Uncertainty Aware Learning from Demonstrations in Multiple Contexts using Bayesian Neural Networks
This work addresses the challenge of robust robotic control in diverse environments, though it is incremental as it applies an existing uncertainty method to a known problem.
The paper tackles the problem of robotic controllers failing due to environmental discrepancies by using Bayesian Neural Networks to detect uncertain situations where performance would be sub-par, showing that this approach enables informed fallback decisions and increases data-efficiency.
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.