A Closed-Loop Brain-machine Interface System Based On Single-Joint Information Transmission Model And Model Predictive Control

Brain-machine interface(BMI)system can control external devices by constructing information transmission pathways between brain and external devices.The system mainly includes brain,decoder,encoder and external device,which provides an effective method for the recovery of impaired motor function and is of great significance for research.The main links of the system include the brain,decoder,encoder and external device.In this paper,based on the single-joint information transmission(SJIT)model,decoders and encoders are designed and compared,a closed-loop BMI system with the auxiliary controller is formulated,and an improved SJIT model is purposed.The details are as follows.1)Aiming at the problem of constructing the information pathway from brain to artificial prosthesis,this paper designed decoders based on the Weiner filter,Kalman filter and BP neural network respectively to convert the average firing rate of cerebral cortex motion-related areas into the joint torque information acceptable to artificial prosthesis.Among them,the decoder training set and testing set are obtained by SJIT model.In addition,the offline and online performance tests of the above three decoders show that the Wiener filter decoder has the best performance.2)Aiming at the general poor motor function recovery effectiveness of existing BMI systems,firstly,the information encoding efficiency of the integrated firing neuron model with synaptic connections and the randomly connected Izhikevich neuron model are compared,and the last one with higher efficiency is selected to design the encoder and build the artificial feedback pathway from the outside world to the brain.Secondly,based on the model predictive control strategy,an auxiliary controller that can solve the feedback information in real time is designed to construct a closed-loop BMI system that can effectively recovery the limb motor function,and an adaptive decoder is introduced to further improve the recovery effectiveness of motor function.3)To solve the problem of model mismatch existing in the SJIT model,several neuron populations are introduced in this paper to improve it,and the effectiveness of model improvement and the designed BMI system on the basis of the improved model are tested.The simulation results show that the improved model can overcome the SJIT model mismatch problem to some extent.On the basis of the improved model,the closed-loop BMI system designed in this paper can still recover arm motor function very well.In this paper,based on the research of decoder,encoder and auxiliary controller,the closed-loop BMI system is formulated,and the motor function recovery effectiveness of the closed-loop BMI system is further improved by designing adaptive decoder and improving SJIT model.In general,the closed-loop BMI system proposed in this paper is composed of different link combinations and has strong universality and flexibility.The research results in this paper can provide theoretical basis for the design of the closed-loop BMI system and have a strong promoting effectiveness on the practical application of BMI system.