Method for controlling intelligent equipment based on reinforcement learning strategy

Abstract

The invention belongs to the technical field of control and reinforcement learning of artificial intelligence, and particularly relates to a method for controlling intelligent equipment based on a reinforcement learning strategy. The method comprises the following steps: obtaining an environment parameter set of the intelligent equipment in a current disturbance environment, taking the environment parameter set as the disturbance environment of the intelligent equipment, and marking the environment parameter set to obtain an environment parameter set with a label; inputting the environment parameter set with the label into a pre-trained antagonism generation network to obtain a brand new environment parameter set; according to the brand-new environment parameter set, updating the current reinforcement learning strategy to obtain an updated reinforcement learning strategy adapted to the brand-new environment parameter set, and inputting the updated reinforcement learning strategy to the intelligent device; and enabling the intelligent equipment to execute an action corresponding to the current state of the intelligent equipment according to the updated reinforcement learning strategy to complete control of the intelligent equipment i0n the disturbance environment.

Description

technical field [0001] The invention belongs to the technical field of artificial intelligence control and reinforcement learning, and in particular relates to a method for controlling an intelligent device based on a strategy of reinforcement learning. Background technique [0002] Reinforcement learning is one of the core technologies of artificial intelligence. Through continuous interaction with the application environment, it can learn the optimal strategy and provide intelligent decision support for intelligent devices such as robots and unmanned vehicles. Public materials show that AlphaGo has been trained using reinforcement learning before the public competition. However, there are always differences between the training environment and the real environment, and the training environment and the real environment cannot be completely consistent due to the existence of noise. Therefore, how to reduce the impact on training due to the difference between the training en...

AI Technical Summary

Problems solved by technology

However, the existing methods cannot effectively control the noise disturbance. Excessive disturbance will cause reinforcement learning to fail to train the optimal strategy, resulting in training failure; and limiting the disturbance will lead to less robust training.

Images