BriefGPT.xyz
May, 2020
机器人强化学习的平滑探索
Generalized State-Dependent Exploration for Deep Reinforcement Learning in Robotics
HTML
PDF
Antonin Raffin, Freek Stulp
TL;DR
采用状态相关的探索方法（SDE）来代替当前深度强化学习算法中常用的无结构步骤探索，提出了一种新的通用状态相关探索方法 (gSDE)，通过定期重新采样噪音来解决真实机器人上运动抖动的问题, 在仿真环境和三个不同的真实机器人上进行了评估并提高了表现。
Abstract
reinforcement learning
(RL) enables robots to learn skills from interactions with the real world. In practice, the unstructured step-based exploration used in
deep rl
-- often very successful in simulation -- lea
→