Based On Information Fusion And Multi-actor Decision-making Autonomous Robot Navigation System

With the development of mordem industry, robots have been applied widely in all fields such as social production, defense industrys, service trade, intellectrual education and so on. Automobile robot ,which is an integrated system containing all kinds of functions by which the robot can perceive the changes of its environmentjnake plan and decision dynamically, execute predefined decision and control the robot's behaves, has been one of important research branches of robotics. Many experts and engineers from different countries are more and more intrested hi this research field.In automobile robotics the key intellectual techonologies are the automatic plan techonique and the intelligence of sensors . The machine's intelligence includes perception of knowledges, presume, feeling, recognition, reasoning, induction, deduction, plan, response, learning and problem solution etc. The fields it covers include image perception, language and character symbol's treatment and perception and the knowledge's expression, acquisition , learning, intelligent motion.The research of an automobile robot, which is based on multisensor' s information fusion with a velocity vector model and multibehave decision made by an expert system is the main content of this paper. At the beginning of this thesis , the development of research on automobile robot is presented. The main present research methods are discussed and the meaning of automobile robot is emphasized . Then some principles and techoniques used hi the development of this system are presented. After that we discussed three control architectures : the control architecture based on functions , the control architecture based on behaves and the hybrid control architecture and their structures. We also put forward a mutisensor's information fusion arithmetic based on velocity vector mode .La this part , A velocity vector mode based on sonar's data procession is made by fuzzy control theories.To modify this soanr's velocity vector mode and get a more reliable mode, we introduce the realtime image procession into the construction of velocity vector. For this purpose, we analysis the choice of the HSI color mode and SCT color mode and design a template forimage edge segmentation . In this way , we can obtain the reliable visual information of environment , realize the information fusion of sonar data and vision data and provide reliable reason for higher control level, that is , the behave schedule level.In this paper , a location method of robot based on map matching is dwelled on and we dwell on the robot's high level plan based on expert system.In conclusion, we give the results of experiment , meanwhile , we summarize our work and put up our goals for the futher research.