A Proactive User Optimum-Oriented Route Guidance System Incorporating Individual Users' Route Choice Preferences

The route guidance system is one of the most effective ways of reducing traffic congestion. Existing route guidance systems are mostly reactive and self-interested, and simplify drivers’ route choice preferences by assuming drivers only pursue the shortest travel time/distance. These features keep the route guidance system from adequately accommodating drivers’ heterogeneous route choice preferences, and also from proactively avoiding congestion. Researchers designed more advanced route guidance systems that can optimize transportation efficiency or proactively avoid congestion, but drivers’ heterogeneous route choice preferences have not been fully incorporated. Because of the lack of considering drivers’ preferences and possible consequent reactions to the guidance, simplifying drivers’ preferences in route guidance systems may lead to the discrepancy between the expected and actually generated traffic conditions, and might also undermine the performance of traveler information related intelligent transportation system strategies. Meanwhile, emerging information technologies applied in the transportation domain make it possible to collect drivers’ behavior data even at the individual level, which provides great resources for analyzing drivers’ preference heterogeneity. Therefore, this research proposes a proactive user optimum-oriented route guidance system that incorporates individual users’ preferences in order to achieve users’ better satisfaction and transportation system performance improvement. Individual drivers’ route choice preferences can be captured from his/her historical preference data, then are adequately considered and coordinated by incorporating individual route choice models into the process of searching for user optimal conditions. Then, routes recommendations can be generated for each user based on the user optimal condition in which no user can improve his/her experience by changing to another route.

In order to make the route guidance system accurately capture, predict thus fully consider each driver’s route choice preference, several approaches were first explored to establish individual route choice models, including traditional discrete choice model, mixed logit model, support vector machine and multi-task linear model adaptation (MT-LinAdapt). Three stated preference datasets collected from 102 participants as well as three synthetic datasets were used to evaluate and compare the performance of different approaches. The evaluation showed that MT-LinAdapt has the highest prediction accuracy which is up to 8% and 18% higher than other approaches when there is adequate and inadequate historical preference data, respectively. Additionally, it has implementation feasibility advantages: (1) does not require segmentation criteria (e.g., sociodemographic information) to distinguish drivers’ heterogeneous preferences; (2) also works well with limited amount of individual preference data and very heterogeneous preference data; and (3) can be updated in real time as individuals’ preference data accumulates. Therefore, MT-LinAdapt is recommended to establish the individual-level route choice preference models in the application of route guidance systems.

The framework of a proactive user optimum-oriented route guidance system was proposed which contains two components: (1) established individual route choice models and (2) incorporated individuals’ route choice preferences in searching for user optimum conditions. Such user optimum conditions are used as guidance information. With a commonly used Sioux Falls network and user population whose preferences were synthesized from surveyed participants, the proposed route guidance system at both perfect and imperfect market penetration rates was compared to existing route guidance strategies including travel time based real-time guidance and travel time based User Equilibrium (UE) guidance. An evaluation platform which is made of a traffic simulation module (DTAlite) and a route choice module (Matlab) was established and utilized to conduct the evaluation. The proposed route guidance system demonstrated advantageous performance in aspects of users’ satisfaction (up to 22% more satisfied users), system mobility and sustainability (up to 10% of travel time reduction and up to 42% of delay reduction), and future traffic conditions estimation (up to 70% links having more accurate volume estimation). At imperfect market penetration rates, users of the proposed route guidance system interact with those drivers who use real-time guidance system and who take habitual routes. The generated performance improvement gradually increases as the market penetration rate increases. In addition, the proposed route guidance system has the potential to be extended for additional traffic control and management strategies so that further system performance improvement could be achieved, such as personalized incentive scheme.
The proposed route guidance system framework and the evaluation results extend the existing literature and have broader impacts on the following aspects: (1) Established individual route choice models to capture individual drivers’ route choice preferences; (2) Proposed a proactive user optimum-oriented route guidance system for system performance and users’ satisfaction improvements; (3) Prepared the foundation for designing personalized traffic control and management strategies that have great potential to further improve transportation system performance.