| The reconfigurable manipulator is assembled by a series of modules with standard interfaces,that can be reconfigured to adapt to different working conditions by changing the degree of freedom and module combination.By virtue of the unique characteristics of "reconfigurable" and "modular",the reconfigurable manipulator overcomes the shortcomings of the traditional manipulator such as single configuration,poor structural flexibility and limited types of tasks,so as to meet the requirements of high openness and portability of modern production and service.However,with the development of the times,reconfigurable manipulator is more widely applied in space exploration,aerospace,deepsea exploration,disaster rescue and other people cannot directly participate in the extreme environment where energy and resources are limited.Due to the complexity and uncertainty of the task,the reconfigurable manipulator is inevitably constrained by the external environment and its own conditions,thus affecting the overall control performance of the manipulator system.Different from the previous control methods,which considered the stability and accuracy of the reconfigurable manipulator system as the main control objective,our research also need to balance the internal/external disturbance,energy consumption,calculation and transmission performance and other factors to develop a suitable control method that can improve the efficiency of resource utilization while maintaining the expected control performance.Therefore,this thesis mainly studies the modeling,optimization and control scheme of the reconfigurable manipulator system which is constrained by the external environment and its own conditions,and solves the decentralized robust zero-sum optimal control of the reconfigurable manipulator which is disturbed by the uncertain environment and the cooperative game optimal control of the reconfigurable manipulator with constrained input.This thesis takes the adaptive dynamic programming algorithm based on differential game and the event-triggered mechanism as the theoretical basis,and studies the energy consumption optimization control of the reconfigurable manipulator as the main objective.Specific research contents include:(1)Research on dynamic modeling of reconfigurable manipulator systemFor a reconfigurable manipulator system composed of n modules in series,each module constitutes an independent subsystem.The established dynamic model of the closedloop system is transformed into a dynamic model of n subsystems.A torque sensor is used to compensate the coupling torque between the joints in the reconfigurable manipulator system with uncertain environmental disturbance and constrained motor input.Based on the joint torque feedback technique,a subsystem dynamic model with friction terms and interconnected dynamic coupling terms is established and the properties of model are analyzed deeply.(2)Event-triggered mechanism-based decentralized robust zero-sum optimal control of reconfigurable manipulator with uncertain environmentA decentralized robust zero-sum optimal control method based on event-triggered mechanism is proposed to solve the problem that reconfigurable manipulator systems are susceptible to external uncertain environment when performing tasks.Considering the influence of external disturbances on the system,the decentralized robust optimal control problem of the reconfigurable manipulator system can be transformed into the optimal control problem of a two-player zero-sum differential game in which the controller and the disturbance items are participants by taking the disturbance term as a control input of the system.According to the event-triggered mechanism and adaptive dynamic programming algorithm,the critic neural network is used to approximate the event-triggered HamiltonJacobi-Isaacs equation of each subsystem,and then the optimal control strategy under the sampling state is obtained.The proposed control method not only avoids the estimation of uncertain environmental disturbance,but also saves limited communication resources by reducing the frequency of controller update.(3)Cooperative game optimal control of reconfigurable manipulator with constrained input based on event-triggered mechanismDue to the limitation of physical conditions,the output torque and speed of the motor are with maximum values,which are often accompanied by the deterioration of the system performance or even the loss of stability.Therefore,this thesis proposes an event-triggered cooperative game optimal control method for the reconfigurable manipulator system with constrained input.Firstly,the performance index function containing the global information of the system is designed,and the tracking control problem of the reconfigurable manipulator system can be transformed into the optimal control problem of the cooperative game with each subsystem controller as the participant.By introducing adaptive dynamic programming algorithm and event-triggered mechanism,the optimal control strategy is obtained by approximating the event sampling Hamilton-Jacobi-Bellman equation.The optimal control strategy of the event-triggered cooperative game proposed in this thesis is aperiodic updated based on the event-triggered condition derived from Lyapunov stability theory,which effectively reduces the updating frequency between the model and the controller and improves the utilization efficiency of system resources. |
