Automatic Calibration of Behavior Parameters for Variable Message Signs-Based Route Guidance Consistent with Driver Behavior

摘要

Variable message signs (VMS) can be used to divert traffic to less congested areas of roadnetworks and enhance network performance. Central to these goals are ensuring the increasedacceptability of the suggested guidance and improving the credibility of VMS information amongdrivers. This study develops a new model for quantifying the effect of VMS messages on drivers’en-route diversion behavior for real-time VMS-based behavior-consistent route guidance. The modelincorporates attributes that can be obtained on-line, including traffic information, traffic flow,weather, and incident duration. The parameters, observation noise, and system noise characterizingthis model are calibrated dynamically according to real-time detected traffic data and adaptiveKalman filtering theory. In the calibration, the consistency between time-dependent actual andpredicted traffic states is verified. The generalized structure of the calibration model enables it tosimultaneously incorporate other sources of state inconsistency, such as traffic flow modelparameters. The results indicate that the calibration model enhances the accuracy of traffic stateprediction, thereby increasing the effectiveness of the proposed route guidance. The model is alsosuitable for application to real-world situations because it can more accurately estimate and predictthe likely route choices of drivers using aggregate-level traffic data. Such data can be used to trackthe evolutionary process of en-route diversion behavior, as well as design meaningful route guidancecontrol goals and strategies for traffic decision makers.