Extending Artery/OMNeT++ for Evaluating V2X Sensor Fusion Algorithms with Infrastructure Support in Cooperative Advanced Driver Assistance Systems

Artery is a well-known V2X (Vehicle-to-Everything) simulation framework for evaluating vehicular communication systems. It is built on top of the OMNeT++ simulation environment. Although the framework is considered one of the best V2X simulators, serious deficiencies related to the error model in the representation of positioning were identified, prohibiting modeling of various advanced sensor fusion solutions and related ADAS technologies that rely on Cooperative Awareness Messages (CAMs), Collective Perception Messages (CPMs), and Vulnerable Road User Awareness Messages (VAMs). The main contribution lies in error models that were introduced to the framework to incorporate realistic measurement uncertainties, enhancing the fidelity of V2X message simulations and enabling a more accurate representation of real-world GPS and sensor errors. This paper extends Artery by addressing these challenges through improved modeling of position error, implementing geodetic-to-Cartesian coordinate transformations, correcting inconsistent scaling methods, and improper validation mechanisms leading to positional mismatches and unreliable object perception within the simulations. The study introduces robust geodetic transformations ensuring precise coordinate alignment between ego vehicles, roadside units (RSUs), and detected objects. A key enhancement includes extending RSU capabilities within Artery, enabling RSUs to actively generate and transmit CPM messages based on their own sensor data rather than solely receiving vehicle-originated messages. This allows RSUs to detect non-V2X-equipped vehicles and vulnerable road users (VRUs), allowing the RSU to share the data via V2X, thereby improving situational awareness and cooperative perception. Our framework treats RSUs as active CPM generators to model infrastructure-supported collective perception, seamlessly integrates configurable error models for the kinematic state vector of V2X messages as well as for perceived objects, and exposes a realistic and appropriate approach for V2X-based object level sensor fusion algorithms. The paper presents a comprehensive evaluation that includes reproducible scenarios, visualization tools, and metrics for perception accuracy and scalability under varying vehicle and RSU densities to enable systematic benchmarking of collective perception for cooperative ADAS applications. The validation is conducted through diverse urban and industrial scenario simulations. The results show that the refined coordinate handling significantly improves the model of object positioning and the reliability of the synthetic message generation, while the applied error models implement highly realistic behavior in a configurable manner.