Towards Safe Learning-Based Non-Linear Model Predictive Control through Recurrent Neural Network Modeling

TL;DR

This paper introduces Sequential-AMPC, a neural policy for nonlinear model predictive control that reduces online computation, improves safety, and enhances learning efficiency on complex systems.

Contribution

It proposes Sequential-AMPC, a sequential neural policy sharing parameters across the horizon, with safety mechanisms, outperforming naive policies in feasibility, safety, and learning speed.

Findings

Requires fewer expert MPC rollouts.

Higher feasibility rates and safety in control sequences.

Better learning dynamics on high-dimensional systems.

Abstract

The practical deployment of nonlinear model predictive control (NMPC) is often limited by online computation: solving a nonlinear program at high control rates can be expensive on embedded hardware, especially when models are complex or horizons are long. Learning-based NMPC approximations shift this computation offline but typically demand large expert datasets and costly training. We propose Sequential-AMPC, a sequential neural policy that generates MPC candidate control sequences by sharing parameters across the prediction horizon. For deployment, we wrap the policy in a safety-augmented online evaluation and fallback mechanism, yielding Safe Sequential-AMPC. Compared to a naive feedforward policy baseline across several benchmarks, Sequential-AMPC requires substantially fewer expert MPC rollouts and yields candidate sequences with higher feasibility rates and improved closed-loop safety. On high-dimensional systems, it also exhibits better learning dynamics and performance in fewer epochs while maintaining stable validation improvement where the feedforward baseline can stagnate.