Random parameter analysis of discretionary lane-changing decision models to account for the driver heterogeneity

초록

This study presents human lane-changing models with random parameters at the operational level, focusing on capturing driver heterogeneity. For three existing operational models - an acceleration-based model, a gap-based model, and a payoff-based model - we assess the validity and effectiveness of integrating random parameters on the models' ability to account for diverse drivers' lane-changing behaviours. First, our findings indicate that the inclusion of random parameters significantly improves model performance to accurately reproduce the real lane-changing behaviours. Second, the model calibration results offer an approach to more intuitively account for driver diversity. Consequently, when random parameters are incorporated, all three models result in enhanced explanatory power for real-world lane-changing behaviours. Among the three models, the acceleration-based model with random parameters outperforms the others in both performance and explanatory power. The estimated distribution of random parameters also demonstrates driver diversity and enhances understanding of real-world lane-changing behaviours.

키워드