Learning-based control of preception for mobility

To overcome the lack of flexibility and inadequacy in performance speed of perception systems for use in real-time tasks, the authors have applied integrated learning techniques to a perception system that is based on a selective sensing paradigm. The incorporation of multiple learning algorithms at different levels provides a great deal of flexibility and robustness when different perceptual task are performed. Using a selective sensing paradigm allows the system to eliminate a large amount of nonpertinent sensory data so that processing speed is greatly increased. Such a perception system is being implemented on an autonomous mobile agent. The methodology and a preliminary example of learning within the perception system are presented