Study On Control Methods For Bionic Prosthetic Hand And Its Electrotactile Feedback

The human hand is a powerful tool for perceiving and interacting with the environment.Hand amputation due to disease or accident seriously affects the level of autonomy of people with disabilities,causing great difficulties in their daily lives and social interactions.As an important tool for the rehabilitation of people with hand disabilities,prosthetic hands can compensate for the aesthetics of disabled hands while restoring their missing functions in terms of movement and perception.However,traditional prosthetic hands are mostly decorative and lack of functionality.With advances in mechanical design and electronics,the emergence of active bionic prosthetic hands has improved the performance of prosthetic hands in terms of movement and feedback.However,the high degree of freedom and multi-channel sensing information of the bionic prosthetic hand also impose higher requirements on motion control and haptic feedback methods.Therefore,in this thesis,based on the status of commercially available prosthetic hands,the control and feedback methods of bionic prosthetic hands in grasping movements are investigated.The main research contents are summarized as follows.For the current problems of reduced stability and attention consumption caused by the lack of feedback pathways in the grasping control of the bionic prosthetic hand,this thesis designs a shared control method that combines the user’s motion intention with the sensing information of the prosthetic hand.The method uses a deep deterministic policy gradient network-based reinforcement learning algorithm instead of the traditional switching algorithm as the top-level controller for the target joint angle of the prosthetic hand.The neural network is trained in a self-built simulation environment,and the success rate,stability and switching method of the controller are compared and validated.The results show that the algorithm can achieve fast adaptive switching in both free movement and grasping scenarios,and realize stable force control and fast response.For the intention extraction problem and multi-degree-of-freedom continuous control problem in the motion control of wearable myoelectric bionic prosthetic hand,a neural network controller based on a nonlinear autoregressive exogenous model is proposed in this thesis.The myoelectric signals are input as exogenous variables in a nonlinear autoregressive neural network with a recursive structure to extract user intent.Subsequently,the multi-degree-of-freedom continuous target joint angles output from the network are used to complete the action execution through the lower-level controller.The neural network uses real EMG signal datasets as training samples,and achieves good performance in both single-finger independent movement and daily movement data.The results demonstrate that the algorithm can meet the requirements of the finger movements of the prosthesis in daily use.In addition,the algorithm gives a simplified suggestion for the acquisition of training data sets for prosthetic hands.To address the problem of exploring electrotactile feedback stimulation methods and control methods together,this thesis designs and builds an artificial electrotactile feedback experimental platform for bionic prosthetic hand grasping experiments.The platform consists of a bidirectional human-machine interface,a bionic prosthetic hand,and a motion control system with haptic feedback.The platform is built with two sets of hardware systems for normal subjects and amputee subjects under a unified control and communication framework.The platform realizes electrotactile stimulation feedback of grasping and objects softness discriminatinig.In addition,the experiments also verifies the ability of the platform to achieve selective electrotactile stimulation,which provides the basis for biomimic stimulation.Finally,to address the lack of embodiment in the electrotactile feedback in grasping experiments,this thesis proposes a biomimic stimulus model that provides a theoretical basis for waveform generation in electrotactile feedback stimuli.The method proposes a Touch Mime model based on the theory of neurobiology,and optimizes the selection method of transcutaneous electrical stimulation parameters.The method was applied to the artificial electrotactile feedback platform and the effectiveness of the bionic stimulation method was evaluated by a modified Box and Block Test.The results show that the biomimic stimulation method has better embodied cognition and can help the amputee to complete the grasping action faster and more accurately,which provides a new strategy for exploring electrotactile feedback stimulation.