arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1709.06166cs.AIcs.RO
收藏

DropoutDAgger:一种安全模仿学习的贝叶斯方法

DropoutDAgger: A Bayesian Approach to Safe Imitation Learning

Stanford University(斯坦福大学)

Kunal Menda, Katherine Driggs-Campbell, Mykel J. Kochenderfer

摘要:

虽然模仿学习正在成为机器人领域的常见做法,但这种方法通常会出现数据不匹配和复合错误。Dagger是一种迭代算法,它通过不断聚合来自专家和新手策略的训练数据来解决这些问题,但不考虑安全性的影响。我们提出了一个概率扩展Dagger,它使用的分布在行动提供的新手政策,对于一个给定的观察。我们的方法,我们称之为DropoutDagger,使用dropout来训练新手作为贝叶斯神经网络,提供对其信心的洞察。使用分布在新手的行动,我们估计一个概率的安全措施方面的专家行动,调整到平衡的探索和利用。在MuJoCo HalfCheetah和一个简单的驾驶实验中评估了这种方法的实用性,与其他Dagger变体和经典的模仿学习相比,证明了性能和安全性的提高。

英文摘要:

While imitation learning is becoming common practice in robotics, this approach often suffers from data mismatch and compounding errors. DAgger is an iterative algorithm that addresses these issues by continually aggregating training data from both the expert and novice policies, but does not consider the impact of safety. We present a probabilistic extension to DAgger, which uses the distribution over actions provided by the novice policy, for a given observation. Our method, which we call DropoutDAgger, uses dropout to train the novice as a Bayesian neural network that provides insight to its confidence. Using the distribution over the novice's actions, we estimate a probabilistic measure of safety with respect to the expert action, tuned to balance exploration and exploitation. The utility of this approach is evaluated on the MuJoCo HalfCheetah and in a simple driving experiment, demonstrating improved performance and safety compared to other DAgger variants and classic imitation learning.

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。

↑