Design And Implementation Of Brain-inspired Control Platform For Manipulator

In the field of brain simulation and brain-inspired computing,the mechanism of neural system to realize motor function has been paid much attention.A large number of physiological experiments have shown that the central pattern generator(CPG)plays a vital role in the process of human rhythmic movement.Meanwhile,the cerebellum is closely related to the coordination of human motor function and the function of learning and memory.In addition,in recent years,computing platforms inspired by the mechanism of brain cognition and information processing have become the focus of research in the field of artificial intelligence.The effect of brain-inspired computing is largely dependent on the performance of the hardware platform.The field programmable gate array(FPGA)technology,which has parallel computing capability and flexible configuration,provides a good solution for large-scale neural network simulation.Therefore,this paper aims to explore the mechanism of brain to realize motor function and bionic control of rehabilitation manipulator by designing bionic controller of different levels based on the manipulator model and building high-performance brain-inspired computing platform.First,a neural network simulation platform based on FPGA was built.After the detailed analysis of FPGA system design principles,the design scheme based on Matlab/DSP Builder is determined.Under the premise of considering resource consumption,computing speed and other factors,pipeline idea is adopted to layout the system.It is verified that the simulation platform can realize the simulation of large-scale neural network in real physiological scale.Secondly,at the level of neural network,the phase-locking mechanism of CPG network is explored.And a CPG neural network controller is designed.Based on the2-dof manipulator model,the control effects of PID,Spiking neuron-pid and CPG neural network controller are compared.The off-line training method of neural network based on lookup table is proposed,and the feasibility of CPG neural network controller is proved.Finally,the design of brain-inspired controller based on cerebellum model is explored.Based on the research of real physiological data and the reflex experiment of cerebellar motor learning function,the cerebellar microcircuit architecture model was built,and a manipulator trajectory tracking scheme based on cerebellum microcircuit learning function was proposed.The encoding and decoding methods between the moment signal of the manipulator and the discharge signal of the neuron are discussed.On the FPGA simulation platform,the hardware implementation of the cerebellar model controller is carried out,and it is proved that the designed brain-inspired controller can track the predetermined trajectory of the manipulator to a certain extent.