||Many researchers are working on developing robots able to interact and work with people at home. To enable the sharing and reuse of robot code between different developers and end-users, we present a service-based approach that exploits the standard web interface to create reusable robotic services. Our approach includes high-level knowledge ontology planning and low-level neural network learning strategies for robot control. In addition, several service functions, including service discovery, selection, composition, and reconfiguration, have been developed for operating these services. |
In this dissertation, the proposed high-level knowledge ontology provides not only spatial knowledge but also a guide to action for daily home tasks. With the efficient HTN planning method, the planner can exploit domain knowledge and the related techniques defined within the ontology to achieve the application task. On the other hand, a low-level recurrent neural networks (RNN) approach has been implemented for new service creation. That is, we present a procedure of programming-by-demonstration to collect the behavior sequence data of the robot as expression profiles, and then employ our network-modeling framework to infer controllers. More briefly, we present a service-oriented robotic framework to enable the rapid prototyping of robotic services.
Experiments have been conducted to verify the proposed framework, and the results show that our approach can not only be applied to robot control, but can also be used to build robotic services.