Omni-directional Intelligent Wheelchair Motion Control Based On Improved FSVM And Visual Inspection

In order to assist the disabilities and the elderly,the demand of multi-functional intelligent wheelchair has grown increasingly in our society.At the same time,the key technologies related with smart wheelchairs have attracted the attention of many scholars both at home and abroad.The research of smart wheelchair has become a hot topic in the field of mobile robot.To allow users to use wheelchair conveniently and ensure the safety of users,researchers need to design the strategy of motion control and obstacle avoidance based on the characteristics of intelligent wheelchair platform.In this study,we use the intelligent wheelchair platform developed by our laboratory to optimize its motion control system and further discusse its obstacle avoidance strategy.First,based on the change of barycenter coordinates of human body in the chair face 2D projection coordinates,the system determines the desired direction of motion and speed of the user and then controls intelligent wheelchair accurately and efficiently.Considering the shortcomings of the traditional optical flow algorithm,such as large amount of data and complex calculation,we propose a monocular vision obstacle avoidance algorithm based on the deformed mesh.Finally,to solve the shortcomings of the traditional fuzzy neural network obstacle avoidance algorithm,such as the lack of autonomy in the parameter selection process,the low learning efficiency,slow convergence speed and easy to form a local optimum problem,we propose a fuzzy obstacle avoidance algorithm for intelligent wheelchair based on improved bayesian neural network,which fuses improved bayesian neural network and improved fuzzy neural network obstacle avoidance algorithm to realize autonomous obstacle avoidance of smart wheelchair.The main research works are listed as follows:1)Intelligent Wheelchair Motion Control System Based on Improved Fuzzy Support Vector Machine.To accurately determine the movement intention of uses according to projection coordinate of users in the seat bottom plane,we propose a clustering algorithm based on fuzzy support vector machine(FSVM)for the clustering of gravity center data of the human body.To avoid making wrong judgment for the motion intention of users,we should eliminate generated disturbance during moving process and interference for clustering results by data points generated by intermediate points,therefore we further optimize the FSVM clustering algorithm.In the clustering algorithm,the membership function is added to reduce the misclassification rate.2)Monocular Vision Obstacle Avoidance Algorithm Based on Deformable Grid.The traditional monocular vision obstacle avoidance algorithm needs to deal with a large amount of data when calculating the obstacle position information,and the processing time is poor.In this paper,we propose a monocular vision avoidance algorithm based on variable mesh to simplify the moving object segmentation in dynamic background to calculate the azimuth and size of distortion in the grid region.The algorithm reduces the amount of information to deal with,improves the realtime performance,and the developed system is suitable for intelligent wheelchair in the movement of obstacle detection.3)An Improved Neural Network Obstacle Avoidance Algorithm Based on Bayesian Network.Aiming at the problems of low learning efficiency,slow convergence and local optimization in traditional neural network training,an improved method based on conjugate gradient method is proposed.The method multiplies the previous point by the corresponding coefficient,adds the result to the gradient of the point,gets the new search direction,accelerates the convergence of the neural network training,which effectively reduces the storage capacity of the data and the computational complexity of the algorithm.At the same time,to enable the intelligent wheelchair apabt to the more complicated environment and improve the robustness of the system,a barrier algorithm based on Bayesian neural network algorithm is proposed.The algorithm uses the Bayesian principle to optimize the structure and weight of the neural network.At the same time,the Bayesian principle is used to select the neural network model and improve the generalization ability and intelligence of the neural network.Experiments show that the motion control algorithm proposed in this study can effectively eliminate the influence of the noise generated by the disturbance and the intermediate process on the clustering results,and reduce the control error and improve the response speed of the system.Secondly,the monocular obstacle avoidance algorithm proposed in this study can effectively identify the obstacle information and provide the correct obstacle avoidance command for the motion control system.The omni-directional intelligent wheelchair obstacle avoidance algorithm based on fuzzy Bayesian neural network improves the obstacle avoidance path of omni-directional intelligent wheelchair and improves the robustness of intelligent wheelchair obstacle avoidance system.The experimental results showed that the proposed method can improve the obstacle avoidance effect of omnidirectional intelligent wheelchair and have practical application value.