| Research Abstract |
RoboCup proposed as an international joint research project aims at building a new paradigm of AI and robotics where broad range of research issues related to the multi-robot environment have been attacked. In our project, we have focused on learning, development, and evolutionary approaches and verified their validity through the real robot experiments. In the Osaka University group, learning and evolutionary approaches have been adopted in order to realize the basic skills and cooperative behaviors. Especially, the adaptive fitness function worked well for developmental skill improvement which has been demonstrated in the RoboCup competitions. In the Southern California University group, ***phologically reconfigure robots have been proposed and implemented. Cooperative behaviors between such modular robots were controlled by a concept of hormone. In the Zurich University group, a fundamental study on concept acquisition of object from sensori-motor map tang was conducted. Further, a basic architecture for concept learning from row sensory information was proposed based on the information theoretic constraint. These research achievements were presented in the RoboCup workshops or symposiums, and we had very constructs discussion towards the next research projects. Note that the main purpose of the project was to support the travel expenses and the individual research projects have their own research funds.
|
