| With the development of technology,more prostheses are being developed to facilitate patients.By assisting or completely replacing the power output and function of healthy limbs,it can increase gait symmetry,reduce metabolic output and improve their quality of life.However,there are still some problems in this field,such as limitation of motion angle,inaccurate locomotion intention recognition,insufficient precision of the bottom controller,and the inability of high-performance prototype to run offline without practical application potential.To solve the problem,in this paper,active-passive electrohydraulic ankle prosthesis(APEHAP)was developed,and high-precision intelligent control algorithm was studied to realize the practical application of intelligent prostheses,reproduce natural gait and improve walking comfort in tibial amputations.First,we designed an APEHAP with large angle variation and its embedded control panel.By constructing kinematic and dynamic models,we analyzed its performance and carried out functional experiments to verify the design validity.Second,based on the motion data of healthy human beings,a high-accuracy motion intention recognition model was designed as the upper controller.The overall framework of the recognition model was determined and optimized through the comparison experiment of relevant parameters of the model,so as to improve the accuracy of the recognition model.The motion intention recognition model can provide a theoretical basis for the application of intelligent ankle prosthesis intention recognition.Then,a data-driven APEHAP mathematical modeling method was proposed to determine the simulation mathematical model.A new controller was designed based on static parameters for trajectory tracking simulation and prototype experiments to verify the controllability of APEHAP in the full range.Finally,a finite state machine based on the healthy ankle joint gait phase was designed as a middle-level controller to achieve a multi-step human level walking experiment and complete the acquisition of patients’ actual demand trajectory.And then,a new bottom controller based on dynamic parameters was designed to realize optimal trajectory tracking control.The main research content of this paper is as follows:1.Design of Electromechanical Integration System of an Active-passive Electrohydraulic Ankle Prosthesis.To address the shortcomings of traditional prostheses,we designed an wide angular variations APEHAP with precise active drive control,damping adjustment,and energy recovery.The dorsiflexion angle of the APHHAP can reach 17.9°whereas its plantarflexion angle can reach 18.3°.Then,we established the kinematic model of the ankle prosthesis system and the dynamic model of the hydraulic system for the core drive to analyze the performance of the prosthesis.An embedded control board with compact structure and small volume was designed based on STM32F429 ZI that meet the requirements of system sensing,control,storage and debugging.Finally,by the performance and function experiment,the results show that the minimum damping force is 31.6N and it has continuous adjustable damping and locking function.2.Research on High-accuracy Motion intention Recognition Algorithm based on human data.To reduce the calculation cost and the extra burden on the patient’s body,a single inertial measurement unit was used to collect 9 hunman locomotion modes to build the locomotion database.We designed locomotion intention recognition model,including decision tree structure(DTS)design and proposed improved backpropagation neural network(IBPNN)as node judgment basis of DTS.By analyzing the influence of the input,the structure and the time window of the model,it is determined that the IBPNN-DTS B model based on the original value of 200-ms time window has the best recognition accuracy and F1 value(97.29% and 0.9465).Finally,we proposed ABC-IBPNN-DTS model by optimizing the model structure with neuron reduction of IBPNN and the initial weight threshold parameter of IBPNN based on artificial bee colony algorithm(ABC).The results show that the accuracy of ABC-IBPNN-DTS is 96.71% and the number of parameters shrank by 66% with only a 0.58% reduction in accuracy while the classification model kept high robustness.3.Research on Modeling and Tracking Control Algorithm of an Active-passive Electrohydraulic Ankle Prosthesis.To solve the problem of difficult measurement of mechanical static and dynamic parameters of APEHAP,the data-driven mathematical modeling method was proposed instead of the traditional physical modeling method.Through comparative experiment of four kinds of modeling methods,the piecework polynomial model with high accuracy and reflecting the motion law was selected as the simulation model.To improve control performance,we proposed improve particle swarm optimization(IGOPSO)with generalized opposition-based learning,dynamic inertia factor,learning factor and adaptive elite mutation strategy.Then,three kinds of controllers(proportional integral derivative(PID),PSO-PID and IGOPSO-PID)were designed to realize trajectory tracking simulation and experiment of slope and sinusoidal reference of dorsiflexion and plantarflexion.The results show that IGOPSO algorithm has better global search ability,faster convergence rate and can avoid falling into local optimal,the three kinds of controllers can track the reference trajectory effectively,which proves that APEHAP has the controllability in the full range,and IGOPSO-PID achieves optimal control effect in terms of root mean square error(RMSE)and maximum error.4.Research on Optimal Motion Planning and Tracking Control Algorithm Based on Human Experiment.By analyzing the gait phase changes of healthy ankle joint and combining the characteristics of APEHAP,a finite-state machine based on gait phase was designed for a multi-step human level walking experiment to obtain the patients’ actual demand trajectory.Aiming at the inertia and comfort problems in the experiment,a minimum-jerk trajectory planning method was proposed to optimize the active plantarflexion and dorsiflexion trajectory in the finite-state machine for the trajectory tracking control algorithm.And then,we proposed improve slime mould algorithm(ISMA)with random opposition-based learning,cauchy mutation strategy and chaotic search strategy.Then,three kinds of controllers(PID,Fuzzy-PID and ISMA-Fuzzy-PID)were designed to realize trajectory tracking experiment of optimal reference of dorsiflexion and plantarflexion.The results show that ISMA has better global optimization and local search ability and faster convergence rate,and ISMA-Fuzzy-PID with optimizing controller parameters achieves the best trajectory tracking effect in terms of RMSE,maximum error and convergence rate. |
