Cyber-Physical Co-Simulation Frameworks for Autonomous Vehicles

Abstract

The integration of cyber‑physical co‑simulation frameworks is increasingly recognized as a foundational technique for advancing autonomous vehicle (AV) research and development. By tightly coupling traffic microsimulation, high‑fidelity vehicle dynamics, and realistic network emulation, such frameworks enable researchers and engineers to evaluate perception, planning, and control algorithms under conditions that closely mimic real‑world operations. This manuscript presents an enhanced exploration of a novel co‑simulation architecture built on ROS 2, orchestrating SUMO for urban traffic scenarios, CARLA for detailed vehicle and sensor modeling, and NS‑3 for end‑to‑end communication emulation. We detail the architectural components, synchronization mechanisms, and data‑exchange protocols that ensure sub‑50 ms round‑trip latencies and near‑real‑time execution. A comprehensive statistical analysis over 100 randomized scenarios—including latency, real‑time factor, CPU utilization, and memory footprint—is provided, with performance metrics rigorously profiled and confidence intervals computed. Validation against single‑simulator baselines demonstrates trajectory divergence under 0.05 m over 1 km, confirming high modeling fidelity. Finally, we discuss extensibility features, including plug‑and‑play modules for threat injection and hardware‑in‑the‑loop (HIL) integration, and identify future research avenues such as standardized FMI‑ROS interfaces and multi‑agent V2X coordination. Through this work, we contribute a robust, open‑source co‑simulation platform that bridges critical gaps in AV systems validation and paves the way for more resilient, scalable, and secure autonomous driving solutions.