Perception Enhanced Robot Arm Oriented Task Learning And Planning

As an emerging kind of robot, service robot has a huge market potential. Compared with industrial robot, service robot is faced with more challenge. Environment service robot works changed frequently and the person it serves is general consumer and even disabled people. Service robot needs perception ability, task planning ability and task learning ability to server people better. So far there are many researches about it, but there still lacks of a framework which synthesizes these three abilities and serves people.Based on current situation, this paper takes robot arm for example and proposed perception enhanced robot arm oriented task learning and planning framework. This framework is enable learning a new task by demonstration of people and executed the task based real-time environment information.The main contributions of this paper are:1) Propose a task learning paradigm; then propose perception enhanced task planning a learning model based on the paradigm;2) Define a task model containing entity and environment, action and behavior; based on DFA realized task planning;3) Propose a task learning method based on demonstration; robot arm can learn a new task by demonstration using this method;4) Design and realize perception enhanced robot arm oriented task learning and planning framework based on perception enhanced task learning and planning model;5) Deploy the framework on the real environment and assess the result of framework learning; take the robot arm system as "my house service desk" and it can learn "take apple to people" and "pour water for people" by demonstration.