The Research Of Multi-sensor Information Fusion Technology Applied In Mobile Robot Localization

Multi-sensor information fusion technology is a very hot research topic recent years. It combines the development of control theory, signal process, intelligence, probability and statistics and so on. This technology means that integrating the data perceived from several sensors and then to generate more reliable, more accurate or more precise information. The fused multi-sensor information has following characteristics: redundancy, complementary, real-time character and low cost character. This technology provides a technical solution for the robot working in all kinds of complex, dynamics or unknown environment.In this thesis, we take the Multi-sensor information fusion technology as the key research point, studied on the theory and practice based on the application in robots. The main research contents are as follows.First, Introduced the Multi-sensor information fusion technology's and mobile robots' development performance and trends in china and abroad, information produced by single sensor could not meet the needs of modern mobile robots and the Multi-sensor information fusion technology begins to be used widely in the robot field.Second, Analyzed the basic principle, fusion levels, fusion method of Multi-sensor information fusion technology in detail and researched the commonly used method of Multi-sensor information fusion technology in the field of robot. Recently, the commonly used multi-sensor information fusion methods in the field of mobile robot are weighted average method, kalman filter, bayesian estimation, Dempster-Shafer evidential reasoning theory, fuzzy logic, neural network etc. This thesis mainly analyzed the mobile robot location method based on probability, according to the analysis of advantages and defects of the common location methods of modern robots, such as Kalman filter, markovian and particle filter and so on. Based on the Kalman filter algorithm, designed the mixed structure fusing data level and feature level. Make a primary simulation demonstration on Kalman filter, this method usually could realize self-location of the robot, but a strict motion model is needed, it requires a linear robot motion model, but in fact, the robot is a nonlinear system, so there are defects in using the classical Kalman filter to realize robot location, here we adopt the extended Kalman filter system to resolve the nonlinear problem.Finally, introduced the RFID location system, analyzed the advantages and disadvantages of RFID location algorithm and the Multi-sensor information fusion location algorithm, and take a simulation on them. The results shows that, the self-location algorithm fused milemeter, electronic compass, ultrasonic and RFID can reduced the error update problem of the classical RFID location method. In order to implement the self-localization of robot in the known structural environment, putting up an indoor experiment environment to test the robot.