Traffic Delay Reduction at Highway Diverges Using an Advance Warning System Based on a Probabilistic Prediction Model

Abstract

This paper presents an on-board advance warning system for vehicles based on a probabilistic prediction model that advises them on when to change lanes to reach a highway diverge on time. The system is based on a model that estimates the probability of reaching a goal state on the road using one or multiple lane changes. This estimate is based on several traffic-related parameters such as the distribution of inter-vehicle headway distances as well as driver-related parameters like lane change duration. For an upcoming diverge, the advance warning system uses the model to continuously calculate the probability of reaching it and advise the driver to change lanes when the probability dips below a certain threshold. To evaluate the performance of the proposed system in reducing traffic delay at highway diverges, it was used on a segment of a four-lane highway to advise vehicles taking an off-ramp on when to change lanes. Results show that using the proposed system reduces average delay up to 6% and maximum delay up to 16%, depending on traffic flow and the ratio of vehicles taking the off-ramp.