Evaluation of probabilistic demands usage for the online dial-a-ride problem

Abstract

The objective of this study is to investigate if it is possible to reduce the operational cost of an online Demand Responsive Transportation System (DRT) by using probabilistic trip demands while leaving the optimization algorithm intact. The idea is that we use probabilistic demands in order to predict actual ones. If the prediction is accurate enough then the DRT’s vehicle fleet reassigned in a better state. The innovation lies in the assumption that, given enough historical data on trip demands, the system’s online nature can be reduced, resulting in a better solution (problem objective). The basic steps of the proposed methodology are: (a) Based on a real historical data set, a demand distribution probability created to describe online DRT’s demands behavior. (b) During operation, for each incoming demand, create a set of additional probabilistic demands based on the distribution in (a) and calculate an initial solution. (c) Remove the probabilistic demands and optimize the solution further. (d) Comparatively analyze these solutions against those that would be produced without the use of probabilistic demands. The study revealed that using probabilistic demands improved the solutions in terms of cost (objective). Test data were recorded during an actual 30-day online DRT operation at the same location, the former municipality of Philippi in northern Greece. © Springer Nature Switzerland AG 2019.