Research On Pattern Recognition Algorithms Of Human Robot Interaction Contact State Based On Tactile Information

With the rapid development of modern industrial technology,the intelligent robot has become an increasingly important part in people's work and life.In daily life,many people would like to raise pet dogs,pet cats and so on,but it takes too much time on feeding,bathing and taking a walk.The appearance of pet robot can greatly solve these problems,pet robot can be served as an object of emotional care and emotional communication without spending a lot of time on feeding and cleaning.A pet dog can communicate with a human being through visual sensors,auditory sensors,tactile sensors and so on.This paper focus on studying the application of tactile information in human-robot contact state pattern recognition.Firstly,a data acquisition platform of pet robot based on tactile sensors is designed,then the improved BP neural network algorithm is used to classify the collected data.The experimental results show that the highest accuracy can reach 98% while the average accuracy rate can reach 95%.The main research work of this paper is as follows:(1)The acquisition platform of pet robot tactile information is designed.To simulate the tactile function of the pet robot,a data acquisition system is built by assembling K60 MCU,Bluetooth,3D printing model,FSR sensors,nano25 sensor and so on.A total of seventeen kinds of data to a pet dog robot are collected including pat twice,a heavy strike,push in one direction and push in transformation directions and so on.(2)An improved PSO algorithm is presented.As the PSO algorithm easily falls into the local optimum,a nonlinear inertial weight and learning factor combined with genetic algorithm are introduced in the PSO algorithm,and the experimental results show that this improved method can get higher precision and can balance the contradiction well between local search and global search.(3)An improved BP algorithm is proposed.Due to the choice of initial weight value is closely related to the accuracy and convergence speed in the BP algorithm.The improved PSO algorithm is used to optimize the weights in the BP algorithm.the experiment results show that in the human robot contact state pattern recognition,the improved algorithm can get a higher recognition accuracy than before.(4)The collected data processing : including filtering processing,feature extraction,classification and recognition processing.Firstly,the adaptive Kalman algorithm is used to filter the experimental data,and PCA algorithm is used to extract the feature.Finally,the improved algorithm is used to classify and identify the data.It can be showed that the accuracy of the recognition of human-robot contact state is improved.