Motion Planning for Soccer Playing Robots

Carlos Ernesto Guestrin








 
Abstract
Introduction
Objectives
Potential Fields
The Simulator
The Applet
Results
Conclusions and Future Work


Abstract
In this paper,


1. Introduction

This paper presents the foundations and results of the application of potential fields for motion planning in soccer playing robots.  The figure bellow shows a typical configuration of a robot soccer environment.
 
 

Soccer Field





This application environment has certain key characteristics, including:
Very dynamic environment: the configuration of the robots and ball changes constantly and rapidly in time.
Simple workspace: the workspace is simply composed of a few convex polygons  (the robots) and a circle (the ball).
Two degree of freedom actuation against three degree of freedom robots: the robots are actuated through linear and angular velocity commands, while their state is defined by three variables: two for position and one for orientation.
High-dimension configuration space for the team: if motion planning is performed for the team as a whole, the dimension of the configuration space is high, three times the number of robots.
Continuos variables: the state of the world is defined by continuous variables.
Based on these characteristics, the approach chosen has the following features:
Plan for each robot individually: the low level planning is performed for each individual robot of the team, based on the current state of the system (configuration of all robots and the ball).
Use potential fields for motion planning: a set of potential fields will be associated with each robot; the motion of the robot will be determined by the resultant of these fields.
Work around the assumption of a very dynamic environment:  a concern about local minima, which is usually associated with potential fields, is not very significant in this case, because the environment changes constantly, thus, making local minima unlikely.


This project was conducted as part of CS326a Motion Planning class taught by Jean-Claude Latombe.

The RoboCup is an international competion for soccer playing robots. Related work can be found at the RoboCup website.

2. Objectives
 

3. Potential Fields
 

4. The Simulator
 

5. The Applet

Run the SoccerSimulator applet. It uses Java 1.1, so you will need Netscape 4.5 or latter to run it.

6. Results
 

7. Conclusions and Future Work
 
 

Comments, questions and suggestions, please contact guestrin@cs.stanford.edu.