Vehicle scheduling for on-demand vehicle fleets in macroscopic travel demand models

Johann Hartleb,Emely Richter,Markus Friedrich

doi:10.1007/s11116-021-10205-4

Johann Hartleb, Emely Richter + Show 1 more

Open Access

https://doi.org/10.1007/s11116-021-10205-4

Copy DOI

Abstract

The planning of on-demand services requires the formation of vehicle schedules consisting of service trips and empty trips. This paper presents an algorithm for building vehicle schedules that uses time-dependent demand matrices (= service trips) as input and determines time-dependent empty trip matrices and the number of required vehicles as a result. The presented approach is intended for long-term, strategic transport planning. For this purpose, it provides planners with an estimate of vehicle fleet size and distance travelled by on-demand services. The algorithm can be applied to integer and non-integer demand matrices and is therefore particularly suitable for macroscopic travel demand models. Two case studies illustrate potential applications of the algorithm and feature that on-demand services can be considered in macroscopic travel demand models.

Highlights

In densely populated areas public transport is more efficient than private means of transport for two reasons
This paper presents an algorithm for the vehicle scheduling process of on-demand services, which can be embedded in macroscopic travel demand models
The contribution of this paper is twofold: First, we develop a vehicle scheduling algorithm to estimate the vehicle fleet size of on-demand services efficiently

Summary

Introduction

In densely populated areas public transport is more efficient than private means of transport for two reasons. This paper presents an algorithm for the vehicle scheduling process of on-demand services, which can be embedded in macroscopic travel demand models. The scheduling step identifies empty vehicle trips which are required for vehicle relocation Approaches to replicate this step differ for microscopic and macroscopic travel demand models as discussed in the following. Most vehicle scheduling contributions consider an operational setting and aim at providing an optimal solution for a certain demand situation Similar to early approaches as presented in Bodin (1983), we model the vehicle scheduling problem as a flow problem This design choice is motivated by the huge demand data of realistic instances considered in this paper that include up to 100 million vehicle trips. Calling the function FlowConservation() in line 8 ensures that the number of arriving and departing vehicles at the considered node vzt are equal and, that the vehicle flow is feasible

Initialisation

Output

Findings

Conclusion and outlook