A Traffic Assignment Model Considering Heterogeneous Speed and Route Choice Behaviors under Mixed Driving Environments

This study aims at quantitatively evaluating the impacts of travelers'heterogeneous behaviors of connected autonomous vehicle (CAV) and human-driven vehicle (HDV) on road network performance. A traffic equilibrium assignment model is proposed under mixed driving environments, which considers the heterogeneity in joint speed-route choice behaviors of CAV and HDV. Specifically, the quantitative relationship is established between driving speed and perceived crash risk, travel time and speeding ticket risk. By using the quantitative relationship, a speed choice behavior model is then formulated based on utility theory. The model could analytically characterize the behavior heterogeneity in optimal speed selection and speed limit (SL) obeying decisions of CAV and HDV. The concept of path time surplus (PTS) is introduced to characterize travelers'non-compensatory decision rules when weighing path travel time and monetary cost. The user equilibrium (UE) and Logit-based stochastic user equilibrium (SUE) principles with PTS maximization are used to describe the heterogeneous route choice behaviors of CAV and HDV, respectively. The speed choice behavior would intertwine with the path choice behavior via the PTS. By utilizing variational inequality (VI) theory, an equivalent mixed traffic equilibrium assignment model is finally proposed. A heuristic double-layer loop iterative algorithm is also developed to solve the model. Nguyen-Dupuis network and Huainan road network are used to validate the proposed model and algorithm. The total travel time (TTT) and total accident risk (TAR) are compared and analyzed in a road network under different traffic regulation conditions. The results indicate that with the market penetration rate (MPR) of CAVs increasing from 20% to 80%, the TTT shows an increasing trend in both 40 km/h and 50 km/h SL scenarios. Meanwhile, the TAR would increase first and then decrease across all SL scenarios. Under a high SL scheme (80 km/h), the TTT would decrease as the CAV traffic conversion factor reduces from 0.9 to 0.1, while the TAR exhibits a rising trend. Under a low SL scheme (40 km/h), however, the influences of CAV traffic conversion factor would become very insignificant. As the SL value increases from 40 km/h to 80 km/h, both the TTT and TAR decrease initially, and then stabilize or increase. The numerical results demonstrate that an appropriate speed limit scheme can reduce both TTT and TAR simultaneously.