Fine-Tuning of Neural Network Approximate MPC without Retraining via Bayesian Optimization

2025-12-16

AI Summary

Adaptive Model Predictive Control (AMPC) deployment suffers from poor practicality due to manual parameter tuning and repeated neural network retraining. Method: This paper proposes a data-driven, Bayesian optimization–based online adaptive method for AMPC parameter refinement—marking the first application of Bayesian optimization to AMPC parameter tuning. Crucially, it enables real-time hardware-in-the-loop adaptation without retraining the underlying neural network model. The approach eliminates subjectivity and counterintuitive behavior inherent in manual tuning, especially in high-dimensional systems or when the cost function is difficult to model analytically. Results: Hardware experiments on swing-up control of an inverted pendulum cart and yaw control of an underactuated unicycle demonstrate that the proposed method achieves significantly superior performance over baseline AMPC using only minimal experimental data, validating its efficiency and cross-task generalizability.

Technology Category

• Search and Optimization: Sampling/Simulation-based Search
• Reasoning under Uncertainty: Stochastic Optimization
• Machine Learning: Optimization

Application Category

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

Abstract

Approximate model-predictive control (AMPC) aims to imitate an MPC's behavior with a neural network, removing the need to solve an expensive optimization problem at runtime. However, during deployment, the parameters of the underlying MPC must usually be fine-tuned. This often renders AMPC impractical as it requires repeatedly generating a new dataset and retraining the neural network. Recent work addresses this problem by adapting AMPC without retraining using approximated sensitivities of the MPC's optimization problem. Currently, this adaption must be done by hand, which is labor-intensive and can be unintuitive for high-dimensional systems. To solve this issue, we propose using Bayesian optimization to tune the parameters of AMPC policies based on experimental data. By combining model-based control with direct and local learning, our approach achieves superior performance to nominal AMPC on hardware, with minimal experimentation. This allows automatic and data-efficient adaptation of AMPC to new system instances and fine-tuning to cost functions that are difficult to directly implement in MPC. We demonstrate the proposed method in hardware experiments for the swing-up maneuver on an inverted cartpole and yaw control of an under-actuated balancing unicycle robot, a challenging control problem.

Problem

• Fine-tuning AMPC parameters without retraining neural networks
• Automating labor-intensive manual adaptation for high-dimensional systems
• Enabling data-efficient adaptation to new system instances and cost functions

Innovation

• Bayesian optimization for AMPC parameter tuning
• Model-based control combined with direct local learning
• Automatic adaptation without neural network retraining