Intelligent Robot Multi-sensor Information Fusion Algorithm Research

For the limitations of single sensor system to obtain information and the complexity of the intelligent robot working environment, multi-sensor information fusion as a new signal processing technology, has high research value. Compared with the information obtained by the single sensor, the multi-sensor information fusion greatly reduces the information uncertainty and improves the credibility of information.Firstly, this thesis studies the multi-sensor information fusion level, structure and algorithms form the perspective of the system. In a number of commonly used algorithms, the fuzzy neural network has easily understood if-then rules in fuzzy system and learning ability of neural networks. Thus, information fusion algorithms based on fuzzy logic and T-S fuzzy neural network are the main research topics in this paper. The learning algorithm of fuzzy neural network is improved to simplify the calculation process of the error function gradient, reduce the network training time, enabling the network convergence process of error function more smoothly, and in-depth analysis of the above fusion algorithm in intelligent robot obstacle avoidance of application. According to NWU-RR-Ⅱ configurated ultrasonic sensor and infrared sensor and it is larger uncertainty by a single sensor measuring distance, a two-step fusion scheme is proposed to complete the robot obstacle avoidance task. The fuzzy neural network is used to the first step fusion algorithm to preprocess the same direction sensors date; the second fusion algorithm respectively adopts fuzzy logic and fuzzy neural network realization robot obstacle avoidance control, and respectively complete intelligent robot obstacle avoidance experiment with the platform of NWU-RR-Ⅱ robot. The experiment results show that both algorithms could achieve robot obstacle avoidance control, but using fuzzy neural network is more accurate and flexibility than using fuzzy logic in information fusion.