Digital Twin-Assisted Explainable AI Framework for Signal Integrity Prediction and Communication Reliability in CAN FD-Based Autonomous Vehicle Networks

Abstract

The increasing adoption of autonomous vehicles has created a demand for faster and more reliable in-vehicle communication systems. Controller Area Network with Flexible Data Rate (CAN FD) improves communication efficiency by supporting higher data transmission rates and larger payloads than traditional CAN networks. However, maintaining signal integrity and communication reliability remains challenging because of electromagnetic interference, network congestion, timing variations, and hardware-related disturbances. Existing monitoring methods mainly detect faults after they occur and often provide limited information about the factors affecting communication performance.
This paper proposes a Digital Twin-Assisted Explainable Artificial Intelligence (XAI) framework for predicting signal integrity and evaluating communication reliability in CAN FD-based autonomous vehicle networks. The framework integrates a digital twin of the communication network with an intelligent prediction model to continuously analyze network conditions using multiple communication indicators, including signal quality, latency, frame error rate, bus utilization, and message loss. To improve transparency and user confidence, SHAP-based Explainable AI is incorporated to identify the contribution of each communication parameter to the prediction process. The predicted communication state is then classified into Normal, Degraded, and Critical conditions for reliability assessment.
The proposed approach enables proactive communication monitoring by identifying potential network degradation before it significantly affects vehicle operation. Compared with conventional threshold-based methods, the framework provides more accurate prediction while offering clear explanations for model decisions. The study demonstrates that combining Digital Twin technology with Explainable AI can improve communication reliability assessment and support the development of trustworthy intelligent monitoring systems for future autonomous vehicle networks.