Learning to control from expert demonstrations

TL;DR

This paper presents a method to learn stabilizing controllers from expert demonstrations for feedback linearizable systems, optimizing demonstration selection and extending to systems embedded in chains of integrators, validated on a quadrotor.

Contribution

It introduces a novel approach to synthesize stabilizing controllers from finite expert demonstrations, including optimal selection of demonstrations and extension to complex systems.

Findings

Successfully stabilizes a CrazyFlie 2.0 quadrotor

Demonstrates effective selection of demonstrations for control

Extends methodology to systems embedded in chains of integrators

Abstract

In this paper, we revisit the problem of learning a stabilizing controller from a finite number of demonstrations by an expert. By first focusing on feedback linearizable systems, we show how to combine expert demonstrations into a stabilizing controller, provided that demonstrations are sufficiently long and there are at least $n+1$ of them, where $n$ is the number of states of the system being controlled. When we have more than $n+1$ demonstrations, we discuss how to optimally choose the best $n+1$ demonstrations to construct the stabilizing controller. We then extend these results to a class of systems that can be embedded into a higher-dimensional system containing a chain of integrators. The feasibility of the proposed algorithm is demonstrated by applying it on a CrazyFlie 2.0 quadrotor.