Research On Clamping Force Feedback Of Minimally Invasive Surgical Robot From Hand

At present,minimally invasive surgical robot systems,such as Da Vinci system,lack an effective clamping force feedback mechanism.Doctors cannot perceive the force between surgical instruments and human tissues,and they still need repeated training to adapt to the actual surgical operation.However,it is prone to the phenomenon of tissue clamping or insufficient clamping force.Because of the small size of surgical instruments,it is not convenient to install additional force sensors,and the preoperative disinfection will also damage the sensors.With the development of deep learning neural network,it is possible to fit the clamping force of instruments only by some parameters of the driving motor.From this point of view,this paper proposes a new prediction model based on neural network,which can estimate the clamping force of instrument fingers well,enable doctors to perceive the force during operation and improve the operability of minimally invasive surgery.The experimental equipments used in this project are all independently developed by the laboratory.Firstly,the structural analysis of surgical instruments and the dynamic modeling of wire rope transmission are carried out,the forward and inverse kinematics calculation of the armed arm part of the slave hand is carried out,the degree of freedom is allocated between the master hand and the slave hand equipment,the workspace of the armed arm is solved,and the master-slave control program is written,which basically realizes the master-slave remote operation.By analyzing the clamping process of surgical instruments,the LSTM algorithm is selected as the prediction algorithm,and the relevant superparameters are set.The MLP algorithm is used for comparison experiment,and the training set and test set are collected and made.The algorithm is written and the network is trained under tensorflow environment.The prediction performance of the network model is tested offline,and the influence of clamping track,clamping speed and object hardness on the prediction effect is analyzed.Combining the fixed network model with the master-slave control program,the model can realize online prediction,and the calculation results can be fed back to the master hand device.The thread pool technology is used to ensure the online prediction effect of the network and reduce the program delay.Aiming at the problem of instrument clamping state identification,a solution based on SVM algorithm,random forest algorithm and convolutional neural network algorithm is proposed,and the improved BMFLC algorithm is used to filter the hand signal to reduce the influence of hand jitter on the prediction effect of the model.A simple experimental platform was built for surgical instruments.Relevant data were collected on the experimental platform to make training sets and test sets,and the functions of position following and posture mapping of master and slave hands were tested.The online running effect of the prediction model was tested with force sensors,and the prediction effect under different postures of instruments was tested.The effect of the main hand force feedback was tested with Flexi Force thin film force sensors.The transparency of the system was analyzed,and finally the validity of the prediction model was verified.