Research On Iterative Learning Control Strategies For Several Classes Of Non-strict Repetitive Systems

Iterative Learning Control(ILC)is an intelligent Control algorithm which simulates the ability of human self-learning and self-improvement.By learning the difference between the actual operation and the expected value,the control of the next operation can be adjusted to approach the expected value gradually.ILC is data-driven,easy to understand and does not require much prior knowledge,and has unique advantages for control problems with highly uncertain and complex dynamical system.The dynamic performance of the traditional ILC system must be strictly consistent every time when it runs.It is the precondition to guarantee the specific control performance of the system.But in the actual control process,the controlled system is affected by various unknown factors easily.The dynamic performance changes to affect the control directly.In this paper,several new control strategies are proposed to improve the control performance of the system.The main research contents are given as follows:Firstly,an optimal iterative learning control algorithm based on quadratic error is proposed for a class of non-strictly repetitive nonlinear continuous systems with uncertain parameters and initial iteration error.The algorithm modifies the control item by calculating the difference between the expected error and the actual error,so that the tracking control can not be limited by the initial fixed value(E~2 ILC).The iterative learning algorithm is used to identify the unknown parameters of the uncertain part of the system,and the inverse optimal control based on improved Sontag formula is used to improve the convergence speed and stability margin of the controller in time domain.The simulation results show that the proposed optimal ILC algorithm has better stability and convergence in both time domain and iteration domain,and has faster tracking speed.Secondly,a fast iterative learning control algorithm based on parallel error is proposed for a class of non-strictly repetitive nonlinear continuous systems with control delays and initial iteration errors.The algorithm uses variable parameters and filters to reduce the negative effect of the initial error step by step,and does not require the initial value to be fixed.It can satisfy the requirements of fast convergence speed and small steady-state error.The control item can be corrected by using the difference between the running error of this iteration and that of the last iteration(i.e.parallel error).The oscillation instability caused by using high-order differential operation is avoided effectively.The simulation results show that the proposed algorithm has good tracking performance for the robot arm control system with time delay and initial iteration error,and the system can still maintain good robustness and real-time performance when the parameters change.Finally,an iterative learning control algorithm is proposed for a class of long-range non-strictly repetitive nonlinear discrete-time systems with two-way network transmission errors.When the data is transmitted through the network,the closed-loop control in the sense of iterative domain is adopted,and the stored historical data is used to compensate the system error.At the same time,the feedback is added in time domain.The closed-loop control improves the stability and learning speed of the system,and the robustness of the system is greatly enhanced.The simulation results show that the algorithm still has good control effect even if each component in the multivariable output or input vector has different data transmission error.