Path Planning for Multi Agent Systems by ikn20172

VIEWS: 133 PAGES: 21

									 Path Planning for
Multi Agent Systems

        by
    Kemal Kaplan
    Multi Agent Systems (MAS)

   A multi-agent system is a system in
    which there are several agents in the
    same environment which co-operate at
    least part of the time.
   Complexity of the path planning
    systems for MAS (MASPP) increase
    exponentially with the number of
    moving agents.
           Problems with MASPP

   Possible problems of applying ordinary
    PP methods to MAS are,
       Collisions,
       Deadlock situations, etc.
   Problems with MASPP are,
       Computational overhead,
       Information exchange,
       Communication overhead, etc.
     Classification of Obstacles
                                             Usually other
         OBSTACLES                            agents are
                                              modelled as
                                              unscheduled,
STATIC            MOBILE
                                              non-negotiable,
                                              mobile obstacles
     NEGOTIABLE
                         NON-
                      NEGOTIABLE
                                              in MASPPs.

                                             Category of
            SCHEDULED       UNSCHEDULED
                                              Obstacles from
                                              Arai et. al. (89)
         Proposed Techniques

   Centralised Approaches
   Decoupled Approaches
   Combined Techniques
        Centralised Approaches
   All robots in one composite system.
    + Find complete and optimum solution
        if exists.
    + Use complete information
    - Computational complexity is
        exponential w.r.t the number of robots
        in the system
    - Single point of failure
        Decoupled Approaches
   First generate paths for robots
    (independently), then handle
    interactions.
    + Computation time is proportional to
        the number of neighbor robots.
    + Robust
    - Not complete
    - Deadlocks may occur
         Combined Techniques

   Use cumulative information for global path
    planning, use local information for local
    planning

            “Think Global Act Local”
    Utilities For Combined Techniques

   Global Planning Utilities:
        The aim is planning the complete path from
         current position to goal position.
        Any global path planner may be used.
         (e.g. A*, Wavefront, Probabilistic Roadmaps,
         etc.)
        Requires graph representation achieved by
         cell decomposition or skeletonization
         techniques.
    Utilities For Combined Techniques
                    (II)
   Local Planning Utilities:
        The aim is usally avoid obstacles. However,
         cooperation should be used also.
        Any reactive path planner can be used.
         (e.g. PFP, VFH, etc.)
        No global information or map representaion
         required. Decisions are fast and directly
         executable.
      Improvements for Combined
             Techniques
   Priority assignment
   Aging (e.g. the forces in a PFP varies in
    case of deadlocks)
   Rule-Based methods (e.g. left agent first,
    or turn right first)
   Resource allocation (leads to suboptimal
    solutions)
        Improvements for Combined
             Techniques (II)
   Robot Groups
       A leader and followers
       Many leaders (or hierarchy of leaders and
        experience)
       Virtual leader
   Virtual dampers and virtual springs
   Assigning dynamic information to edges
    and vertices
      Possibe MAS environmets for
                MASPP
   Robocup 4-Legged League
   Robocup Rescue
   SIMUROSOT, MIROSOT (?)
   Games (RTS, FPS)
   ...
    MASPP Example [ARAI & OTA 89]

   Measures
       Computational Load
       Total length of the generated trajectories
       The radius of curvature of the generated trajectories
       Total motion time

   Preferred measure is the first one
    MASPP Example [ARAI & OTA 89]

   Properties of agents
    MASPP Example [ARAI & OTA 89]

   Problem 1
    MASPP Example [ARAI & OTA 89]

   Problem 2
    MASPP Example [ARAI & OTA 89]

   Virtual Impedance Method
MASPP Example [ARAI & OTA 89]
MASPP Example [ARAI & OTA 89]
     Questions?
kaplanke@boun.edu.tr

								
To top