter is especially well suited to our
  trilateration localization technology. Especially
  good for formations of robots and obstacle
  avoidance.

19
Run-Time Checking

• Although Artificial Physics is very robust to
  sensor noise, disruptions in a formation of
  robots can occur.
• Run-time checking can detect and repair such
  problems if they occur.

20
Search Algorithms

• We are currently enhancing Artificial Physics so
  that robot formations can search for areas of
  interest in an environment.
• The formation acts to provide an implicit
  mechanism for consensus, filtering out sensor
  noise.

21
Built-In Test

• As our robot platforms are augmented with
  additional sensors and modules, the probability
  of hardware failures increases quickly.
• As a consequence, built-in test capabilities will
  be incorporated as the sensors and modules are
  added.

22
Demonstrations

• Collective Pulling/Pushing
• Obstacle Avoidance
• Map Building
• Search

23
Collective Pulling
Box that is heavier on one end
Real Tethers
Virtual Tethers Created by Trilateration
and Control Algorithms
24
Obstacle Avoidance
25
Map Building
Linear formation of robots created using
Artificial Physics. A map of the maze should be
displayed to a human user.
26
Search
This is a top down view of robots exploring an
environment for a hot-spot. One robot wanders.
Three robots in formation find the hot-spot (the
diamond). The dark squares are obstacles.
27
Credits

• Rodney Heil, Dimitri Zarzhitsky, Tom Kunkel, Paul
  Maxim, Suranga Hettiarachchi, Derek Green, Anton
  Rebguns, and Christer Karlsson.