Behavior-consistent deployable real-time traffic routing under information provision.

摘要

This research proposes a new methodological perspective to address the real-time information-based control of vehicular traffic systems problem. Existing approaches do not realistically consider the interdependencies between driver response behavior and the information provision strategies. The proposed methodology determines effective and realistically deployable information strategies by explicitly factoring the controller's objectives and its estimation of driver response behavior in the generation of these strategies, thereby circumventing realism issues with existing dynamic traffic assignment (DTA) models that pre-specify driver behavior. This leads to a behavior-consistent deployable traffic routing approach that can enhance system performance in light of the actual driver behavior. It illustrates the need for explicit demand-supply integration. The proposed approach uses a controller-estimated driver behavior model based on aggregate behavioral if-then rules to estimate driver route choices under information provision. This circumvents the need for data at the individual driver level, and the calibration is based on measurable traffic data. The proposed behavior-consistent approach enables the classification of routes based on their relevance to the drivers and controller, leading to the definition of controllable routes which provides a realistic deployment mechanism to simultaneously enhance system performance and driver compliance in a behavior-consistent manner. The concept of behavior-consistency gap is developed to illustrate the behavioral inadequacies of DTA models. Experimental insights suggest that the behavior-consistent approach provides superior performance compare to DTA models. This is because lack of behavior-consistency can potentially deteriorate system performance as the controller may over- or under-recommend, or recommend routes that are not considered by the drivers. A key insight from this study is that there are trade-offs between the number of driver-preferred routes considered by the controller for routing and the quality of routes relative to the controller objective. That is, even when routing is performed in a behavior-consistent manner, higher compliance rates need not necessarily translate to better system performance as it also depends on the quality of the routes. From a computational standpoint, the behavior-consistent approach can be parallelized at the origin-destination level enabling real-time deployment. In addition, model parameters are optimized to further enhance the computational efficiency of the proposed solution framework.