analiza aiqbench
This paper combines a temporal convolutional network, a graph attention network, a Transformer-VAE anomaly detector, and Shapley-based attribution to detect and explain faults in fixed-wing drones, then feeds the evidence to an LLM for diagnosis. The reported results (window-level PR-AUC of 0.976, flight-level F1 of 0.944) come from cross-flight holdout tests on a single public dataset, ALFA, so generalization to other platforms and real-world fault distributions remains untested. The LLM component is constrained by structured evidence, which is a useful pattern for interpretable diagnostics, but the paper's own claims are limited to post-flight review.
TGTV: An Interpretable Framework for UAV Fault Detection and Traceability via Spatiotemporal Graph Learning and Large Language Models
arXiv:2610.38106v1 Announce Type: new Abstract: Fixed-wing unmanned aerial vehicles experience operating-condition transitions, asynchronous sensor sampling, and control-system compensation that can obscure fault signatures. Existing temporal detection methods have limited capacity to represent coupled faults across subsystems, track their evolution, and constrain language-model diagnoses with quantitative evidence. This paper proposes TGTV, an interpretable fault detection and traceability framework that integrates spatiotemporal graph learning with a large language model. A temporal convolutional network and a graph attention network with weighted fused adjacency extract multiscale temporal patterns and spatial dependencies among sensors. A hybrid temporal Transformer and variational autoencoder then evaluates window-level anomalies using prediction errors and reconstruction probabilities. Continuous-time Shapley attribution identifies critical sensors and high-contribution intervals while preserving the temporal continuity of fault evolution. A structured evidence package combines detection results, graph relationships, temporal attribution, sensor attribution, and operating context. An evidence-constrained diagnosis protocol uses this package to generate flight-level conclusions with traceable evidence identifiers. Cross-flight holdout experiments on the public ALFA fixed-wing flight dataset yield a window-level area under the precision-recall curve of 0.976, accuracy of 0.973, recall of 0.902, and F1 score of 0.920. Flight-level fault diagnosis achieves recall of 0.943 and an F1 score of 0.944. These results demonstrate accurate anomaly detection and quantitatively supported fault-tracing cues, providing a basis for interpretable post-flight fault review and analysis.
Abstract:Fixed-wing unmanned aerial vehicles experience operating-condition transitions, asynchronous sensor sampling, and control-system compensation that can obscure fault signatures. Existing temporal detection methods have limited capacity to represent coupled faults across subsystems, track their evolution, and constrain language-model diagnoses with quantitative evidence. This paper proposes TGTV, an interpretable fault detection and traceability framework that integrates spatiotemporal graph learning with a large language model. A temporal convolutional network and a graph attention network with weighted fused adjacency extract multiscale temporal patterns and spatial dependencies among sensors. A hybrid temporal Transformer and variational autoencoder then evaluates window-level anomalies using prediction errors and reconstruction probabilities. Continuous-time Shapley attribution identifies critical sensors and high-contribution intervals while preserving the temporal continuity of fault evolution. A structured evidence package combines detection results, graph relationships, temporal attribution, sensor attribution, and operating context. An evidence-constrained diagnosis protocol uses this package to generate flight-level conclusions with traceable evidence identifiers. Cross-flight holdout experiments on the public ALFA fixed-wing flight dataset yield a window-level area under the precision-recall curve of 0.976, accuracy of 0.973, recall of 0.902, and F1 score of 0.920. Flight-level fault diagnosis achieves recall of 0.943 and an F1 score of 0.944. These results demonstrate accurate anomaly detection and quantitatively supported fault-tracing cues, providing a basis for interpretable post-flight fault review and analysis.
21 pages, 8 figures, 4 tables. Submitted to IEEE Sensors Journal
Machine Learning (cs.LG)
arXiv:2610.38106 [cs.LG]
(or arXiv:2610.38106v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2610.38106 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 317 cuvinte
Subiecte conexe
SpaTeoGL: Spatiotemporal Graph Learning for Interpretable Seizure Onset Zone Analysis from Intracranial EEG
arXiv · 12 iun. 2026
EraseSAE: Surgical Concept Erasure in Text-to-Video Diffusion Models via Sparse Autoencoders
arXiv · 5 sept. 2026
CaC: Advancing Video Reward Models via Hierarchical Spatiotemporal Concentrating
arXiv · 28 sept. 2026
Perceiving and Predicting Dynamic Sports: Spatiotemporal Intelligence in Vision Models for Egocentric Tennis Video
arXiv · 28 oct. 2026