| Abstract of "Learning Symbolic Maps from Robot Navigation": |
| Navigation in real life inspires powerful approaches used in robotics. Giving an overview of various biological navigation systems and their influence on the field of robotics, this thesis will present an extension of such an approach by introducing a knowledge-based map learning simulation. The presented system will be enlarged to mimic the biological correlates even more realistically by simulating differently skilled agents. The focus will lie on the agent's knowledge representation and therefore on the acquisition and maintenance of symbolic maps representing the environment the agent is operating in. The different skills of the agents will mainly be simulated by distorting their sensory perception that is equivalent to the implementation of measurement noise from the technical point of view. The presented system will be able to deal with some of the problems concerning map building that occur when distorted sensory input is integrated in the simulation. Moreover the performances of diversely skilled agents will be extensively tested and evaluated showing the robustness of the approach. |
|