Towards Learning Path Planning for Solving Complex Robot Tasks

For solving complex robot tasks it is necessary to incorpo rate path planning methods that are able to operate within di erent high dimensional con guration spaces containing an unknown number of obstacles Based on Advanced A algorithm AA using expansion matrices instead of a simple expansion logic we propose a further im provement of AA enabling the capability to learn directly from sample planning tasks This is done by inserting weights into the expansion ma trix which are modi ed according to a special learning rule For an exam plary planning task we show that Adaptive AA learns movement vectors which allow larger movements than the initial ones into well de ned di rections of the con guration space Compared to standard approaches planning times are clearly reduced