Connected and Automated Vehicle Evaluation using Real-Twin Auto-Calibrated Traffic Microsimulation

Abstract: Traffic microsimulation models are widely leveraged to study the mobility, safety, energy, and environmental impacts of emerging vehicle and intelligent transportation infrastructure technologies. In recent years, these models have become crucial for developing and evaluating Connected and Autonomous Vehicle (CAV) technologies and applications. However, the development of these microsimulation models can be a time-consuming process, primarily due to data integration, simulation network development, and model calibration. To address this, Real-Twin tool has been developed. Real-Twin provides a capability of automatically generating calibrated scenarios in three widely used traffic microsimulation software—SUMO, VISSIM, and AIMSUN. One of the objectives of the Real-Twin tool is to ensure consistent results across different software platforms when evaluating emerging technologies such as CAV applications. In this work, the developed Real-Twin tool is applied to generate simulation scenarios across these platforms to evaluate the consistency of the performance of a CAV eco-approach algorithm. Firstly, this study presents the Real-Twin workflow for the automated generation of calibrated microscopic traffic simulation scenarios. Secondly, the capability of Real-Twin to automatically calibrate traffic simulation scenarios to make them consistent across different microsimulation software is demonstrated using a real-world traffic scenario. The automatic calibration process includes the calibration of traffic demand and turn movement ratios at locations with no data availability, such as a driveway or unsignalized intersection, as well as the calibration of driver behavior parameters. For automatic calibrations, the tool is developed to use three optimization algorithms—Genetic Algorithm, Simulated Annealing, and Tabu Search. Lastly, the Real-Twin-generated simulation scenario is applied to evaluate an external CAV eco-approach algorithm to demonstrate Real-Twin scenario consistency to study CAV applications. Real-Twin enables researchers and engineers to efficiently develop realistic, calibrated microscopic simulation models that are consistent across various traffic simulation software, both open-source and commercial, for CAV studies.