Two-Dimensional Model-Free Optimal Tracking Control for Batch Processes With Packet Loss

Aiming at the 2-D batch processes with unknown parameters and packet loss under the network transmission, a novel data-dependent 2-D off-policy algorithm with compensation of packet loss is developed for realizing the precise and timely control. First, in order to better compensate for the negative impact of data loss on the system performance, a Smith predictor combined with preceding 2-D data is introduced into the design of the controller. Second, an optimization performance index is defined and a 2-D optimal tracking control problem for the batch process with data loss is constructed. Furthermore, a completely data-driven 2-D off-policy
Q
-learning algorithm is derived by taking into account the reinforcement learning theory and the dynamic programming method. Only the data in time and batch directions are used to acquire the optimal controller gains, and the data-driven optimal tracking control of batch processes with data loss can also be realized. Ultimately, the unbiasedness and convergence of the proposed algorithm are demonstrated by a strict proof process, and its effectiveness is verified by simulation experiments.