Simulation Tools for Real Robots

RoboCup Special Interest Group (SIG)

Overview

• Design of the system and
• Rapid prototyping of the control.

The design stage involve modelling the whole system, including the environment like a soccer field and the constraints, like the rules of a game, but also the robot design, i.e., what size, locomotion system, shape for the robot, where the sensor should be set, what kind of sensor should be used, etc. A good simulation of sensors is a very important issue. Using simulation tools saves a lot time in this design process and thus allows reseachers to test several possibilities before they choose one.

The development of control stage is the programming part. Using simulation tools is often easier than using directly real robots. Real robots needs to be physically present in their environment (that is several robots, built, fully working, with full batteries, existing physical soccer field, etc.). This often prevents several researchers to work at the same time on the same project. Using simulation tools allows each researcher to setup his/her own robot soccer environment in minutes and share it with colleagues as one would copy files. Moreover, the CPU resources of a regular computer is often much higher than the CPU resources of a mobile robot. Hence computer expansive techniques, such as genetic evolution, learning, neural networks can be bred on the computer simulation and moved to the real robot after it has converged to a good solution. Finally, developers are more efficient on regular computers, due to the presence of several programming tools including debuggers, development environment, visualisation facilities, etc.

However, to be useful, simulation tools have to be as much realistic as possible, to facilitate the transfer of simulation results to the real robots. This is why a simulation tool has to be designed with this tranfer capability in mind. This involves simulation quality as well as cross-compilation tools.

Finally, simulation tools may have different characteristics to match different needs: a simulator for kinematic models may be good to work on coordination, but not on control or behavior definition, activities that need the ability to simulate dynamical models.

General Goals

• Maintain a web page with a number of information:
  • Overview
    
  • Comparative table of simulation tools for real robots. [TO BE DONE]
  • Advantages / disadvantage of using simulation tools. [TO BE DONE]
  • Download: papers, user contributions, etc. [TO BE DONE]
  • Links to resources, including existing simulation tools, robot projects using simulation tools, robot soccer projects using simulation tools, related topics, etc. [TO BE DONE]
• Run a mailing list to discuss that topic: simtools@robocup.biglist.com.
• Set up a workshop at RoboCup symposiums with some demos / presentations on simulation tools for real robots.
  

Next Meeting

Preliminary program

• Oliver Obst: 3D Simulator used for Soccer Simulation League
• Olivier Michel: Webots 4: Physics Simulation and Transfer to Real Robots
• Thomas Röfer: SimRobot and common formats for simulated worlds and robot controllers
• Javier Ruiz-del-Solar and Juan Zagal: UCHILSIM: A Dynamically and Visually Realistic Simulator for the RoboCup Four Legged League
• Daniel Polani: XGrabtor: a Design Tool for Agent Strategies for Education and Research - Experiences

Meeting Reports

• RoboCup 2003, Padova.

Comparative Table

Committee Members

• Ansgar Bredenfeld, Fraunhofer Institute AIS, Germany, bredenfeld@ais.fraunhofer.de
• Brett Browning, Carnegie Mellon University, USA, brettb@cs.cmu.edu
• Augusto Loureiro da Costa, Salvador University, Brazil, loureiro@unifacs.br
• Tom Howard, TIBCO Software, Australia tomhoward@users.sourceforge.net
• Pieter P. Jonker, Delft University of Technology, The Netherlands, pieter@ph.tn.tudelft.nl
• Alexander Kleiner, University of Freiburg, Germany, kleiner@informatik.uni-freiburg.de
• Olivier Michel, Swiss Federal Institute of Technology in Lausanne, Switzerland, Olivier.Michel@epfl.ch
• Roozbeh Mottaghi, Sharif University of Technology, Iran, mottaghi@ce.sharif.edu
• Simon Parsons, City University of New York, USA, parsons@sci.brooklyn.cuny.edu
• Oliver Obst, University of Koblenz, Germany, fruit@uni-koblenz.de
• Daniel Polani, University of Hertfordshire, UK, d.polani@herts.ac.uk
• Patrick Riley, Carnegie Mellon University, USA, Patrick.Riley@cs.cmu.edu
• Thomas Röfer, University of Bremen, Germany, roefer@tzi.de
• Shahab Tasharrofi, Iran University of Science and Technology, Iran, shahabt2000@yahoo.com
• Bas Terwijn, University of Amsterdam, The Netherlands, bterwijn@science.uva.nl
• Changjiu Zhou, Singapore Polytechnic, Singapore, zhoucj@sp.edu.sg

Mailing List

Other SIGs