Networked Robot System Based On User Intention Recognition

With the rapid development of the Internet and networked robot technology in the latest years, researches on networked robots have become one of the most concerning issues of the theoretical and applications researches on robotics. However, the real-time communi- cation ability and the operation performance of the system is not very satisfying resulting from the random network delay and the errors within data transmission during the communication through the Internet. The uncertainty of the network leads a negative effect on the stability and performance of the system.To improve the previous control and operation methods of networked robot, the thesis proposes a novel approach, which implements modeling and recognition of the user's intention into Teleoperation system of networked mobile robot, according to which the negative effects to system performance due to the Internet uncertainty can be overcome. The main contributions of the thesis are summarized as follows:(1) Establishment of the Teleoperation Experimental PlatformThe Teleoperation Experimental Platform based on Web browser technology is established to satisfy users'demands of the easy operation, the multimedia interaction and the platform crossing feature of the tele- operation system. The experimental platform could fulfill requirements of testing control methods for teleoperation, completing tasks for network mobile robot experiments, and being open to Internet users.(2) The Two-Layer Model of User's Intention in TeleoperationA two-layer model of user's intention is proposed. The two layers are Action Layer and Task Layer. The action layer defines the behaviors that user want the robot to do during the operation, while the task layer focuses on describing user's expectation of tasks carried out by the robot. The model is the fundamental of intention recognition.(3) Modeling and Recognition of User's Action Layer Intention RecognitionThe thesis analyzes and models the user's intention of action layer according to the two-layer model. During teleoperation, the results of recognition become an important source of information, which influences the robot's scheduling of behaviors. The negative effects of Internet uncertainty are alleviated by employing the results of recognizing user's intention of action layer. Experimental results demonstrate the validity and feasibility of the proposed method.(4) Modeling and Recognition of User's Task Layer Intention RecognitionA recognition approach focusing on user's intentions of the Task and Action layers is proposed, specialized for the mobile robots working in an indoor environment. A direct connected graph is adopted to help modeling user's Task Layer intention. By dividing a complicated trajectory into several simple traces, the robot is able to recognize them and find the user's expected task using a trajectory based recognition method. The Player/Stage simulation results prove the positive effects to the teleoperation system by implementing the proposed method in a cruising simulation restricted in an apartment map.