Digital-Twin-Driven Performance Assessment for IoT-Integrated Autonomous Driving Systems

Digital twin technology is emerging as a key innovation to enhance efficiency and performance in the automobile industry. Despite its potential, several challenges persist, notably the need for accurate performance evaluation and robust predictive capabilities. This article addresses these challenges by presenting a comprehensive model that delineates the vehicle metrics essential for performance evaluation. The novelty of this research lies in the integration of multiple advanced techniques, including a Naive Bayes model for categorizing data segments for quantitative analysis, spatial-temporal mining and regression analysis to abstract temporal data for deeper evaluation, and recurrent neural network (RNN) technology to ensure robust predictive capabilities. Experimental validation in a simulated environment comprising 50 250 data segments demonstrates the efficacy of the proposed model, showing significant improvements in performance metrics. Prediction analysis yields promising results with high Specificity (92.44%), Sensitivity (95.81%), Precision (94.33%), and F1-score (88.33%). Notably, the proposed model achieves temporal efficiency with a minimum time delay of 99.83 s, underscoring its effectiveness in real-time assessment of driverless automobile performance.