Output Feedback Tube MPC-Guided Data Augmentation for Robust, Efficient Sensorimotor Policy Learning

TL;DR

This paper introduces a novel data augmentation method combining imitation learning with a robust tube MPC controller, enabling efficient and robust sensorimotor policy learning from minimal demonstrations, specifically demonstrated on aerial robot trajectory tracking.

Contribution

It presents a new approach that integrates RTMPC-based data augmentation with imitation learning to significantly reduce demonstration requirements and enhance robustness.

Findings

Learned a robust visuomotor policy from a single demonstration.

Reduced demonstration collection and computation time by two orders of magnitude.

Enhanced robustness to sensing and process uncertainties.

Abstract

Imitation learning (IL) can generate computationally efficient sensorimotor policies from demonstrations provided by computationally expensive model-based sensing and control algorithms. However, commonly employed IL methods are often data-inefficient, requiring the collection of a large number of demonstrations and producing policies with limited robustness to uncertainties. In this work, we combine IL with an output feedback robust tube model predictive controller (RTMPC) to co-generate demonstrations and a data augmentation strategy to efficiently learn neural network-based sensorimotor policies. Thanks to the augmented data, we reduce the computation time and the number of demonstrations needed by IL, while providing robustness to sensing and process uncertainty. We tailor our approach to the task of learning a trajectory tracking visuomotor policy for an aerial robot, leveraging a 3D mesh of the environment as part of the data augmentation process. We numerically demonstrate that our method can learn a robust visuomotor policy from a single demonstration--a two-orders of magnitude improvement in demonstration efficiency compared to existing IL methods.