DESIGN OF A GENERIC PATH PATH PLANNING SYSTEM AILAB Path Planning Workgroup
AILAB Path Planning Workgroup 2 OUTLINE Path Planning Basics Current Implementations System Design Conclusion
AILAB Path Planning Workgroup 3 PATH PLANNING BASICS Path Configuration Work Space Configuration Space (Cspace) –Cell Decomposition –Roadmap (Skeletonization) Free, Obstacle, Unknown Space Dimension and Degrees of Freedom
AILAB Path Planning Workgroup 4 Cell Decomposition Regular Grids Multiresolution Cells Trapezoidal Cells
AILAB Path Planning Workgroup 5 Roadmap (Skeletonization) Meadow Maps Generalized Voronoi Diagrams Visibility Graphs Probabilistic Roadmaps
AILAB Path Planning Workgroup 6 Properties of Path Planners Dynamic vs. static Global vs. local Optimal vs. suboptimal Complete vs. heuristic Metric vs. topological
AILAB Path Planning Workgroup 7 Classification of Obstacles Category of Obstacles from Arai et. al. [Arai89, 28]
AILAB Path Planning Workgroup 8 Path Planning Techniques Reactive Methods –Artificial Potential Fields –Vector Field Histogram Method Graph Traversing Methods –A* Algorithm –Best First / Breadth First / Greedy Search Wavefront Method Other Methods –Wall following, Space filling curves, Splines,Topological maps, etc.
AILAB Path Planning Workgroup 9 Problems with MA-PP Possible problems of applying ordinary PP methods to MAS are, –Collisions, –Deadlock situations, etc. Problems with MA-PP are, –Computational overhead, –Information exchange, –Communication overhead, etc.
AILAB Path Planning Workgroup 10 Approaches Cenralised: All robots in one composite system. +Find complete and optimum solution if exists. +Use complete information -Exponential computational complexity w.r.t # of robots -Single point of failure Decoupled: First generate paths for robots (independently), then handle interactions. +Proportional computation time w.r.t # of robots +Robust - Not complete -Deadlocks may occur
AILAB Path Planning Workgroup 11 Improvements for MA-PP Priority assignment Aging Rule-Based methods Resource allocation Robot Groups Virtual dampers and virtual springs Assigning dynamic information to edges and vertices...
AILAB Path Planning Workgroup 12 Characteristics of MAS According to Dudek et. al. [Dudek96,53], Team Size 1, 2, limited, infinite Communication Range None, Near, Infinite Communication Topology Broadcast, Addressed, Tree, Graph Communication Bandwidth High, Motion related, Low, Zero Team Composition Homogeneous, Heterogeneous
AILAB Path Planning Workgroup 13 Characteristics of Domain Initial Information None, Partial, Complete Number of Targets 1, Many Target Available True (i.e. go to target), False (i.e. explore for target) Stationary Targets True, False
AILAB Path Planning Workgroup 14 Complexity of Path Planning In 3D work space finding exact solution is NP-HARD. [Xavier92, 54] Path planning is PSPACE-HARD. [Reif79,55] The compexity increases exponentially with, –Number of DOF [Canny88, 9] –Number of agents
AILAB Path Planning Workgroup 15 Imperfect solutions Used in case of compex problems, –Approximation –Probabilistic –Heuristic –Special cases
AILAB Path Planning Workgroup 16 CURRENT IMPLEMENTATIONS Sampling Based Algorithms –Incomplete, but efficient and practical Types –Multiple Query –Single Query
AILAB Path Planning Workgroup 17 Multiple Query A map is generated for multiple queries Fill the space adequately Probabilistic Roadmap –Uniform sampling of C-free –Local planner attempts connections –Biased sampling
AILAB Path Planning Workgroup 18 Single Query Suited for high dimensions Find a path as quick as possible RRTs –Grow from an initial state RRT-Connect : Grow from both initial and goal –Expand by performing incremental motions
AILAB Path Planning Workgroup 19 Demos Path Planning –Probabilistic Roadmap (PRM) Different sampling methods –Rapidly-exploring Random Trees (RRTs) RRT RRT-Connect
AILAB Path Planning Workgroup 20 SYSTEM DESIGN * Following slides are based on Lavelle’s Motion Strategy Library, implemented in C++
AILAB Path Planning Workgroup 21 Overview MODULES: Model Geom Problem Solver Scene Render Gui
AILAB Path Planning Workgroup 22 Model Contain incremental simulators that model the kinematics and dynamics of a variety of mechanical systems. The methods allow planning algorithms to compute the future system state, given the current state, an interval of time, and a control input applied over that interval.
AILAB Path Planning Workgroup 23 Geom These define the geometric representations of all obstacles in the world, and of each part of the robot. The methods allow planning algorithms to determine whether any of the robot parts are in collision with each other or with obstacles in the world. (PQP - the Proximity Query Package )PQP - the Proximity Query Package
AILAB Path Planning Workgroup 24 Problem This is an interface class to a planner, which abstracts the designer of a planning algorithm away from particular details such as collision detection, and dynamical simulations. Each instance of a problem includes both an instance of Model and of Geometry. An initial state and final state are also included, which leads to a problem to be solved by a solver (typically a planning algorithm).
AILAB Path Planning Workgroup 25 Planner The most important module. Base for all path planners...
AILAB Path Planning Workgroup 26 CONCLUSION Path planning is a challenging task with many different applications. Each application may device its own path planning strategy. A generic path planning library may provide solution or guidelines for other path planners....
AILAB Path Planning Workgroup 27 QUESTIONS? Thank you...