SDVDiag: Multimodal Causal Discovery for Online Diagnosis in Software-defined Vehicles

2026-06-13

AI Summary

This work addresses the challenges of root cause localization in software-defined vehicles, where complex multi-hop causal chains, reliance on single-modality data, and lack of online diagnostic capability hinder existing approaches. To overcome these limitations, we propose a shared embedding framework that fuses log and metric modalities to construct a dynamically updatable causal graph, coupled with an anomaly-triggered mechanism enabling continuous online diagnosis. By innovatively integrating multimodal embedding fusion with an anomaly-driven online triggering strategy, our method supports automatic, real-time multi-hop root cause tracing. Experimental evaluation on an automated parking platform demonstrates that, compared to single-metric baselines, our approach reduces the number of causal graph edges from 182 to 134; after 60 rounds of human-in-the-loop feedback optimization, diagnostic reward improves by 2.4×, successfully identifying true upstream root causes two hops away.

Abstract

The transition toward software-defined vehicles concentrates an increasing share of vehicle functionality into distributed software services, where failures propagate through service dependencies and the surface symptom is often several causal hops away from the underlying defect. Existing approaches to causal root-cause analysis in such systems address this only partially: they typically reason over a single observability modality and operate in an offline, operator-driven mode that does not match the demands of continuous vehicle operation. This paper presents SDVDiag, a multimodal causal-discovery pipeline that fuses log-based and metric-based service representations into a shared embedding space before graph construction, coupled with an anomaly-driven trigger that converts the diagnostic platform from a manually operated batch tool into a continuously running online system. Evaluation on an Autonomous Valet Parking testbed shows that the multimodal pipeline produces sparser causal graphs than a metrics-only baseline (134 vs. 182 edges on average) and consistently outperforms it in edge-weighted reward against an expert knowledge graph at every stage of human-feedback refinement, showing a 2.4-fold improvement over the baseline after 60 feedback queries. An end-to-end fault-injection scenario further demonstrates that the integrated trigger correctly recovers a true root cause located two causal hops upstream of the observable symptom.

Problem

• software-defined vehicles
• causal discovery
• root-cause analysis
• multimodal observability
• online diagnosis

Innovation

• multimodal causal discovery
• online diagnosis
• software-defined vehicles
• anomaly-driven trigger
• causal graph embedding