Obstacle Avoidance For Omnidirectional Intelligent Wheelchair Based On Fuzzy Bayesian Neural Network

To solve the problem of accelerating population of aging, helping disabled people reintegrate into society and improving freedom of movement, a lot of scholars have focused on the study of intelligent wheelchair's technology. The intelligent wheelchair is an integrated service tool of the mobile robot for the people whose mobility is impaired, which will confront some obstacles when it automatically moves.So to ensure safety and comfort of intelligent wheelchair's obstacle avoidance is the key technical problem. This paper presents an omnidirectional intelligent wheelchair obstacle avoidance algorithm based on fuzzy Bayesian network. The algorithm fusion avoidance algorithm based obstacle avoidance algorithm improved fuzzy neural networks and Bayesian networks achieved fully automatic to the intelligent wheelchair obstacle avoidance. The research works in this paper are listed as follows:1) Obstacle avoidance algorithm for intelligent wheelchair based on improved fuzzy neural networkIn order to solve the problem of selecting the parameters in the traditional fuzzy logic algorithm must be relied on the experience of the designers, we propose an algorithm of obstacle avoidance that can learn the parameters by itself based on improved fuzzy neural network. The algorithm in this paper has the fuzzy logic and neural network's advantages. It trained fuzzy logic parameters by neural network.Consequently, the algorithm can achieve self-learning and optimization parameters.2) Method of neural network optimization based on conjugate gradientIn order to solve the problem of learning efficiency, slow convergence and easy to form a local optimum in the traditional neural network training process, a method based on conjugate gradient was proposed. The algorithm uses the previous point gradient multiplied by the appropriate factor to give the new search direction gradient applied to the point. This method speeds up the training of the neural network convergence, reduce the computational complexity of data storage and algorithms.3) Omnidirectional intelligent wheelchair obstacle avoidance based on Bayesian networkIn order to improve the ability of intelligent wheelchair adaption to complex environment and the system's robustness, the algorithm based on Bayesian network was presented. The algorithm is added to the network right decay objective function,and uses Bayesian principles to optimize the structure and weights of the neural network. Also, the algorithm selectes neural network model using Bayes' theorem. As a result, the algorithm improves the neural network generalization ability and robust of intelligent wheelchair obstacle avoidance.Simulation and hands-on experiments show that the proposed method canoptimize the omnidirectional intelligent wheelchair obstacle avoidance path and the robustness of the intelligent wheelchair system. The method has practical application value.