This paper proposes a framework to enhance the fidelity of Vehicle-in-the-Loop (VIL) simulation by introducing realistic sensor modeling and environmental reconstruction. Despite reflecting real vehicle dynamics, conventional VIL has limitations in th...
This paper proposes a framework to enhance the fidelity of Vehicle-in-the-Loop (VIL) simulation by introducing realistic sensor modeling and environmental reconstruction. Despite reflecting real vehicle dynamics, conventional VIL has limitations in that simplified sensor emulation and generalized virtual environments create a gap between simulation and reality (Sim-to-Real Gap).
To overcome this, this study reproduces LiDAR beam patterns and packet structures, and constructs a realistic environment based on point cloud maps using the Poisson Surface Reconstruction technique. Furthermore, to analyze the individual contributions of sensor and environmental realism, five configurations were evaluated in Euro NCAP Car-to-Car scenarios, and the simulation fidelity was quantified using the normalized RMSE and Pearson correlation coefficient of six key signals.
Experimental results demonstrated that environmental realism yielded the greatest improvement in fidelity by stabilizing positioning through accurate geometric reconstruction. Realistic LiDAR modeling played a complementary, fine-tuning role in improving perception accuracy, and the final integrated configuration achieved a fidelity close to the ground truth. Consequently, the findings suggest that maximizing the realism of both sensor and environmental domains is essential for the reliable and efficient validation of autonomous driving systems.