Assessing the impact of driver compliance on traffic flow and safety in work zones amid varied mixed autonomy scenarios

The safety of work zones is a critical issue for drivers, transportation agencies, and governing authorities. In particular, the vehicles that perform lane changes in the proximity of the work zones involving lane closure, pose a significant threat to the safety of the public and the work zone workers, as they need to complete a forced merging. Yet, there is no comprehensive simulation framework to examine the work zone traffic safety under different compliance distributions of the drivers to the warning delivery for the work zone in a mix-autonomy operation of autonomous and human-driven vehicles. To fill this void, we present an integrated microsimulation framework to assess the correlation between the number of vehicles that perform late merge at the taper (LMT) and traffic mobility and safety under different empirical compliance distributions of the drivers to the warning delivery for the downstream work zone. We employ different work zone configurations to illustrate the relationship between late merges at the taper and performance indicators for traffic mobility and safety of the work zone under a variety of work zone configurations. Simulation results show that compliance distribution significantly impacts the number of late merges at the taper (LMTs) and thereby traffic safety and efficiency. Our findings demonstrate that when human-driven vehicles exhibit high compliance behavior to the merging warning signs, it can offset the impact of the lower percentage of market penetration rate (MPR) levels for autonomous operation to achieve comparable traffic safety and efficiency. We further employ the conflation of microsimulation observations and data-driven models to design a regression model to predict LMTs as an indicator for traffic conditions using the work zone configuration as input variables. In particular, we address the heterogeneity induced by the compliance distribution of drivers by sampling the data points from the distribution to capture the diversity in compliance behaviors of the drivers. Our findings can provide insights for practitioners and researchers regarding the optimal compliance distribution using the performance measurements demonstrated in this work.