Evaluating Driver Speeding Compliance During Curve Speed Warning (CSW) System Application

Speeding on horizontal curves is a leading contributor to roadway departure crashes in the U.S. Unlike straight-section speeding, exceeding curve advisory limits rapidly degrades vehicle handling and increases rollover risk. Historically, behavioral interventions have been limited to static signage, which fails to provide the real-time, personalized feedback necessary to influence in-the-moment decision-making. Emerging Connected Vehicle (CV) applications enable real-time, vehicle-specific warnings, yet empirical evidence on curve speed warning performance, behavioral conditioning, and compliance in naturalistic CV corridors remains sparse. This study addresses these gaps by evaluating whether in-vehicle Curve Speed Compliance (CSPCOMP) alerts (1) produce immediate speed reductions, (2) generate progressive compliance improvements over time, and (3) exhibit consistent effects across different sub-corridors and driver groups.

This paper evaluates the Curve Speed Compliance (CSPCOMP) application deployed during Phase 2 of the Connected Marysville-Dublin Pilot. Using 144 instrumented vehicles in a Vehicle-to-Infrastructure (V2I) corridor, the system provided heads-up display alerts when vehicles exceeded advisory limits (25 and 30 mph). We utilized Wilcoxon signed-rank testing to evaluate speed profiles within 2.5- and 5-second windows pre- and post-alert.

Findings indicate that CSPCOMP alerts consistently produced speed reductions of approximately 1.0 mph within 2.5 seconds and 2.0 mph within 5 seconds across all curve groups. Disaggregated analyses by sub-corridor confirmed these patterns, with Bonferroni post-hoc testing revealing significant time-interval and driver-group effects. Longitudinally, the study demonstrated a progressive compliance effect: the frequency of alerts declined over time while the proportion of drivers responding with speed reductions increased. This suggests that sustained exposure produces a durable shift in behavior, rather than a transient reaction.

Although absolute speed reductions may appear modest, they are highly significant given the nonlinear relationship between speed and crash risk at curves. This paper argues for the systematic integration of curve speed warning functionality into V2I deployments and provides a robust methodological framework for evaluating driver compliance in naturalistic connected environments.