Performance guaranteed MPC Policy Approximation via Cost Guided Learning

TL;DR

This paper introduces a cost-guided learning method for approximating MPC policies with neural networks, directly optimizing for operational cost rather than fitting error, leading to better closed-loop performance.

Contribution

It presents a novel cost-guided learning approach that leverages cost sensitivity to improve MPC policy approximation, with theoretical guarantees and practical validation.

Findings

Cost-guided learning yields tighter optimality guarantees than error-guided methods.

Experiments on CSTR demonstrate improved closed-loop performance.

The approach effectively connects fitting errors with operational objectives.

Abstract

Model predictive control (MPC) is widely used in industries but implementing it poses challenges due to hardware or time constraints. A promising solution is to approximate the MPC policy using function approximators like neural networks. Existing methods focus on minimizing the error between the approximators outputs and the MPC optimal control actions on training data, which is called error guided learning approach in this paper. However, the goals of control law design is not to minimize the fitting error but to minimize the operation cost. This paper proposes a novel cost-guided learning approach that utilizes the cost sensitivity information from the MPC problem to directly minimize the loss in closed-loop performance. A theoretical analysis shows cost-guided learning provides tighter guarantees on optimality loss compared to traditional error-guided learning. Experiments on a continuous stirred tank reactor (CSTR) benchmark demonstrate that the proposed technique results in approximate MPC policies that achieve substantially better closed-loop performance. This work makes an important contribution by connecting the fitting errors with operational objectives, overcoming key limitations of existing approximation methods. The core idea could be applied more broadly for data-driven control.