Event-triggered-based Decentralized Optimal Fault-tolerant Control Of Reconfigurable Manipulators

With the research in the field of robots becoming more and more mature and widely used,robots serving in various industries have become indispensable for human beings.However,with the progress of technology and the development of productivity,the requirements for robots put forward by human beings are getting higher and higher.Therefore,the traditional industrial manipulator which has a large size,single structure and configuration,and high cost is no longer enough to meet the requirements of specific tasks.On the contrary,reconfigurable manipulators which can perform a variety of tasks,low cost,and outstanding performance have received widespread attention.Reconfigurable manipulators which are composed of standard modules and interfaces can complete the task by changing the configuration or replacing modules according to different environmental requirements.It has been widely concerned by researchers due to "reconfigurable" and "modular" characteristics.At present,it has been extensively used in industrial manufacturing,aerospace,deep-sea exploration,military battlefield,and on-orbit assembly.As the reconfigurable manipulators are utilized in some complex and changeable environments that human beings can not directly participate in,in addition to ensuring the system accomplishes the corresponding task objectives stably,how to reduce the energy loss of the system to improve the life ability,decrease the computational complexity by the system software and hardware to improve the working performance has more important research significance.Furthermore,as for the reconfigurable manipulator system which works continuously for a long time in a harsh environment,the adapter components will inevitably break down,which will affect the stability of the whole system,and even lead to the system working in an unpredictable way.Therefore,how to optimize the control performance,energy loss and fault compensation of the reconfigurable manipulator system under an environment of limited resources and energy is an urgent problem to be solved.Aiming at the above problems,this paper studies the event-triggered-based decentralized optimal fault-tolerant control of reconfigurable manipulators.The specific research contents of this paper include:(1)Dynamic modeling of reconfigurable manipulators through joint torque sensorConsidering the practical application of dynamic model,the dynamic model of the whole system is decomposed into dynamic subsystems and modeled respectively.Based on the joint torque feedback technique,the dynamic model of reconfigurable manipulator system is established with the locally known model information.The non-affine failures of actuator and model uncertainties are analyzed in-depth,which lays a solid foundation for the decentralized control,fault-tolerant control and the compensation of model uncertainty.(2)Recurrent neural network identifier-based event-triggered decentralized optimal control of reconfigurable manipulatorsBased on the constructed dynamic model of reconfigurable manipulators,an event-triggered-based decentralized optimal tracking control approach of reconfigurable manipulators using recurrent neural network identifier is proposed.Through this approach,the optimization problem of control performance and energy loss of systems is solved.First,a recurrent neural network identifier is designed to identify the uncertainties of the subsystem model.Then,the performance index function is constructed with the identified values,system state and control law,and reasonable event-triggered conditions are designed.Based on the neuro-dynamic programming algorithm,the critic neural network is utilized to approximate the event-triggered Hamilton-Jacobi-Bellman equation.Finally,the event-triggered decentralized optimal control strategy is obtained.(3)Decentralized optimal fault-tolerant control of reconfigurable manipulators through aperiodic update strategyConsidering unexpected non-affine actuator failures,an event-triggered decentralized optimal fault-tolerant control of reconfigurable manipulators is presented,which solves the problem of trajectory tracking and fault-tolerant control based on the self-learning optimization idea.First,an event-triggered decentralized robust controller is designed to compensate for the uncertainties of the subsystem dynamic model.Secondly,aiming at the non-affine actuator faults,an adaptive fault observer is constructed to estimate the faults effectively.Next,the fault estimation value,system state and control law are combined to form an improved value function,and then the fault-tolerant control problem can be transformed into the corresponding optimal control problem.Finally,combining aperiodic update policy with neuro-dynamic programming algorithm,by utilizing the critic neural network to approach the Hamilton-Jacobi-Bellman equation,and the event-triggered-based decentralized optimal fault-tolerant control of reconfigurable manipulators is realized.