Safe and Efficient Model Predictive Control Using Neural Networks: An Interior Point Approach

TL;DR

This paper introduces a neural network-based approach for model predictive control that explicitly encodes constraints, enabling faster and more efficient solutions while ensuring constraint satisfaction, demonstrated on a robust MPC problem.

Contribution

The paper presents a novel neural network parameterization for MPC policies that explores the feasible set interior, improving speed and constraint handling over existing methods.

Findings

Faster policy computation compared to projection-based methods

Effective constraint satisfaction through interior point exploration

Significant computational gains demonstrated on test systems

Abstract

Model predictive control (MPC) provides a useful means for controlling systems with constraints, but suffers from the computational burden of repeatedly solving an optimization problem in real time. Offline (explicit) solutions for MPC attempt to alleviate real time computational challenges using either multiparametric programming or machine learning. The multiparametric approaches are typically applied to linear or quadratic MPC problems, while learning-based approaches can be more flexible and are less memory-intensive. Existing learning-based approaches offer significant speedups, but the challenge becomes ensuring constraint satisfaction while maintaining good performance. In this paper, we provide a neural network parameterization of MPC policies that explicitly encodes the constraints of the problem. By exploring the interior of the MPC feasible set in an unsupervised learning paradigm, the neural network finds better policies faster than projection-based methods and exhibits substantially shorter solve times. We use the proposed policy to solve a robust MPC problem, and demonstrate the performance and computational gains on a standard test system.