Interpretable machine learning approach for nonlinear control

Tools

Minisci, Edmondo and Avanzini, Giulio (2025) Interpretable machine learning approach for nonlinear control. In: 22nd Conference on Advances in Continuous Optimization, 2025-06-30 - 2025-07-02.

[thumbnail of Minisci-Avanzini-EUROPT-2025-Interpretable-machine-learning-approach]
Preview
 Text. Filename: Minisci-Avanzini-EUROPT-2025-Interpretable-machine-learning-approach.pdf
Accepted Author Manuscript
License: Creative Commons Attribution 4.0 logo
Download (999kB)| Preview

Abstract

The work is about an interpretable machine learning approach for control based on genetic programming with integrated search for continuous coefficients. The method can discover compact, human-readable control laws by combining symbolic expressions with embedded parameter tuning, thus bridging the gap between black-box learning and classical control. The main aim of the work is to demonstrate its potential on textbook cases, including standard systems (e.g., nonlinear oscillators, inverted pendulum) and provide initial results on aeronautical applications, such as stability augmentation of unstable, highly manoeuvrable aircraft. The obtained solutions maintain performance comparable to conventional controllers, while preserving transparency and ease of analysis, and the method enables the treatment of nonlinear systems without reliance on gain scheduling.

ORCID iDs

Minisci, Edmondo ORCID logoORCID: https://orcid.org/0000-0001-9951-8528 and Avanzini, Giulio;
CORE (COnnecting REpositories)