Research On Quantized Iterative Learning Control

The networked control structure widely adopted in the research on iterative learning control(ILC).It has been used in practice due to its unique flexibility,convenience and robustness of the entire control framework.All involved units in the network system are usually located at different sites and communicate with each other through a wired/wireless network.The overload of these networks usually cause data congestion and data dropout,as well as other serious problems.Therefore,reducing the communication burden is of great significance for controlling through the network.In response to these problems,the quantized ILC problem that uses a finite-level uniform quantizer to reduce the burden of network transmission is studied in this paper.Aiming at the problem of data dropout and quantization based on the encoding-decoding mechanism in the unpredictable data dropout environment,the upper bounds of the quantization level of the finite-level uniform quantizer are given in this paper.An additional encoding-decoding mechanism is introduced to eliminate quantization errors.In order to ensure the consistency between the output of the decoder and the internal state of the encoder,a simple maintenance strategy is proposed.By introducing the lifting technique in discrete-time iterative learning control,that is,by lifting signals as super vectors along the time axis,the original state-space model can be expressed as a form of lifted system.The problem of quantized ILC based on the encoding-decoding mechanism in the environment of random data dropout on the measurement side and the actuator side is solved,and the more general algorithm for the quantized ILC problem in the environment of data dropout is given.The upper bounds of the finite-level uniform quantizer calculated by the existing research results are too conservative.In this paper,due to the introduction of the lifting technique,the finite-level quantizers used on the measurement side and the actuator are more accurate,which is an important improvement to the existing results.The algorithm proposed in this paper has been proved to be effective through the numerical model and the linear permanent magnet motor model.