Electromagnetic Environment Classification for Robust UAV and Robotic Operation Using Ensemble and Deep Learning Models

A Machine Learning-Powered Solution for Safe Autonomous Robotic Ground Navigation in Cyber-Contested Environments

In this article, machine learning (ML) is proposed as a solution to detect and classify false message injection attacks in autonomous ground navigation. First, multiple trajectories are designed and simulated to collect authentic feature samples offered by the odometry and inertial measurement unit (IMU) of an autonomous ground vehicle (UGV). Then, a dataset comprising these samples and other injected samples that simulate two cyberattacks, namely path modification (PM) and velocity drift (VD), is created to train, validate, and benchmark various ML classification models. These include decision tree (DT), k-nearest neighbors (KNN), multi-layer perceptron (MLP), random forest (RF), and support vector machine (SVM). The optimum classification model is experimentally evaluated using a UGV platform, and results suggest that the proposed solution allows the detection of authentic and attacked messages with more than 98% average accuracy and sub-millisecond prediction time. Thus, this solution is ideal for real-time classification, especially in fixed-route applications, e.g., public transportation.

Applied machine learning for power grid security UAV and bird classification based on 5G-A trajectories

The increasing prevalence of low-altitude aerial activities presents significant security challenges to power grid infrastructure, as distinguishing non-cooperative unmanned aerial vehicles from natural birds remains a persistent technical hurdle. Traditional sensing modalities—such as electro-optical/infrared imaging and radar—often falter under complex environmental conditions or suffer from prohibitive deployment costs. Addressing these constraints, this study introduces an applied machine learning framework that classifies these targets by mining kinematic features from 5 G-Advanced (5 G-A) integrated sensing and communication trajectory data. Our approach builds upon a high-fidelity real-world dataset of 5856 trajectories, utilizing a multidimensional feature system—highlighted by a novel clustered heading standard deviation—to capture the fundamental kinematic divergence between electromechanical control and biological flight. Extensive evaluations across lightweight machine learning models reveal that the random forest classifier achieves optimal discriminative performance, yielding an accuracy of 91.68% and an area under the curve of 0.993, significantly outperforming both support vector machine and k-nearest neighbor counterparts. These findings validate 5 G-A trajectory sensing as a robust, low-cost technical paradigm for wide-area low-slow-small target monitoring, establishing a firm foundation for future low-altitude power grid security.

Evidential Deep Learning for Multi-Modal Anti-UAV Detection

Anti-UAV systems increasingly fuse multiple sensors, yet their detection heads provide no per-modality reliability signal. This study evaluates whether evidential deep learning (EDL) heads, Dempster-Shafer (DS) evidence fusion, and uncertainty-driven temporal sensor gating improve anti-UAV detection through a controlled ablation on three benchmarks: thermal tracking (AntiUAV600), RGB-audio-RF classification (TRIDENT), and RGB-IR tracking (MM-UAV). The EDL training objective improves accuracy over retrained sigmoid baselines (+5.9 percentage points in accuracy and a tripled tracker-on-absent rate in E1; +4.8 percentage points in classification accuracy in E2, surviving a clip-clustered bootstrap, p = 0.011) and ranks classification errors substantially better (entropy UAUC approximately 0.94 vs. 0.51). The remaining components do not support their respective hypotheses. DS fusion does not outperform simple probability averaging. Dirichlet vacuity adds no ranking power beyond predictive entropy and inverts at the detection level, where extreme background imbalance causes it to encode class membership rather than error likelihood, a failure also observed for entropy and sigmoid confidence. Temporal gating preserves accuracy only when nearly inactive and yields no realised latency saving on shared-backbone hardware. The benefit of evidential learning therefore arises primarily from its training objective rather than its uncertainty estimate; a crop-level control further localises the detection-level breakdown to anchor-level evaluation rather than the learned representation.

ENHANCING UAV CLASSIFICATION WITH RADIO FREQUENCY SIGNALS USING A HYBRID CONVOLUTIONAL-LSTM ARCHITECTURE

Fault diagnosis method for quadrotor UAV based on ResCNN-LSTM-attention

To address the challenges of difficult fault feature extraction and low diagnostic accuracy caused by noise interference in quadrotor unmanned aerial vehicles (UAVs) under complex flight environments, this paper proposes a UAV fault diagnosis model (ResCNN-LSTM-ATT) that integrates residual convolutional neural network (ResCNN), bidirectional long short-term memory (BiLSTM), and an attention mechanism. The proposed model seeks to concurrently acquire deep spatial features and long-term temporal dependencies from noise-corrupted airborne multi-source sensor time-series data. Specifically, ResCNN is first employed to extract local spatial features from high-dimensional observational data and suppress noise via residual connections and one-dimensional convolutional neural networks. Secondly, a BiLSTM structure is constructed to capture the bidirectional temporal dependencies of the flight data. Subsequently, an attention mechanism is introduced to weight the outputs of the BiLSTM, focusing on key fault time steps. Finally, the fused spatiotemporal features are fed into the classifier. Experimental results on typical UAV fault diagnosis tasks, including motor, gyroscope, and magnetometer faults, demonstrate that the proposed method achieves superior diagnostic accuracy over existing state-of-the-art approaches while maintaining a compact architecture suitable for real-time onboard deployment.

A Comparative Study of Machine Learning and Deep Learning Models for State of Charge and Remaining Useful Life Estimation on a Rotary-Wing UAV Battery

Battery state estimation is the main safety constraint for electric rotary-wing unmanned aerial vehicles (UAVs): mission decisions depend on both the instantaneous State of Charge (SOC) and the Remaining Useful Life (RUL). The present study compares seven machine learning and deep learning models (LR, SVM, k-NN, GBT, EL, LSTM, and a simplified RWKV) on real flight data from a rotary-wing helicopter testbed with a Pixhawk autopilot and an NVIDIA Jetson Nano mission computer. The dataset has 1310 samples (∼262 s) of nine on-board sensor signals. Mission-based RUL is defined as the projected time until SOC reaches a 20% safe-landing threshold. All models use an 80/20 random split, five regression metrics (RMSE, MAE, R2, MSE, PRMSE), and five random seeds. GBT wins on SOC with R2=0.9943±0.0014, MAE =0.25%, and 3.8μs per-sample inference on a workstation CPU; this latency leaves headroom for on-board mission planning. Battery temperature and voltage together carry over 90% of the predictive signal. GBT wins again on RUL (R2=0.596±0.042, MAE =583 s). The same ordering (tree ensemble ≻ recurrent ≻ linear) holds for both tasks; the remaining RUL gap reflects the single-flight dataset. The SOC labels originate from the on-board autopilot’s Coulomb-counting-based fuel-gauge estimator, so the SOC numbers should be read as a reproduction of that on-board trace at sub-microsecond inference latency rather than as independent accuracy; the calibration-free Coulomb-counting baseline reaches a marginally higher R2 (0.9950, MAE =0.35%) on the same task.