Data-Consistent Fuzzy Approach for Online Driver Behavior Under Information Provision

Abstract

A fuzzy approach to predict online driver routing decisions under information provision is proposed. The problem is characterized by subjectively interpreted and linguistic data, limited data on qualitative variables, and the need for computational efficiency to enable real-time deployment in a traffic control framework. The appropriateness of a fuzzy model to address these characteristics is highlighted. The proposed model enables theoretical consistency with the underlying behavioral mechanisms revealed through the available data. This is done by transforming the probabilities obtained from the data to possibilities, and then constructing membership functions. An S-shaped curve is proposed as a more robust behavioral indicator than the often-used triangular and trapezoidal shapes for these functions, while retaining their analytical tractability, computational efficiency, and ease of construction. In general, the approach suggests membership functions that are consistent with actual data. The model is analyzed using on-site stated preference data on variable message sign driver route diversion attitudes and is compared with a probabilistic discrete choice model for the same data. The results indicate that the fuzzy model performs more robustly for these data. When its computational efficiency and modularity are factored in, it may represent a better alternative to model online driver routing behavior in the context of real-time deployment.