基于贝叶斯神经网络的不确定性示教学习
Uncertainty Aware Learning from Demonstrations in Multiple Contexts using Bayesian Neural Networks
摘要:
环境的多样性是一个关键的挑战,导致学习机器人控制器失败,由于训练和评估条件之间的差异。通过各种条件下的演示进行培训可以减轻-但不能完全防止-此类故障。诸如神经网络的学习控制器通常不具有允许诊断训练和测试条件之间的偏移的不确定性的概念,并且可能进行干预。在这项工作中,我们建议使用贝叶斯神经网络,它有这样一个概念的不确定性。我们表明,可以利用不确定性来一致地检测高维模拟和真实机器人领域中的情况,其中学习控制器的性能将低于标准杆。此外,我们表明,这样一个基于不确定性的解决方案,允许作出明智的决定,何时调用回退策略。一种回退策略是请求更多数据。我们的经验表明,只有在请求时才提供数据,从而提高数据效率。
英文摘要:
Diversity of environments is a key challenge that causes learned robotic controllers to fail due to the discrepancies between the training and evaluation conditions. Training from demonstrations in various conditions can mitigate---but not completely prevent---such failures. Learned controllers such as neural networks typically do not have a notion of uncertainty that allows to diagnose an offset between training and testing conditions, and potentially intervene. In this work, we propose to use Bayesian Neural Networks, which have such a notion of uncertainty. We show that uncertainty can be leveraged to consistently detect situations in high-dimensional simulated and real robotic domains in which the performance of the learned controller would be sub-par. Also, we show that such an uncertainty based solution allows making an informed decision about when to invoke a fallback strategy. One fallback strategy is to request more data. We empirically show that providing data only when requested results in increased data-efficiency.
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。