Efficiently Learning Robust, Adaptive Controllers from Robust Tube MPC

Abstract

The deployment of agile autonomous systems in challenging, unstructured environments requires adaptation capabilities and robustness to uncertainties. Existing robust and adaptive controllers, such as those based on model predictive control (MPC), can achieve impressive performance at the cost of heavy online onboard computations. Strategies that efficiently learn robust and onboard-deployable policies from MPC have emerged, but they still lack fundamental adaptation capabilities. In this work, we extend an existing efficient Imitation Learning (IL) algorithm for robust policy learning from MPC with the ability to learn policies that adapt to challenging model/environment uncertainties. The key idea of our approach consists of modifying the IL procedure by conditioning the policy on a learned lower-dimensional model/environment representation that can be efficiently estimated online. We tailor our approach to learning an adaptive position and attitude control policy to track trajectories under challenging disturbances on a multirotor. Our evaluation shows that a high-quality adaptive policy can be obtained in about 1.3 hours of combined demonstration and training time. We empirically demonstrate rapid adaptation to in- and out-of-training-distribution uncertainties, achieving a 6.1 cm average position error under wind disturbances that correspond to 50% of the weight of the robot, and that are 36% larger than the maximum wind seen during training. Additionally, we verify the performance of our controller during real-world deployment in multiple trajectories, demonstrating adaptation to turbulent winds of up to 5.2 m/s and slung loads of up to 40% of the robot’s mass, and reducing the average position error on each trajectory to under 15 cm, a 70% improvement compared to a non-adaptive baseline.