arXiv · Artificial Intelligence· Huiwen Yan, Kyriakos G. Vamvoudakis, Mushuang Liu·· 3 小时前AI 评分22
Meta-Multi-Agent Reinforcement Learning for Fast Adaptation of Interactive Policies with Applications to Autonomous Driving
Meta-Multi-Agent Reinforcement Learning for Fast Adaptation of Interactive Policies with Applications to Autonomous Driving
AI 导读
summary_zh 论文提出 meta-MARL 框架，将多智能体强化学习建模为马尔可夫博弈，实现交互策略的快速适应，并定义 meta-NE 概念及梯度博弈算法的充分条件。在自动驾驶任务上的评估表明，该方法比预训练 MARL 基线适应更快，验证了框架有效性。
来源：arXiv · Artificial Intelligence · arxiv.org