An Active Learning Framework for Efficient Robust Policy Search
This work addresses the challenge of learning robust policies for robotics and control applications, offering an incremental improvement in sample efficiency over existing methods.
The paper tackles the problem of robust policy search for transferring policies from simulation to real-world environments by proposing an active learning framework, EffAcTS, which uses linear bandits to selectively choose model parameters, resulting in improved sample efficiency and performance on continuous control tasks.
Robust Policy Search is the problem of learning policies that do not degrade in performance when subject to unseen environment model parameters. It is particularly relevant for transferring policies learned in a simulation environment to the real world. Several existing approaches involve sampling large batches of trajectories which reflect the differences in various possible environments, and then selecting some subset of these to learn robust policies, such as the ones that result in the worst performance. We propose an active learning based framework, EffAcTS, to selectively choose model parameters for this purpose so as to collect only as much data as necessary to select such a subset. We apply this framework using Linear Bandits, and experimentally validate the gains in sample efficiency and the performance of our approach on standard continuous control tasks. We also present a Multi-Task Learning perspective to the problem of Robust Policy Search, and draw connections from our proposed framework to existing work on Multi-Task Learning.