Embedded Model Predictive Control of a Ball-in-Airstream System

Zenodo (CERN European Organization for Nuclear Research)

Abstract

Abstract This dataset and source code provide the implementation of an embedded Linear Model Predictive Control (LMPC) system designed for the position control of a sphere within a ball in airstream system. The project was developed at the Embedded Systems Laboratory (LSE) at the Federal University of Santa Catarina (UFSC). Technical Specifications The control system utilizes the TinyMPC solver, optimized for real-time execution on resource-constrained hardware. The firmware is built upon the Zephyr Real-Time Operating System (RTOS) and targets the STM32G474RE microcontroller. Key technical features include: Controller: Linearized MPC with offline pre-computed Riccati recursion to ensure deterministic timing. Operating Point: System linearized at $h = 0.6$ m for stable operation within the 0.2 m to 0.8 m range. State Estimation: Implementation of an Extended Kalman Filter (EKF) for noise reduction from Time-of-Flight (ToF) sensors and velocity reconstruction. Hardware: Custom-designed PCB (Ball-in-Airstream Control Circuit) featuring industrial safety layers, including crowbar overvoltage protection. Content Description This record contains: Firmware: C++ source code developed for the Zephyr RTOS environment. Hardware Design: KiCad project files for the custom control board. Documentation: System diagrams, mathematical models for the ball in airstream dynamics, and technical schematics. Experimental Data: Results of reference tracking and step response tests demonstrating the stability and performance of the predictive controller. This work serves as a reference for real-time predictive control applications in embedded systems and academic research in automation.

Authors 3

1. Matheus Franco

   Universidade Federal de Santa Catarina

   Affiliation as printed

   Universidade Federal de Santa Catarina

2. Gian Ricardo Berkenbrock

   Universidade Federal de Santa Catarina

   Affiliation as printed

   Universidade Federal de Santa Catarina

3. Robert Göllinger Aachen

   RWTH Aachen University

   Affiliation as printed

   RWTH Aachen University