Distributed MultiRobot Construction of 2D Structures

1
Distributed Multi-Robot Construction of 2D
Structures

• Doug Demyen
• Ken Dwyer

2
Motivation

• What are the possible uses of structure-building
  teams of robots?
• Automated construction
• Able to operate in settings that present danger
  to humans
• Build infrastructure in preparation for human
  inhabitation
• e.g. Underwater or extra-terrestrial environments

3
Problem Description

• Construct specific 2D structures by deploying a
  team of robots
• Building material consists of passive, moveable
  blocks
• Robots have only local knowledge and
  communication
• Solution should be robust, scalable, and also
  take advantage of parallelism

4
Challenges

• Multiple robots working in an area could result
  in congestion (e.g. traffic jam)
• Because of local vision, some portions of the
  structure may be neglected
• Robots could interfere with each other
• Picking up blocks that are part of the structure
• Multiple robots trying to assume one vertex, etc.
• Individual robots may fail during the
  construction process

5
Distributed Multi-Robot Systems

• Advantages
• Can tolerate individual robot failures (at least
  to some extent)
• Algorithms are less computationally intensive
  compared to those required for a centralized
  controller
• Disadvantage
• Since each robot has access to only local
  information, solutions are usually suboptimal

6
Related Work

• Wawerla et al. (2002)
• Construct a straight barrier
• Laser beacon indicates starting point
  orientation (also seeded)

• Stewart Russell (2004)
• Build a loose wall structure
• Organizer robot coordinates activities of builder
  robots

7
Related Work (2)

• Melhuish et al. (1999)
• Wall-building without direct communication

8
Werfels Approach

• Werfel (2004) proposed a technique for building
  pre-specified 2D structures
• A beacon is placed in the workspace which can
  send out long-range signals
• It broadcasts information about the structure
• The robots have a restricted vision and
  communication range

9
Werfels Approach (2)

• Blocks are assumed to be scattered randomly
  across the workspace
• Robots first clear a space within which to work
• C robots act as corners, where C is the number
  of corners in the structure
• Remaining robots collect blocks and build walls
  between the corners
• Once a robot has verified that all walls have
  been filled in, it stops building

10
Werfels Approach (3)
11
Werfels Approach (4)

• A number of shortcomings are apparent
• Simulated on a 2D cellular grid
• Unrealistic, since movement is not accurately
  modeled
• Robustness was not tested
• If the beacon fails, the robots no longer have a
  signal by which to localize themselves
• Overall a good idea, but the assumptions are
  unlikely to hold in reality

12
Our Approach

• Build an arbitrary polygon or line strip
• possibly restricted to non-self-intersecting
• Use local vision and no communication, outside
  what can be sensed through vision
• Robots can both build the structure and serve as
  the vertices of the polygon
• Vertex robots form the template for the
  structure

13
The Algorithm (Overview)

• Robots clear bricks from the work area
• One robot assumes the first vertex of the figure
  to be constructed
• Other robots roam the workspace until the
  structure is complete, performing
• Picking up bricks from the outside of the work
  area
• Placing bricks on edges in front of vertices
• Assuming subsequent vertices of the figure

14
Robots Behaviours

• Worker robots are a finite state machine that
  selects between low level behaviours
• If not holding a brick and sense a brick on the
  outside of the workspace, pick up the brick
• If holding a brick and sense a vertex robot,
  place the brick in any gap in front of the vertex
• If not holding a brick and sense a vertex
  robot, check if the next vertex is already there
• If not, assume the next vertex of the figure
• Otherwise, randomly wander the workspace

15
Forming a Vertex
Worker robot sees a vertex robot
Checks to see if next vertex is present
If it isnt, moves in front of previous vertex
Then it moves to the correct distance
Finally it turns to the correct heading
And signals that its assumed the vertex
2
4
3
16
Project Plan Week 1

• Simulate local vision and communication
• Implement low-level behaviours
• Picking up and placing bricks
• Lining up with vertices
• Collision avoidance and random wandering
• Create each robots finite state machine
• Building the structure
• Assuming the vertices

17
Project Plan Week 2

• Integrating and testing behaviours
• Implement indicators for transition states
• When work area has been cleared
• When the structure (or one of its walls) has been
  completed
• Prepare simulation environment for the building
  task
• Have robots build preliminary structure

18
Project Plan Week 3

• Testing and fixing all previous components
• Tuning behaviour parameters
• Robot speeds, vision distances and ranges
• Collision avoidance, navigation to target, lining
  up, and random wandering parameters
• Performing experiments (see next slide)
• Making additions and changes as needed

19
Experiments

• Test the algorithm on a variety of structures
• Find out how construction time and structural
  accuracy are affected by
• Number of robots
• Range of vision
• Precision of odometry
• Assess robustness by randomly introducing robot
  failures

20
Summary

• Robots can build structures where it might be
  expensive or impossible for humans
• Distributed multi-robot teams should be able to
  build such structures more efficiently
• Existing algorithms incorporate global knowledge,
  communication and many unrealistic assumptions
• Our technique should be more fault tolerant and
  applicable in realistic situations

21
References

• C. Melhuish, J. Welsby, and C. Edwards, "Using
  Templates for Defensive Wall Building with
  Autonomous Mobile Ant-Like Robots," in
  Proceedings of the 2nd British Conference on
  Autonomous Mobile Robots (TIMR-99), 1999.
• R. L. Stewart and R. A. Russell, "Emergent
  structures built by a minimalist autonomous robot
  using a swarm-inspired template mechanism," in
  Proceedings of the First Australian Conference on
  Artificial Life, pp. 216-230, 2003.
• R. L. Stewart and R. A. Russell, "Building a
  loose wall structure with a robotic swarm using a
  spatio-temporal varying template," in Proceedings
  of the IEEE/RSJ International Conference on
  Intelligent Robots and Systems, 2004.
• J. Wawerla, G. S. Sukhatme, and M. J. Mataric,
  "Collective Construction with Multiple Robots,"
  in Proceedings of the IEEE/RSJ International
  Conference on Intelligent Robots and Systems,
  2002.
• J. Werfel, "Building blocks for multi-agent
  construction," Distributed Autonomous Robotic
  Systems 6, 2004.

22
Questions?

• Thank You