A deep reinforcement learning method and equipment based on a plurality of historical optimal Q networks

Abstract

The invention provides a deep reinforcement learning method and device based on multiple historical optimal Q networks for an intelligent robot interaction system comprising an agent, and the method comprises the steps: defining the attributes and rules of the agent, defining the state space and action space of the agent, and constructing or calling an agent motion environment; selecting an optimal plurality of Q networks from all historical Q networks based on the level of the interaction evaluation score; combining the plurality of historical optimal Q networks with the current Q network byusing a maximum operation, guiding an agent to select an action strategy, training parameters of a learning model, and autonomously carrying out a next decision action according to an environment where the agent is located. According to the method, a reasonable motion environment can be constructed according to actual needs, the optimal Q network generated in the training process is used for better guiding the intelligent agent to make decisions, the purpose of intelligent strategy optimization is achieved, and the method has a positive effect on development of robots and unmanned systems in China.

Description

technical field [0001] The invention belongs to the technical field of computer artificial intelligence, and in particular relates to a deep reinforcement learning method and equipment based on multiple historical best Q-networks in an intelligent robot interactive system. Background technique [0002] In recent years, reinforcement learning has been widely used in artificial intelligence because of its excellent decision-making ability, so it is often used in the interactive system of intelligent robots. In reinforcement learning (RL), an agent seeks an optimal policy to solve a continuous decision problem by optimizing the accumulated future reward signal (see [1]). Over time, many popular reinforcement learning algorithms have been proposed including Q-learning (see [2]), temporal difference learning (see [3]), and policy gradient methods (see [4]). However, the reinforcement learning (RL) algorithms of these early methods mainly rely on manual feature extraction, and th...

AI Technical Summary

Problems solved by technology

Based on the above-mentioned inventive concept, aiming at the existing intelligent robot interactive system, the present invention proposes a deep reinforcement learning method based on multiple historical best Q-networks, which makes the training process more stable and solves the problem of late training in the prior art. , the lower the interaction evaluation score is for technical issues

Images