Bayesian Optimization-Based Tunable Explicit MPC on a Pocket-Sized Embedded Platform

The paper presents a compact embedded platform developed for validating advanced control methods. It is built on an ESP32-S3 microcontroller and includes a miniaturized heat exchange device, making it suitable for temperature control experiments.

The platform supports the implementation of a real-time tunable explicit Model Predictive Control (MPC) algorithm, allowing online adjustment of the weighting parameter in the MPC cost function. In addition, the paper introduces a novel application of Bayesian optimization to efficiently explore the parameter space and identify optimal performance settings.

The performance metric combines multiple aspects of controller behavior, specifically actuator power consumption and control signal variability. Experimental results confirm the effectiveness of the approach, with both exploration and exploitation strategies evaluated.

Both the explicit MPC controller and the Bayesian optimization algorithm are implemented directly on the embedded platform, demonstrating its capability for real-time control. The results suggest strong potential for future research and development in advanced control strategies.