Estimating Vehicle Suspension Characteristics for Digital Twin Creation with Genetic Algorithm

Abstract

Usage of simulation techniques like Vehicle-in-the-Loop, Scenario-in-the-Loop, and other mixed-reality systems are becoming inevitable in autonomous vehicle development, particularly in testing and validation. These methods rely on using digital twins, realistic representations of real vehicles, and traffic in a carefully rebuilt virtual world. Recreating them precisely in a virtual ecosystem requires many parameters of real vehicles to follow their properties in a simulation. This is especially true for vehicle dynamics, where these parameters have high impact on the simulation results. The paper's objective is to provide a method that can help reverse engineering a real car's suspension characteristics with the help of a genetic algorithm. A detailed description of the method is presented, guiding the reader through the whole process, including the meta-heuristic function's settings and how it interfaces with IPG Carmaker. The paper also presents multiple measurements, which can be effortlessly recreated without expensive devices or the need to disassemble any vehicle parts. Measurements are reproduced in two separate simulation tools with special scenarios providing an efficient way to analyze and verify the results. The provided method creates vehicle suspension characteristics with adequate quality, opening up the possibility to use them in the creation of digital twins or creating virtual traffic with realistic vehicle dynamics for high-quality visualization. Results show satisfying accuracy when tested with OpenCRG.