Research And Development Of Supervisory Control System Of Reconnaissance Robot

The reconnaissance robots are used to assist human perform task in hazard or dangerous circumstance by providing video/audio and other data around the robot to the operator before he arrives the accident spot. The human operators monitor the robot's surroundings and robot states by human-robot interface, and command the robot to detect the hazardous materials, grasp samples, and search for specified objects, etc. In this dissertation, several key issues about the research of reconnaissance robot, including real-time and error-resilient wireless video transmission under low bandwidth, scout mission plan, and implementation methods of mission executor and supervisory software system based on multi-agent, are studied.Firstly, the functional requirements of semi-autonomous reconnaissance robot are introduced, which followed by designing of mobile vehicle, selecting of sensors, and dynamic modeling of vehicle. Then, a layered hardware structure with following characteristics: modularization, network linkage of modules, and local intelligent, is described. A hybrid control architecture which combined the key features of hierarchical and behavior-based architecture is also introduced.In the Chapter 3, according to the restriction of bandwidth and the instability of wireless network, while taking every part of video transmission system such as video capturing, video encoding, video packaging, video transmission, and video decoding into account, an integrated, real-time and error-resilient wireless video transmission framework which is bandwidth adaptive and h.264 codec based is proposed in order to get good quality of service under low bandwidth. The realized framework is tested in wireless network which has only data transferring rate of 512 kbps and perform well.For the convenience of planning reconnaissance mission for operator at the scene of the accident, the mission plan method based on graph assemblage is brought forward in the Chapter 4. In the abstract layer, it uses RS operator to describe robot missions based on behaviors. With the developed visual mission planner, the operator can use it to plan reconnaissance mission without programming word-for-word, just need assemble the graph and synthesize robot behaviors easily. Then the generated mission description can be downloaded to the robot and be interpreted and executed.As to the mission execution system, a novel implementation method is applied using Python language. The system comprise of two parts: mission interpreter and mission executor. The mission interpreter receives the abstract mission description from the mission planner, then begins to interpret, finally makes out the mission description using Python language. While, the mission executor with Python interpreter embedded executes the output of mission interpreter. The executing mechanism is the hybrid behavior coordination mechanism which is also introduced. This mechanism makes the robot be autonomous to perform a goal-oriented mission. By using Python it becomes possible that all the robot behaviors can be stored as Python script files and be managed conveniently. We could generate new robot behaviors by constructing existed behavior script files easily. The robot mission description using Python language could be put into execution promptly, not need to be complied. The results of test and simulation show that the implemented system could interpreter and execute mission description from mission planner rightly, and it is adaptable.Since the semi-autonomous reconnaissance robot system is a large complex system, it should be constructed in opening framework with distributed intelligences and local autonomous capacities. The supervisory software system based on multi-agent is built up. The whole system is divided into robot interface agent and human-robot interface agent. All the agents in the system communicate with each other by agent link, and they follow the model of client/server, thus a client software development framework named as SAMRIA is put forward. By using it as API, programmers can focus their attention to robotic logic and the development cycle can be cut obviously. A surveillance and navigation system developed mainly with Visual C++ based on multi-agent is implemented. It is adaptable and extendable.