Safety effects of take-over requests on vehicle conflicts in realistic mixed traffic simulation

This study investigates how take-over-related dynamics in Automated Driving (AD) influence safety interactions in mixed traffic using a spatial Generalized Additive Model (GAM) applied to a calibrated microsimulation of a real highway corridor in central Greece. The simulation reproduced automated driving transitions (Level 2/3) and take-over events (Take-Over Request (TOR) → driver response → manual driving → re-engagement) under baseline, lane-closure, and ODD-exit conditions. Vehicle trajectories were analyzed with the Surrogate Safety Assessment Model (SSAM) to extract conflict-level Time-to-Collision (TTC) indicators, enabling a quantitative assessment of take-over safety across an entire network rather than at the driver or maneuver level. The fitted spatial GAM effectively captured the variability in safety outcomes and revealed significant effects of Market Penetration Rate (MPR), take-over Time Budget (TB), scenario context, speed limit, vehicle type (first or second in the conflict), and conflict geometry. Higher automation shares and speed limits were associated with longer TTC, reflecting smoother and more stable interactions, whereas take-over events consistently reduced TTC regardless of TB, confirming elevated short-term risk during control transitions. The spatial smooth revealed localized low-TTC zones near the TOR area and merge lanes. Methodologically, this study combines microsimulation, SSAM-based surrogate safety analysis, and multivariate spatial GAM modeling to quantify TOR effects on AD-Human Driven Vehicle (HDV) interactions at the network level. The framework bridges simulation and statistical analysis by linking modeled behavior with interpretable safety outcomes. Substantively, the findings show that automation benefits scale with market share but are constrained by transition management and roadway geometry, emphasizing the importance of spatially aware, take-over-sensitive safety strategies in mixed traffic.