Algorithmic design and implementation considerations of deep MPC

2025-11-21

AI Summary

This work addresses the challenge of dynamically allocating control authority between neural network–based learning of model uncertainty and model predictive control (MPC) for constraint enforcement. We propose a dynamic co-design framework wherein a deep neural network learns unmodeled system dynamics online, while MPC ensures satisfaction of state and input constraints and guarantees closed-loop safety; their outputs are fused in real time via a learnable weighting mechanism that governs responsibility allocation. Our key contributions include: (i) a novel online weight adaptation strategy that jointly ensures Lyapunov-based stability and learning adaptability, and (ii) support for lightweight, data-driven neural network fine-tuning during operation. Numerical experiments demonstrate substantial improvements in trajectory tracking accuracy and closed-loop stability during learning transients, effectively mitigating performance degradation inherent in conventional serial or hard-switching architectures.

Technology Category

• Search and Optimization: Learning to Search
• Constraint Satisfaction and Optimization: Constraint Learning and Acquisition
• Machine Learning: Calibration & Uncertainty Quantification

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Economics, Online Markets and Human Computation: Incentives in network design for Web infrastructures and ecosystems
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization

Abstract

Deep Model Predictive Control (Deep MPC) is an evolving field that integrates model predictive control and deep learning. This manuscript is focused on a particular approach, which employs deep neural network in the loop with MPC. This class of approaches distributes control authority between a neural network and an MPC controller, in such a way that the neural network learns the model uncertainties while the MPC handles constraints. The approach is appealing because training data collected while the system is in operation can be used to fine-tune the neural network, and MPC prevents unsafe behavior during those learning transients. This manuscript explains implementation challenges of Deep MPC, algorithmic way to distribute control authority and argues that a poor choice in distributing control authority may lead to poor performance. A reason of poor performance is explained through a numerical experiment on a four-wheeled skid-steer dynamics.

Problem

• Integrating deep learning with model predictive control for system optimization
• Distributing control authority between neural networks and MPC controllers
• Addressing implementation challenges to prevent unsafe behavior during learning

Innovation

• Deep neural network integrated with MPC loop
• Neural network learns model uncertainties online
• MPC enforces constraints during learning transients