| Abstract: |
The information delivered by a robot's sensors always contains a high degree of uncertainty. When there are multiple communicating robots, there is an opportunity to combine sensor information to obtain a more complete model of the world. However, combining uncertain information in noisy and dynamic environments is fraught with many difficulties and can often make reasoning about the world more difficult. In this paper, we describe our attempts to combine world models of multiple robots in the RoboCup robot soccer domain. We work with Sony's Aibo legged robot. A team consists of four such robots, each of which operates autonomously, sensing its environment primarily through an onboard colour camera. Each robot is equipped with a wireless LAN card enabling it to communicate with its team mates. The robots attempt to localise themselves and other objects and construct a world model which may be shared. The problem we discuss is how the world models from other robots can be incorporated into a single model. We also discuss the dangers in attempting to do so.
|