Robert B. Noland, PhD, Director, Voorhees Transportation Center
Taxi GPS data provides traffic insights
The Voorhees Transportation Center has created a research project to leverage valuable data from taxi GPS systems. The project goals are twofold: first, to use GPS data from taxis to understand actual taxi travel characteristics, such as the busiest origins and destination, and second to understand overall network conditions using average travel speeds, distances, and times obtained from the data. Research team members include Eric J. Gonzales, PhD; Ci Yang; Ender F. Morgul; and Kaan Ozbay, PhD.
Sample density of taxi pick-ups (top) and drop-offs (bottom) in New York City..
Congestion times, locations revealed
Taxi traffic accounts for almost 12% of the total street traffic in New York City. This means the GPS data from taxi trips is a valuable resource to understand citywide traffic conditions. Traditional traffic surveillance methods, such as loop detectors or closed circuit video cameras, are expensive to deploy and can measure only traffic flows, speeds, and densities at fixed locations in the network. However, GPS data from the linked origins and destinations of taxi trips provide detailed information about where traffic congestion occurs and at what times of day.
Travel demand patterns studied
The GPS data from New York taxis consist of nearly 40 billion data points per year, which provide sufficient detail to support analysis of specific spatial and temporal variations in taxi use and traffic conditions. All together, these data show patterns of travel demand. Maps of taxi trip origin and destination densities from a sample of data during the morning peak, for example, show the concentration of trips starting at major transit hubs and ending in Manhattan’s busiest business districts.
Travel times, traffic delays predicted
The abundance of data is also useful for comparing the characteristics of individual trips. Therefore, an additional aspect of this project is to analyze the reliability of taxi travel times for specific origin-destination pairs compared to other modes of travel, such as bus and rail transit. Using Google Maps API, a tool is in development to provide estimated transit travel times corresponding to actual observed taxi trips. Ultimately, the data set will provide insights for where and when traffic congestion has the greatest detrimental effect on travel times for taxis. This might also indicate delays for other road users, including buses.
The methods developed as part of this study can be used to study demand patterns in other cities with high taxi use. The methods for using GPS data to estimate traffic conditions and travel time variability on the road network may be applied to data from other types of vehicle fleets as well.
Understanding bus transit driver availability
Rutgers News –Transit agencies must employ a sufficient number of transit vehicle operators to meet their scheduled service requirements and to account for unforeseen absences, such as illness. To do so, the agencies employ extraboard operators (on-call backups) to account for these unexpected situations and ensure that service is not interrupted. Transit agencies have a burden to meet minimum service requirements for day-to-day operations. However overestimating the number of extraboard operators can also have a significant cost for these usually cash-strapped transit agencies. The Voorhees Transportation Center is investigating a way to address this issue. The research team includes Kaan Ozbay, Ender F. Morgul, Eric J. Gonzales, Hani Nassif, and Ozlem Cavus.
Goal: reliable bus service at minimum cost
In this study funded by the Mineta National Transit Research Consortium, we are investigating the transit agencies’ day-to-day extraboard management problem of adequately managing supply while maintaining an acceptable level of service.
Monthly number of absent transit drivers.
The main research objective is to create an extraboard management plan primarily for transit agencies – such as New Jersey Transit, New York Metropolitan Transportation Authority, and Rutgers University – to properly and efficiently allocate their personnel throughout their respective bus transit systems.
The goal is to help the transit agencies provide reliable bus service throughout the system, but to do so efficiently and at a minimum possible cost. The observed demand and supply data required for the model validation can be obtained from the transit agencies in the New York/New Jersey area that have already deployed advanced technologies such as smart cards, on-board GPS, and other emerging transit fare collection and asset tracking technologies.
Reliability and risk assessed
Stochastic mathematical models that can adequately capture the complex structure of the agencies’ decision making processes through observed data are employed to develop an extraboard management plan that takes into account various reliability and risk measures. Recently, the Rutgers research team has completed extensive work to develop similar probabilistic mathematical models required to develop an extraboard management plan for emergency evacuation conditions. These models are now being modified and applied to non-evacuation conditions where risk and reliability measures can be substantially different from those for emergency management operations.
Reduced costs, better management play key roles
The research team also plans to implement these models in a user-friendly computer tool that enables decision-makers to easily create various scenarios and to conduct “if-then” scenarios for quick decision making. The end product of this study will be leveraged to help develop several recommendations that transit agencies can easily adopt to reduce operating costs with more effective extraboard management strategies.