A Ride-Sharing System with High Capacity Vehicles and Flexible Meeting Points

In many major cities, fixed route transit systems such as bus and rail serve millions of trips per day. These systems have people collect at common locations (the station or stop), and board at common times (for example according to a predetermined schedule or headway). By using common service locations and times, these modes can consolidate many trips that have similar origins and destinations or overlapping routes. However, the routes are not sensitive to changing travel patterns, and have no way of identifying which trips are going unserved, or are poorly served, by the existing routes. On the opposite end of the spectrum, personal modes of transportation, such as a private vehicle or taxi, offer service to and from the exact origin and destination of a rider, at close to exactly the time they desire to travel. Despite the apparent increased convenience to users, the presence of a large number of small vehicles results in a disorganized, and potentially congested road network during high demand periods. The focus of the research presented in this paper is to develop a system that possesses both the on-demand nature of a personal mode, with the efficiency of shared modes. In this system, users submit their request for travel, but are asked to make small compromises in their origin and destination location by walking to a nearby meeting point, as well as slightly modifying their time of travel, in order to accommodate other passengers. Because the origin and destination location of the request can be adjusted, this is a more general case of the Dial-a-Ride problem with time windows. The solution methodology uses a graph clustering algorithm coupled with a greedy insertion technique. A case study is presented using actual requests for taxi trips in Washington DC, and shows a significant decrease in the number of vehicles required to serve the demand.

Behavioral Response to Environmental Taxation: Evidence from the Transportation Sector

This dissertation analyzes how individuals respond to the introduction of taxation aimed to reduce vehicle pollution, greenhouse gases and traffic. The first chapter analyzes a vehicle registration tax based on emissions of carbon dioxide (CO2), a major greenhouse gas, adopted in the UK in 2001 and subject to major changes in the following years. I identify the impact of the policy on new vehicle registrations and carbon emissions, compared to alternative measures. Results show that consumers respond to the tax by purchasing cleaner cars, but a carbon tax generating the same revenue would further reduce carbon emissions. The second chapter looks at a pollution charge (polluting vehicles pay to enter the city) and a congestion charge (all vehicles pay) adopted in 2008 and 2011 in Milan, Italy, and how they affected the concentration of nitrogen dioxides (NOx). I use data from pollution monitoring stations to measure the change between areas adopting the tax and other areas. Results show that in the first quarter of their introduction, both policies decreased NOx concentration in a range of -8% and -5%, but the effect declines over time, especially in the case of the pollution charge. The third chapter examines a trial conducted in 2005 in the Seattle, WA, area, in which vehicle trips by 276 volunteer households were recorded with a GPS device installed in their vehicles. Households received a monetary endowment which they used to pay a toll for each mile traveled: the toll varied with the time of the day, the day of the week and the type of road used. Using information on driving behavior, I show that in the first week a $0.10 toll per mile reduces the number of miles driven by around 7%, but the effect lasts only few weeks at most. The effect is mainly driven by a reduction in highway miles during trips from work to home, and it is strongly influenced by past driving behavior, income, the size of the initial endowment and the number of children in the household.