Data-Driven Optimal Control of Bilinear Systems

TL;DR

This paper introduces a data-driven method for optimal control of bilinear systems that learns control strategies directly from data without needing prior system models, using an iterative convexification approach.

Contribution

It proposes a novel data-based framework that transforms the optimal control problem into a convexified optimization, enabling control synthesis without explicit system knowledge.

Findings

Performance comparable to model-based methods

Effective online experiment design guarantees data sufficiency

Iterative convex-concave procedure finds locally optimal controls

Abstract

This paper develops a method to learn optimal controls from data for bilinear systems without a priori knowledge of the system dynamics. Given an unknown bilinear system, we first characterize when the available data is suitable to solve the optimal control problem. This characterization leads us to propose an online control experiment design procedure that guarantees that any input/state trajectory can be represented as a linear combination of collected input/state data matrices. Leveraging this data-based representation, we transform the original optimal control problem into an equivalent data-based optimization problem with bilinear constraints. We solve the latter by iteratively employing a convex-concave procedure to convexify it and find a locally optimal control sequence. Simulations show that the performance of the proposed data-based approach is comparable with model-based methods.