Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Enhancing Field-Oriented Control of Electric Drives with Tiny Neural Network Optimized for Micro-controllers

Martin Joel Mouk Elele, Danilo Pau, Shixin Zhuang, Tullio Facchinetti

TL;DR

This work tackles the challenge of enhancing Field-Oriented Control for PMSMs on resource-constrained micro-controllers by integrating a tiny neural network, TinyFC, into the speed control loop. Through a two-branch, 1,400-parameter architecture optimized via Bayesian hyperparameter search, PCA-based pruning, and 8-bit quantization, the authors demonstrate significant reductions in overshoot and improved speed tracking compared to PI-based FOC in simulated test cases. Key contributions include a high-fidelity Simulink-based testbed, a dataset-ground-truth framework using adjusted PI predictions, and a practical deployment evaluation on an STM32G4 MCU showing favorable memory and timing characteristics. The findings indicate that TinyNN augmentation can enhance real-time motor control in edge devices, while also highlighting the need for physics-informed loss criteria and further work to replace PI with a fully data-driven control approach for PMSMs.

Abstract

The deployment of neural networks on resource-constrained micro-controllers has gained momentum, driving many advancements in Tiny Neural Networks. This paper introduces a tiny feed-forward neural network, TinyFC, integrated into the Field-Oriented Control (FOC) of Permanent Magnet Synchronous Motors (PMSMs). Proportional-Integral (PI) controllers are widely used in FOC for their simplicity, although their limitations in handling nonlinear dynamics hinder precision. To address this issue, a lightweight 1,400 parameters TinyFC was devised to enhance the FOC performance while fitting into the computational and memory constraints of a micro-controller. Advanced optimization techniques, including pruning, hyperparameter tuning, and quantization to 8-bit integers, were applied to reduce the model's footprint while preserving the network effectiveness. Simulation results show the proposed approach significantly reduced overshoot by up to 87.5%, with the pruned model achieving complete overshoot elimination, highlighting the potential of tiny neural networks in real-time motor control applications.

Enhancing Field-Oriented Control of Electric Drives with Tiny Neural Network Optimized for Micro-controllers

TL;DR

This work tackles the challenge of enhancing Field-Oriented Control for PMSMs on resource-constrained micro-controllers by integrating a tiny neural network, TinyFC, into the speed control loop. Through a two-branch, 1,400-parameter architecture optimized via Bayesian hyperparameter search, PCA-based pruning, and 8-bit quantization, the authors demonstrate significant reductions in overshoot and improved speed tracking compared to PI-based FOC in simulated test cases. Key contributions include a high-fidelity Simulink-based testbed, a dataset-ground-truth framework using adjusted PI predictions, and a practical deployment evaluation on an STM32G4 MCU showing favorable memory and timing characteristics. The findings indicate that TinyNN augmentation can enhance real-time motor control in edge devices, while also highlighting the need for physics-informed loss criteria and further work to replace PI with a fully data-driven control approach for PMSMs.

Abstract

The deployment of neural networks on resource-constrained micro-controllers has gained momentum, driving many advancements in Tiny Neural Networks. This paper introduces a tiny feed-forward neural network, TinyFC, integrated into the Field-Oriented Control (FOC) of Permanent Magnet Synchronous Motors (PMSMs). Proportional-Integral (PI) controllers are widely used in FOC for their simplicity, although their limitations in handling nonlinear dynamics hinder precision. To address this issue, a lightweight 1,400 parameters TinyFC was devised to enhance the FOC performance while fitting into the computational and memory constraints of a micro-controller. Advanced optimization techniques, including pruning, hyperparameter tuning, and quantization to 8-bit integers, were applied to reduce the model's footprint while preserving the network effectiveness. Simulation results show the proposed approach significantly reduced overshoot by up to 87.5%, with the pruned model achieving complete overshoot elimination, highlighting the potential of tiny neural networks in real-time motor control applications.

Paper Structure

This paper contains 16 sections, 5 equations, 7 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Preliminaries and Related Work
  3. System Design
  4. Setup and hardware
  5. Test cases
  6. Training of the Neural Network
  7. Neural network design
  8. Hyperparameters optimization
  9. Pruning
  10. Quantization
  11. Experimental Results
  12. Accuracy evaluation
  13. Control loop evaluation
  14. Deployability on the MCU
  15. Discussion
  16. ...and 1 more sections

Figures (7)

  • Figure 1: Workflow diagram to deploy NN-augmented FOC
  • Figure 2: System model of input (in green), control (in blue) and motor/inverter (in gray) implemented in Simulink
  • Figure 3: Measured speed collected from PI-based FOC
  • Figure 4: Quadrature current from PI-controller simulation adjusted using speed observations threshold
  • Figure 5: Quadrature current adjusted based on capping and rectification of measured speed
  • ...and 2 more figures