In this paper we provided a detailed description of our hardware implementation of mini-robots to play Robotic Soccer. Although we considered using off-the-shelf general-purpose robots, we found it to be less expensive to build our own. Furthermore we were able to include the exact features that we needed for our approach to Robotic Soccer.
As reported in this paper, we successfully achieved the implementation of the mini-robot players. We consider this a major step towards the development of the overall system. We are now ready to move on to the next phase. Therefore, our on-going and future work includes the incorporation of the vision system, the perception-based control of the individual agents, and the high-level strategy and behavior learning for the real robotic system. For the perception-based control, we will build upon and extend multiple robot path planning which assumes global control [Ferrari et al. 1995]. Our added challenge is that each mini-robot is controlled by a separate client. We may eventually include limited on-board sensing and intra-robot communication as in [Suzuki et al. 1995] to aid in path planning and obstacle avoidance. Since we have been working on learning issues in a simulated environment [Stone and Veloso1995, Stone and Veloso1996], we expect to use similar techniques in the real robotic system.
We plan to complete the system and participate at MIROSOT96 and RoboCup97. We hope that our research and this detailed description of our design will encourage others to participate as well.