Design and Application of A Parallel Agent

The parallel agent learns the model according to the environment feedback, and selects the appropriate amount of action. It consists of automatic modeling system, automatic simulation system and parallel execution system. The automatic modeling system sends an excitation signal to the environment and establishes the model according to the system state data. The automatic simulation system automatically carries out the step response simulation experiment by using predictive dynamic optimization technology. The parallel execution system will automatically switch with the actual system without disturbance, and carry out step disturbance test. If the performance is excellent, it will be put into application. The parallel agent is divided into two processes: learning and experiment. The learning process refers to that agents collect feedback information and learn the model. The experimental process refers to that the agent constructs the predictive dynamic optimization closed-loop control system based on the obtained model and carries out the step response simulation experiment, and switches to the actual system. If the step disturbance test has excellent performance, it will be put into operation. Based on the parallel control theory, the parallel agent automatically completes the whole process according to the step sequence and has a certain adaptive ability.