Life long learning and adaptation for embedded agents operating in unstructured environments

摘要

The authors introduce a novel technique for lifelong learning and adaptation of mobile robotic agents operating in unstructured environments based on our patented fuzzy-genetic system (British patent 99-10539.7). The life long learning strategy tunes the controllers learnt in our previous work by a process of gradual improvement and adaptation to the surrounding environment. We have applied this work to two different unstructured environments: the agricultural environment and the intelligent buildings environment.