In the last decade, the use of Driver-in-the-Loop (DiL) simulators has significantly increased in research, product development, and motorsports. To be used as a verification tool in research, simulators must show a level of correlation with real-world driving for the chosen use case. This study aims to assess the validity of a low-cost, limited travel Vehicle Dynamics Driver-in-Loop (VDDiL) simulator by comparing on-road and simulated driving data using a statistical evaluation of longitudinal and lateral metrics. The process determines if the simulator is appropriate for verifying control strategies and optimization algorithms for longitudinal vehicle dynamics and evaluates consistency in the chosen metrics. A validation process explaining the experiments, choice of metrics, and analysis tools used to perform a validation study from the perspective of the longitudinal vehicle model is shown in this study.
Validity is measured statistically, where a simulator is said to have absolute validity if the chosen metrics are not statistically different between on-road and simulator driving. Metrics showing statistically significant differences but with a consistent trend imply relative validity. In this study, an instrumented vehicle is used to collect on-road driving data in a defined urban route with 14 drivers with no professional driving experience. The vehicle model used in the simulator was matched to the actual vehicle in the longitudinal dynamics using CarSim. The urban route used for on-road driving was replicated in the virtual world with geometric accuracy.
VDDiL was rated to have moderate consistency with the real-world driving experience implying medium physical fidelity with a combination of absolute and relative validity for longitudinal metrics. On average, the speed-based metrics in the simulator were approximately 5 mph higher, with acceleration and jerk-based metrics about 1.5 - 3 times of corresponding on-road metrics.