Adaptive game playing algorithm based on deep reinforcement learning

Abstract

The invention relates to the field of data processing, and discloses an adaptive game playing algorithm based on deep reinforcement learning, comprising the following steps: (A) acquiring strategies for different cooperation levels; (B) generating strategies for different cooperation levels; (C) detecting the cooperation strategy of the opponent; and (D) developing different coping strategies. Theadaptive game playing algorithm based on deep reinforcement learning has the beneficial effects of using the trained detector and the strategy of different cooperation degrees to implement the existing ideas such as Tit for tat in the sequential social dilemmas; improving the scalability of the agent agent; and obtaining more intuitively better than its own competitive strategy.

Description

【Technical field】 [0001] The invention relates to the field of data processing, in particular to an adaptive game algorithm based on deep reinforcement learning. 【Background technique】 [0002] Reinforcement learning is used in various fields, from games to robot control. Traditional reinforcement learning uses tables or linear functions to represent value functions or strategies. It is difficult to expand to complex problems. Deep reinforcement learning combined with deep learning uses neural networks to extract The ability to approximate features and functions has seen some successful applications [DQN][Alpha Zero][PPO]. Prisoner's Dilemma (PD game) has always been the research focus of Matrix game (Matrix game). PDgame regards cooperation and competition as an atomic action (atomic action), but in the real world, the game is composed of a series of actions, which will be timed The temporally extended PD is called Sequential Prisoner's Dilemma (SPD). In the PD game, most...

AI Technical Summary

Problems solved by technology

[0003] In order to solve the problems in the prior art, the present invention provides an adaptive game algorithm based on deep reinforcement learning to solve the problem of poor scalability of multi-agents in the prior art

Images