Grid-Based Occupancy Mapping and Automatic Gaze Control for Soccer Playing Humanoid Robots

With advances in walking abilities of autonomous soccer playing humanoid robots, the world modeling and state estimation problem moves into focus, as only sufficiently accurate and robust modeling allows to leverage improved locomotion capabilities. A novel approach for dense grid-based obstacle mapping in dynamic environments with an additional application for automatic gaze control is presented in this paper. It is applicable for soccer playing humanoid robots with external sensing limited to human-like vision and strongly limited onboard computing abilities. The proposed approach allows fusion of information from different sources and efficiently provides a single consistent and robust world state estimate despite strong robot hardware limitations.