RoboCup is an initiative to foster research activities on
multi-agent and multi-robot systems.  Researches on these systems
focus on how to build complex systems that act robustly in dynamic
environments.  RoboCup provides a standard test-bed to compare these
systems and to investigate performance of these systems from many
viewpoints.  Currently, we use soccer as an example of the test-bed,
in which individual agents and robots should act intelligently and
cooperatively in a dynamically changing environment.

  RoboCup consists of several leagues, that is, simulation league,
small-size and middle-size robot league, and legged robot league.  In
the simulation league, a participant builds a team of 11 soccer player
programs, and have a match in the simulator running on a computer
network.  Other three leagues are matches of real-robots that play
soccer.  Currently, participants of RoboCup are attacking the
following research issues.

    * machine learning under multi-agent system
    * design of an agent/robot to act reactively and intelligently
    * design of a multi-agent system to act cooperatively
    * robust mechanism of robot to act in a dynamical environment

  In addition to these leagues, other fields of research are involved
in RoboCup.  In the simulation league, a couple of researchers are
developing systems to show the situation on the field in 3D view and
generate comments of games in natural language in real time.  In these
systems, they are attacking the important issue of human-computer
interface, how to show complex phenomena by selecting suitable
aspects.