Abstract
Reinforcement learning, as an important direction of machine learning, has huge development potential due to its unsupervised and model-free characteristics. Deep Q-Learning Network(DQN) is a neural network combining Q-learning and deep learning, which can effectively approximate the Q-learning function. This paper proposes a feedforward feedback control system based on the DQN algorithm, which uses the strong generality of reinforcement learning to solve the difficult problem of feedforward controller design for nonlinear systems. A single-loop water tank liquid level control system model is built, and artificial disturbance factors are introduced to check the performance of the control system. Based on the water tank liquid level control, a simulation experiment is designed to compare the feedforward feedback control system based on DQN with a feedback control system with the same parameters. The simulation results show that the feed-forward controller based on DQN can significantly improve the control effect of the control system, verifying the effectiveness of the algorithm.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have