Edge AI-Enabled Real-Time Gesture Recognition for UAV Control

View at publisher Summarise with VIRAT
Authors: Weihang You, Junchen Wu
Cited by: 0
References: 16

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Gesture-based UAV control provides an intuitive alternative to conventional remote controllers, yet existing approaches often fail to satisfy real-time, robustness, and edge deployment constraints. This paper presents a lightweight edge AI framework for real-time UAV gesture control using a MobileNetV3-based convolutional neural network optimized for embedded platforms. The system supports ten control gestures, achieves sub- 20 ms end-to-end latency, and integrates directly with MAVLink-based flight controllers. Experimental results on multiple edge devices demonstrate favorable accuracy-efficiency trade-offs compared with state-of-the-art baselines. In addition, the proposed framework maintains stable performance across varying lighting conditions and backgrounds, highlighting its robustness for outdoor and indoor applications. A comprehensive ablation study further analyzes the impact of model depth, input resolution, and quantization on both accuracy and latency. The framework is evaluated in closed-loop flight tests, confirming that the recognized gestures translate into safe, smooth, and responsive UAV maneuvers.

Fields of Study & Subjects

Hand Gesture Recognition Systems Advanced Neural Network Applications Aerospace Engineering and Control Systems

Publication details

Journal / source
International Conference on Electrical Automation and Artificial Intelligence (ICEAAI)
Pages
600–604
Publication date
16 January 2026
Language
EN