A view direction planning strategy for a multi-camera vision system

A multi-camera view direction planning strategy for mobile robots is discussed. Two concurrent tasks are considered: self-localization and object tracking. The approach is to assign the different tasks to different cameras, such that for each task an individual optimal view direction is selected based on the information gain maximization. Thereby, the individual task performance is significantly improved. The performance of the proposed strategy is evaluated in simulations, considering a humanoid robot navigation scenario, and compared with another two coupled or partly coupled gaze control strategies.