Search

Throughout the last decade, the advent of novel mobility services such as ride-hailing, car-sharing, and ride-sharing has shaped urban mobility. While these types of services offer flexible on-demand transportation for customers, they may also increase the load on the, already strained, road infrastructure and exacerbate traffic congestion problems. One potential way to remedy this problem is the increased usage of dynamic ride-sharing services. In this type of service, multiple customer trips are combined into share a vehicle simultaneously. This leads to more efficient vehicle utilization, reduced prices for customers, and less traffic congestion at the cost of slight delays compared to direct transportation in ride-hailing services. In this thesis, we consider the planning and operation of such dynamic ride-sharing services. We present a wider look at the planning context of dynamic ride-sharing and discuss planning problems on the strategical, tactical, and operational level. Subsequently, our focus is on two operational planning problems: dynamic vehicle routing, and idle vehicle repositioning. Regarding vehicle routing, we introduce the vehicle routing problem for dynamic ridesharing and present a solution procedure. Our algorithmic approach consists of two phases: a fast insertion heuristic, and a local search improvement phase. The former handles incoming trip requests and quickly assigns them to suitable vehicles while the latter is responsible for continuously improving the current routing plan. This way, we enable fast response times for customers while simultaneously effectively utilizing available computational resources. Concerning the idle vehicle repositioning problem, we propose a mathematical model that takes repositioning decisions and adequately reflects available vehicle resources as well as a forecast of the upcoming trip request demand. This model is embedded into a real-time planning algorithm that regularly re-optimizes the movement of idle vehicles. Through an adaptive parameter calculation process, our algorithm dynamically adapts to changes in the current system state. To evaluate our algorithms, we present a modular simulation-based evaluation framework. We envision that this framework may also be used by other researchers and developers. In this thesis, we perform computational evaluations on a variety of scenarios based on real-world data from Chengdu, New York City, and Hamburg. The computational results show that we are able to produce high-quality solutions in real-time, enabling the usage in high-demand settings. In addition, our algorithms perform robustly in a variety of settings and are quickly adapted to new application settings, such as the deployment in a new city.

With the increasing interest in creating Smart Cities, traffic speed prediction has attracted more attention in contemporary transportation research. Neural networks have been utilized in many studies to address this problem; yet, they have mainly focused on the short-term prediction while longer forecast horizons are needed for more reliable mobility and route planning. In this work we tackle the medium-term prediction as well as the short-term. We employ feedforward neural networks that combine time series forecasting techniques where the predicted speed values are fed into the network. We train our networks and select the hyper-parameters to minimize the mean absolute error. To test the performance of our method, we consider two multi-segment routes in Istanbul. The speed data are collected from floating cars for every minute over a 5-month horizon. Our computational results showed that accurate predictions can be achieved in medium-term horizon.

Bus rapid transit systems in developing and newly industrialized countries are often operated at the limits of passenger capacity. In particular, demand during morning and afternoon peaks is hardly or even not covered with available line plans. In order to develop demand-driven line plans, we use two mathematical models in the form of integer programming problem formulations. While the actual demand data is specified with origin-destination pairs, the arc-based model considers the demand over the arcs derived from the origin-destination demand. In order to test the accuracy of the models in terms of demand satisfaction, we simulate the optimal solutions and compare number of transfers and travel times. We also question the effect of a selfish route choice behavior which in theory results in a Braess-like paradox by increasing the number of transfers when system capacity is increased with additional lines.