Skip to yearly menu bar Skip to main content


Oral
in
Workshop: Safe and Robust Control of Uncertain Systems

Reinforcement Learning with Feedback from Multiple Humans with Diverse Skills

Taku Yamagata · Ryan McConville · Raul Santos-Rodriguez


Abstract:

One promising approach to improve the robustness and exploration in Reinforcement Learning is by taking human feedback and in that way incorporating prior knowledge of the target environment. It is, however, often too expensive to obtain enough feedback of good quality. To mitigate the issue, we aim to rely on a group of multiple experts (and non-experts) with different skill levels to generate enough feedback. Such feedback can therefore be inconsistent and infrequent. In this work, we build upon prior work -- Advise, a Bayesian approach attempting to maximise the information gained from human feedback -- extending the algorithm to accept feedback from this larger group of humans, the trainers, while also estimating each trainer's reliability. We show how aggregating feedback from multiple trainers improves the total feedback's accuracy and make the collection process easier in two ways. Firstly, this approach addresses the case of some of the trainers being adversarial. Secondly, having access to the information about each trainer reliability provides a second layer of robustness and offers valuable information for people managing the whole system to improve the overall trust in the system. It offers tools for improving the feedback collection process or modifying the reward function design if needed. We empirically show that our approach can accurately learn the reliability of each trainer correctly and use it to maximise the information gained from the multiple trainers' feedback, even if some of the sources are adversarial.