Localization is a fundamental problem in robot navigation. In this thesis we present a new cooperative localization method for the multi-agent robotic system.
Cooperative localization deals with the estimation of poses of multiple robots in associations with their sensor data. Most existing localization approaches refer to the single robot case. That is, a pose of one robot is decided by its own sensor data. In this case, all robots should have an expensive sensor that can measure the robot pose in the global frame. But if we use a advantage of multi-agent system, cooperation, localization can be done more effectively.
The system in this thesis is a heterogeneous multi-agent robotic system. One robot has a laser range finder and others have vision camera. We may think that the laser range finder has a better ability for localization than vision camera. So if the laser range finder is used to get information of others as well as of itself.
Robot soccer system is used in this paper. It gives us good features, line segments, edges. Using these features, we perform cooperative localization and show some simulations and experimental results.