Home > Research > Publications & Outputs > Modeling and Optimization of Demand Responsive ...

Electronic data

View graph of relations

Modeling and Optimization of Demand Responsive Systems and Urban Congestion

Research output: ThesisDoctoral Thesis

Published
Publication date4/04/2014
Number of pages169
QualificationPhD
Awarding Institution
  • École Polytechnique Fédérale de Lausanne (EPFL)
Supervisors/Advisors
  • Geroliminis, Nikolas, Supervisor, External person
Award date4/04/2014
<mark>Original language</mark>English

Abstract

This study is motivated by planning on-demand transportation systems in large scale urban networks. We specifically handle emergency response systems and car-sharing services in congested urban areas. Our ultimate aim is to improve the quality of service for both emergency response and car-sharing systems
with the support of operational research tools.

In the first part of the thesis, we deal with a method that can evaluate the performance of spatial queueing systems with different server-customer service rates. More precisely, we propose two new spatial queueing models, which utilize service rates that are the functions of both the server providing the service and the customer receiving it. These two new models can be regarded as two extensions to the known hypercube queueing models. These models have a lot of application areas including improvement of emergency (e.g. ambulance, police, emergency repair) and transportation (taxicabs, on-demand transportation, para-transit) services in cities.

The first contribution of this part of the thesis to the literature is developing two hypercube models that apply different service rates according to the distance between server and the customer. In order to keep problems tractable, we assume that there are two different service rates for each server: The service rate when a server is serving its own region (i.e. intradistrict service) is different when the same server is serving outside its own region (i.e. interdistrict service). The second contribution is proposing an approximation algorithm for the problems that are intractable because of their size. We also test both methods inside some efficient heuristics to show the applicability of the two methods with algorithms for real case problems.

In the second part of the thesis, we work on improving the services in (non-floating) one-way (electric) car-sharing systems. We regard this problem in three different levels: Strategic, tactical and operational. Strategic decisions are regarded as the decisions related to the infrastructure (e.g. location and size of the stations) and tactical decisions are about the vehicles and the personnel (e.g. fleet and personnel size); whereas operational decisions are problems related to daily operations (e.g. relocation operations, personnel shift assignments). The first two levels of the decisions (i.e. strategic and tactical) are taken into consideration together in the first chapter of this part. A multi-objective (different objectives are applied for the users and the operator of the system) mixed integer linear programming formulation with its relaxation is proposed and solved for different scenarios. The model is applied on a real case, a car-sharing service in the city of Nice.

In the second chapter of this part, the operational problem of one-way car-sharing systems are handled. A mixed integer linear programming formulation is proposed for the problem which decides on initial vehicle locations, relocation operations and assignment of relocation personnel to the shifts. In order to keep the decisions robust, flexible and applicable, some extra soft constraints are added. We aim to improve flexibility of the service from the users point of view, it is assumed that the users might pick-up and drop-off vehicles earlier or later than their reservations occasionally. In order to cope with these situations, set of soft constraints are introduced to the model that keeps vehicles and empty spots at the right place at the right time. This chapter is still work in progress but is included in the dissertation to represent preliminary results.

In the last part of the thesis, we propose a tool for a parsimonious travel time estimation for the first two parts of the thesis. We mostly benefit from Variational theory and macroscopic fundamental diagram literature. We start with extending the prior work on the travel time estimation works on homogeneous networks and apply similar methods to the systems with heterogeneous system characteristics; i.e. link lengths, offset between traffic signals and incoming turns. The research is conducted for both unimodal and multimodal networks. More specifically, in this part of the research, we explore the effect of network parameters on the two key characteristics of macroscopic fundamental diagram: (i) the network capacity and (ii) the density range when the network capacity is maximum.

Although scarce data do not enable us to do, in the process of improving the tool we have reached to some conclusions that are applicable not only to travel time estimation but also to traffic in urban networks. A closed analytical formulation, that utilizes system characteristics, to calculate the density range of a homogenous network is proposed and proven. Then, the effects of the changes on the system characteristics are investigated for both homogeneous and heterogeneous networks. In addition, the effect of the incoming turns is modeled and its intensity is explored. In an extended research, similar investigations are conducted for the multimodal networks, i.e. networks with public transportation buses. These tools can be utilized for the development of hierarchical control strategies for large scale congested transport networks.