Research On Force/Position Control And Fault Tolerant Control For Reconfigurable Manipulators Based On Adaptive Dynamic Programming

Nowadays,robot is known as one of the key technologies that affect human production and life,which is a disruptive technology leading the third industrial revolution and global economic change.Countries around the world regard robots as the one of the national plans.China has also issued the “China-made 2025” strategy,which has promoted the development of robots to the strategic level.The robots also evolved from the traditional fixed-structure industrial robot to the different mechanical-structures and different connection-modes robots,such as flight robots,soft robots,the reconfigurable modular robots,etc.Among them,reconfigurable manipulators have become a research hotspot in the field of robots in recent years,since its wide application and outstanding performance.Reconfigurable manipulator is a kind of manipulator with standard module and interface,which can be reassembled and reconfigured in order to adapt different task requirements.On the basis of its structural characteristics,reconfigurable manipulator has been widely concerned in industry,aerospace,medical rehabilitation,military and other fields of all over the world,especially in developed countries.With the continuous progress of science and technology,the single trajectory tracking task of manipulators is not enough to satisfy demands.Reconfigurable manipulators with strong robustness,high adaptability and universality,are expected to implement more tasks in contact with complex dangerous environments.Considering more situations,the force control of the manipulator is significance same as the position control.When the end-effector of the manipulator contacts with the surrounding environment,only the position control cannot satisfy the task requirements.In addition,reconfigurable manipulators are mostly used in the environment that human beings cannot directly participate in,and its components will inevitably be scuffed or damaged,which will directly affect the stability and even safety of the manipulator.Furthermore,since the complexity of tasks implemented by the reconfigurable manipulator is constantly improved,the control accuracy and stability are not the only seizing objectives.How to balance the control performance objectives and other necessary requirements such as energy saving,robustness for unknown space interference and fault tolerance,etc.,and to construct an optimal control strategy for reconfigurable manipulators to adapt the complex and extreme environments,is the ultimate goal we have been seeking.The main contents of the thesis include:1.The background and significance of the research are described.The introduction of the research status of reconfigurable manipulators and the research progress of the control methods are summarized.2.The traditional dynamic model of reconfigurable manipulator are obtained by Newton-Euler iteration algorithm.Considering the external space constraints,the dynamic model of the constrained reconfigurable manipulator is established.In addition,a joint torque feedback technique-based modeling method is proposed.A universal dynamic model of reconfigurable manipulator subsystem is obtained by using harmonic reducer and torque sensor in joint module,and its properties are analyzed in detail.3.For the constrained reconfigurable manipulator dynamics model,firstly,the unknown environmental constraints are analyzed.And the adaptive estimation algorithm is designed to approximate the unknown constraint parameter vector and to obtained the estimated constraint torque of joint.Secondly,the multi index performance index function including the motion tracking error and the contact force tracking error of end-effector is designed.Combining with adaptive dynamic programming(ADP)and policy iteration(PI)algorithm,the approximate optimal force/position control of reconfigurable manipulator contacting with unknown environments is realized.4.Addressing the uncertain actuator failures in constrained reconfigurable manipulator system,the optimal force-position fault-tolerant control strategy is developed.Firstly,since reconfigurable manipulators is capable to changing its configuration with the task requirements,the neural network(NN)-based identifier is utilized to estimate the uncertain dynamic model,realizing the real-time control of the model free reconfigurable manipulator systems.Secondly,the adaptive fault observer is designed to estimate the actuator fault which regarded as the one of the performance indices in improved performance index function.Then,the approximate optimal force/position fault-tolerant control for constrained reconfigurable manipulator is obtained.5.Different from the existing control methods based on time-triggered mechanism,an event-triggered mechanism-based distributed neural optimal tracking control strategy is designed for the reconfigurable manipulator in aspect of uncertain environment.Firstly,a model-based robust compensator is designed to compensate the model uncertainty,and the performance index function of tracking target is established.Secondly,based on event-triggered mechanism,the control policy can only calculate and update when the system state violates the trigger conditions.On the basis of the critic NN,the event-triggered Hamilton equation can be approximately addressed,and the appropriate event triggering conditions of the joint module are designed.6.Considering the subsystem of reconfigurable manipulator with actuator failures,the decentralized near-optimal fault-tolerant controller is designed.Firstly,considering the limited output of joint module actuator,a joint torque feedback(JTF)-based decentralized near-optimal anti-windup control strategy is designed.Secondly,considering the reconfigurable manipulator subsystem with actuator failures,the Longberg state observer is developed to estimate the unknown actuator failure,a decentralized robust compensation controller based on the measurable model information is established to compensate the uncertainties of the model in real time,and the optimal tracking controller is approximated by utilizing critic NN structure.Then,an observer-compensator-critic structure-based decentralized near-optimal fault-tolerant controller is obtained.7.Finally,the works of the whole thesis are summarized,and the future research work is prospected.