Network Resource Optimization for ML-Based UAV Condition Monitoring with Vibration Analysis

TL;DR

The study tackles real-time UAV condition monitoring under edge-network constraints by proposing a resource-aware framework that optimizes data aggregation and applies PCA-based dimensionality reduction to minimize feature throughput while preserving anomaly-detection performance. It systematically evaluates multiple ML models (SVC, KNN, DT, RF) across aggregation intervals and feature-subset configurations, highlighting significant gains when applying PCA to STFT features combined with a Random Forest classifier. The key finding is a dramatic reduction in network traffic (up to 99.92%) with near-perfect ML metrics, particularly at 4000-sample aggregation using STFT PCA, demonstrating a practical route for deploying ML-based UAV CM in resource-limited edge environments. These results offer actionable guidance for designing scalable, real-time UAV health analytics pipelines in smart-city contexts.

Abstract

As smart cities begin to materialize, the role of Unmanned Aerial Vehicles (UAVs) and their reliability becomes increasingly important. One aspect of reliability relates to Condition Monitoring (CM), where Machine Learning (ML) models are leveraged to identify abnormal and adverse conditions. Given the resource-constrained nature of next-generation edge networks, the utilization of precious network resources must be minimized. This work explores the optimization of network resources for ML-based UAV CM frameworks. The developed framework uses experimental data and varies the feature extraction aggregation interval to optimize ML model selection. Additionally, by leveraging dimensionality reduction techniques, there is a 99.9% reduction in network resource consumption.

Figures (5)

• Figure 1: System Model
• Figure 2: Experiment Overview
• Figure 3: No PCA applied
• Figure 4: PCA applied to STFT features
• Figure 5: PCA applied to all features