Modeling Lane-Choice Behavior to Optimize Pricing Strategy for HOT Lanes: A Support Vector Regression Approach

Abstract

High-occupancy toll (HOT) lanes can better utilize road resources by allowing low-occupancy vehicle (LOV) drivers to pay a toll and use high-occupancy vehicle (HOV) lanes. In such a system, the toll price plays a key role in dynamically allocating LOVs over the HOT and general purpose (GP) lanes to improve the overall system performance. This paper presents a method to model the lane-choice behavior and dynamically determine the toll price in response to traffic conditions. First, a combined model of k-fold cross validation (k-CV), particle swarm optimization (PSO), and support vector regression (SVR) is proposed to predict the lane-choice behavior of LOVs, which is an important factor to determine the optimal toll. Five-minute tolling data collected from Interstate 405 have been used for the study. Compared with five different methods, this combined model showed the highest accuracy of 92.37%. Based on the model, the relationship between the toll price, the real-time traffic speed and volume in GP and HOT lanes, and the number of lane-changing LOVs can be identified. Next, a toll pricing optimization model was developed in order to minimize the total travel time as well as determine the corresponding optimal volume of lane-changing LOVs. Based on the previously identified relationship, an optimal toll can be calculated dynamically according to the optimal number of lane-changing LOVs and the real-time speed and volume. The study further analyzes the benefit of the proposed price optimization method in terms of travel time saving. The results show that the dynamic tolling strategy helps to better utilize the capacity of HOT lanes as well as bring significant reduction in the total travel time.