Research On Tremor Elimination Algorithm Based On Broad Learning System In Teleoperation

In recent years,teleoperation system can assist the operator to finish the operation task remotely,which has the advantages of remote operation,master-slave tracking and hand-eye coordination.These advantages make the tele-operating system have the broad application prospect in the industrial field,the medical field,the aerospace field,the deep sea exploration field and the space field and so on.However,the development of teleoperating system is restricted by its disadvantages include accuracy,timeliness,reliability and security.In the process of controlling the main terminal equipment,the vibration generated by the operator's arm will affect the accuracy of the slave manipulator,so as to affect the accuracy,security and reliability of the teleoperating system.Therefore,how to eliminate the interference of physiological tremor to the control of mechanical arm from the end becomes a key problem in remote operation.In the field of medicine,physiological tremor is defined as a non-stationary,rough,and inevitable low-frequency vibration inherent in normal motion.Different from pathological tremor,the frequency of physiological tremor exists in all normal daily behaviors,and the general frequency range is in a certain frequency band,such as frequency band 8-15 hz.In the teleoperating system,the operator inputs the operation signal in the main device and sends the signal to perform the operation from the mechanical arm of the terminal through certain communication methods.During this process,the operator's hand has an inevitable physiological disturbance signal that will eventually lead to a tremor at the end of the robotic arm.Aiming at the above problems,this paper studies the methods to eliminate physiological tremor in remote operation,including the mathematical model of tremor compensation and the prediction and estimation algorithm for physiological tremor.In recent years,Broad learning system(BLS),a new neural network architecture,has been proposed,which has potential application in various fields.Based on this new neural network architecture,this paper improves the modeling and estimation of tremor and uses the tremor value at the first N moments to predict the tremor value at the next moments.In the strategy of signal compensation,the estimated predicted tremor value is compensated to the actual disturbed signal to eliminate the influence of physiological tremor on the accuracy,safety and reliability of teleoperation.In this paper,the wavelet neural network and the broad learning system are organically combined to propose a wavelet braod learning system to model and estimate the tremor.Furthermore,considering that the quaternion algorithm can take advantage of the coupling between different dimensions in multi-dimensional signals,this paper proposes a quaternion broadd learning system based on the quaternion neural network to improve the performance of the system.In the previous work,the features of the data are extracted in the time domain,frequency domain and even quaternion domain.In the further work,we introduce the fuzzy domain and innovatively proposed a lightweight three-domain fuzzy wavelet broad learning system.In the simulation part,we take a tremor time series signal to test and compare the performance of the new algorithm.Finally,we can verify the effectiveness of the new method from the results.