Effective Prediction of Road Segment Occupancy for the Route-Reservation Architecture

The effect of traffic congestion within an urban region can be eliminated by curbing the number of vehicles that use the congested part of the transport network. To do so, earlier work by the authors introduced a novel route-reservation architecture that manages the vehicle departure times and makes the appropriate routing decisions so that vehicles arrive at their destination at the earliest possible time avoiding the congested parts of the network Menelaou et al. (2017a,b). Under the proposed route-reservation architecture, the future state of each road segment is predicted based on the received reservations and assuming that all vehicles travel at a constant speed which is set equal with the free-flow speed. Even within a homogeneous region, this assumption is not always valid which can lead to non accurate estimation of the time that a vehicle occupies a road segment which in turn affects the accuracy of the overall architecture. Therefore, the key objective of this work is to investigate different prediction schemes that improve the prediction-accuracy of vehicles travel times within each road segment. In this paper we explore two prediction methods, an Exponential Moving Average (EMA) and a Multiple Linear Regression (MLR) method. The performance of the two prediction schemes is also presented. Finally, realistic simulation results across an urban region of San Francisco demonstrate the gains that can be achieved applying the proposed prediction methods.