BriefGPT.xyz
Jun, 2012
具有局部最优示例的连续逆优化控制
Continuous Inverse Optimal Control with Locally Optimal Examples
HTML
PDF
Sergey Levine, Vladlen Koltun
TL;DR
本文介绍了一种适用于大规模连续任务的概率反向最优控制算法，通过使用奖励函数的局部估计值，该方法可以学习来自非全局最优演示的例子，并消除全局最优的假设。
Abstract
inverse optimal control
, also known as
inverse reinforcement learning
, is the problem of recovering an unknown reward function in a
markov decisi
→