Network-Level Traffic and Safety Effects of Automated Vehicle Market Penetration Under Take-Over Events: A Microsimulation Study

Abstract

This study investigates the network-level traffic and safety effects of increasing Market Penetration Rate (MPR) of SAE Level 2 and Level 3 automated vehicles under take-over conditions. A calibrated microsimulation model of a real-world 50 km highway corridor in Greece (Nea Odos) was used to develop 18 scenarios combining two take-over contexts (Lane Closure and ODD Exit), two Time Budget (TB) configurations, and four MPR levels (25–100%), complemented by two baselines. Traffic performance was assessed through five network-level indicators (speed, delay, lane-changing frequency, travel time, and density). Safety was evaluated through 674 conflict events extracted via the Surrogate Safety Assessment Model (SSAM), using Time-to-Collision (TTC) as the primary indicator. MPR is the dominant feature of both dimensions. Increasing MPR produces a monotonic reduction in speed (−19.2% at 100% MPR), a sharp decline in lane-changing (−59.8%), and increases in density (+21.1%) and travel time (+22.4%), delay peaks non-monotonically at 50% MPR (+107.9%). Spearman correlations between MPR and the traffic indicators are very strong for speed, lane changes, travel time, and density, while delay time yields a weaker correlation reflecting its non-monotonic response. All operating points remain within free-flow conditions throughout the MPR range. The MPR effect on TTC is marginal (p = 0.065) but non-monotonic, with 75% MPR yielding significantly higher mean TTC than 50% and 100% MPR. Neither TB duration nor TOR context produces a statistically significant effect on traffic indicators or TTC distributions, with negligible effect sizes in both cases. Traffic costs of mixed-traffic automation appear at partial penetration, while network-level safety benefits remain limited and non-linear.