多智能体强化学习基本概念

基本概念"/>

多智能体强化学习基本概念

Multi-Agent Reinforcement Learning:Concepts and Challenges

1. Multi-Agent Settings

1.1. Fully cooperative 合作关系

e.g., 机器人协作

1.2. Fully competitive 竞争关系

e.g., 零和博弈，机器人搏斗，捕猎

1.3. Mixed cooperative and competitive

e.g., 足球赛、游戏赛（两队：队间&队内）

1.4. Self-interested 利己主义

e.g., 无人车

2. Terminologies

2.1. S，A，P

2.2. Rewards

2.3. Returns

2.4. Policy Network

2.5. State-Value Function

2.6. Multi-Agent Policy Learning

2.7. 三种架构