Exoskeleton robot power-assisted control system and method based on neural network

Abstract

The invention aims to provide an exoskeleton robot power-assisted control system and method based on a neural network. The exoskeleton robot power-assisted control system is applied to elbow joints and comprises a sensing detection module and a data collecting and processing module. According to the exoskeleton robot power-assisted control system based on the neural network and the method, sensorssuch as a myoelectric sensor and a bending sensor are used for data fusion through the radial basis neural network, and motion power-assisted auxiliary is carried out on the aged people and patientswith motor dysfunction to meet the requirements of daily life of the aged people and the patients with motor dysfunction. According to a control strategy, a torque estimation index based on a myoelectric signal and an energy estimation index based on the myoelectric signal are adopted to objectively and quantitatively evaluate the motion power-assisted effect.

Description

technical field [0001] The invention relates to the field of information technology of wearable exoskeleton equipment. Background technique [0002] In contemporary society, our country is facing increasingly serious social problems such as aging population, younger incidence of stroke, and frequent traffic accidents. The result is motor dysfunction for the elderly and physically disabled patients. Therefore, upper limb rehabilitation robots have become a medical Research priorities and hotspots in the field of rehabilitation. Exoskeleton robots provide a feasible way out to solve these problems. They form an intelligent hybrid of human-machine coordination by wearing them in parallel on human limbs, and combine the wearer's motion intentions to perform assisting or rehabilitation training tasks. It has broad application prospects and important research significance in the fields of medical rehabilitation and national defense security. In the research of exoskeleton robots...

AI Technical Summary

Problems solved by technology

[0003] In the Chinese invention patent application with the publication number CN102799937A, a method for predicting the trajectory of lower limb movement based on the fusion of EMG signals and joint angle information is disclosed. By extracting the EMG characteristic parameters of the six muscles of the lower limbs during exercise, the root mean square value and the spatial three-dimensional Using coordinates to build a model and calculate the angle signal of the side leg and knee, the operation is complicated, and the error in the middle link is relatively large, so the feasibility is not strong

[0004] In the Chinese invention patent application with the publication number CN103440498A, a surface electromyographic signal recognition method based on the LDA algorithm is announced. The whole process needs to collect the electromyographic signal, perform overlapping windowing, dimensionality reduction, average calculation, and input into the classifier. It takes a lot of time, and the onlineness and simplicity of the algorithm are difficult to guarantee

[0005] In the Chinese invention patent application with publication number CN106109174A, a rehabilitation robot control method based on myoelectric feedback impedance self-adaptation is disclosed. The control method described in this invention combines the characteristic value of myoelectric signal, plantar pressure signal and angular velocity signal. Identify the joint extension and flexion state of the training object, determine its subjective movement intention, and establish a method for adaptively adjusting impedance parameters and adjusting static balance force in stages, making it difficult to achieve continuous compliant control in the entire control process

[0006] However, for the control technology and strategy of exoskeleton robots, there are still the following deficiencies: (1) Although the accuracy of offline classification has reached a higher level of more than 90%, the real-time rate of online and the simplicity of the algorithm cannot be guaranteed. , and the control strategy based on the classifier still needs to consume a lot of training time (2) The traditional control strategy based on the classifier, due to the discreteness of the classification results and the switching of the classification mode, the output of the controller is not completely continuous. The control effect of the skeletal robot is not stable enough, and the comfort of the human-computer interaction process is poor

(3) The control quantities in the power assist control strategy, such as the estimation of joint force, torque and angle, are too complicated, which increases the difficulty of operation; and most of the process uses empirical parameters, and the accuracy of the model is not enough

Images