arXivDaily arXiv每日学术速递 周一至周五更新
北美退伍军人节(Veterans Day),12 日无更新
arXiv 2206.06009cs.LGcs.AI
收藏

面向快速策略迁移的相对策略-转移优化

Relative Policy-Transition Optimization for Fast Policy Transfer

Tencent Robotics X(腾讯机器人X实验室);The Chinese University of Hong Kong, Shenzhen(香港中文大学(深圳))

Jiawei Xu, Cheng Zhou, Yizheng Zhang, Baoxiang Wang, Lei Han

AI总结:

针对两个 MDP 间的快速策略迁移问题,提出 RPTO 框架,通过 RPO 优化跨环境策略、RTO 缩小动力学差距,并在 MuJoCo 任务上验证有效性。

摘要:

我们研究两个 Markov 决策过程(MDP)之间的策略迁移问题。基于强化学习中已有的理论结果,我们引入一个引理来衡量任意两个 MDP 之间的相对性差距,即在不同策略和环境动力学上定义的任意两个累积期望回报之间的差异。基于该引理,我们提出两种新算法,分别称为 Relative Policy Optimization(RPO,相对策略优化)和 Relative Transition Optimization(RTO,相对转移优化),它们分别实现快速策略迁移和动力学建模。RPO 将在一个环境中评估过的策略进行迁移,以在另一个环境中最大化回报;RTO 则更新参数化动力学模型,以缩小两个环境动力学之间的差距。将两种算法整合后得到完整的 Relative Policy-Transition Optimization(RPTO,相对策略-转移优化)算法,其中策略同时与两个环境交互,使来自两个环境的数据收集、策略更新和转移更新在一个闭环中完成,从而形成一个有原则的策略迁移学习框架。我们通过变体动力学构造策略迁移问题,并在一组 MuJoCo 连续控制任务上证明了 RPTO 的有效性。

英文摘要:

We consider the problem of policy transfer between two Markov Decision Processes (MDPs). We introduce a lemma based on existing theoretical results in reinforcement learning to measure the relativity gap between two arbitrary MDPs, that is the difference between any two cumulative expected returns defined on different policies and environment dynamics. Based on this lemma, we propose two new algorithms referred to as Relative Policy Optimization (RPO) and Relative Transition Optimization (RTO), which offer fast policy transfer and dynamics modelling, respectively. RPO transfers the policy evaluated in one environment to maximize the return in another, while RTO updates the parameterized dynamics model to reduce the gap between the dynamics of the two environments. Integrating the two algorithms results in the complete Relative Policy-Transition Optimization (RPTO) algorithm, in which the policy interacts with the two environments simultaneously, such that data collections from two environments, policy and transition updates are completed in one closed loop to form a principled learning framework for policy transfer. We demonstrate the effectiveness of RPTO on a set of MuJoCo continuous control tasks by creating policy transfer problems via variant dynamics.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑