Neural Network Verification for Gliding Drone Control: A Case Study

2025-05-01

AI Summary

This work addresses the formal robustness verification of a neural-network-based trajectory tracking controller for centimeter-scale biomimetic gliding micro-air vehicles—inspired by *Alsomitra macrocarpa* seeds—to fill a critical gap in verification methodologies for micro-air vehicles operating under passive wind transport. We propose a novel robust training framework tailored for regression-type neural networks, construct the first verified neural network (VNN) benchmark case for gliding micro-air vehicles, and integrate vehicle dynamics simulation (Vehicle), reachability analysis (CORA), and high-fidelity aerodynamic modeling. Experimental results demonstrate that our training method substantially improves both tracking accuracy and robustness of the closed-loop system. Moreover, the study exposes systematic limitations of existing VNN verification tools when applied to highly nonlinear, physics-driven dynamical systems. Collectively, these contributions establish a scalable, physics-informed methodology for engineering-grade safety verification of autonomous micro-air vehicles.

Technology Category

• Intelligent Robots: Behavior Learning & Control
• Computer Vision: Motion & Tracking
• Machine Learning: Adversarial Learning & Robustness

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

As machine learning is increasingly deployed in autonomous systems, verification of neural network controllers is becoming an active research domain. Existing tools and annual verification competitions suggest that soon this technology will become effective for real-world applications. Our application comes from the emerging field of microflyers that are passively transported by the wind, which may have various uses in weather or pollution monitoring. Specifically, we investigate centimetre-scale bio-inspired gliding drones that resemble Alsomitra macrocarpa diaspores. In this paper, we propose a new case study on verifying Alsomitra-inspired drones with neural network controllers, with the aim of adhering closely to a target trajectory. We show that our system differs substantially from existing VNN and ARCH competition benchmarks, and show that a combination of tools holds promise for verifying such systems in the future, if certain shortcomings can be overcome. We propose a novel method for robust training of regression networks, and investigate formalisations of this case study in Vehicle and CORA. Our verification results suggest that the investigated training methods do improve performance and robustness of neural network controllers in this application, but are limited in scope and usefulness. This is due to systematic limitations of both Vehicle and CORA, and the complexity of our system reducing the scale of reachability, which we investigate in detail. If these limitations can be overcome, it will enable engineers to develop safe and robust technologies that improve people's lives and reduce our impact on the environment.

Problem

• Verifying neural network controllers for gliding drones
• Ensuring adherence to target trajectory in bio-inspired drones
• Overcoming limitations in robust training and verification tools

Innovation

• Robust training for regression networks
• Verification with Vehicle and CORA tools
• Bio-inspired gliding drone neural control