Back to search

Georgia Institute of Technology

EVALUATING THE SUSTAINABILITY IMPACTS OF INTELLIGENT CARPOOLING SYSTEMS FOR SOV COMMUTERS IN THE ATLANTA REGION

Abstract

dc:description.abstract

Community-based carpooling has more potential to help alleviate traffic congestion and reduce energy use during peak hours than ride-hailing services, such as Uber or Lyft, because community-based carpooling avoids deadheading operations. However, community-based carpooling is not fully exploited due to communication, demographic, and economic barriers. This thesis proposes a top-down computation framework to estimate the potential market-share of community-based carpooling, given the outputs of activity-based travel demand models. Given disaggregate records of commute trips, the framework tries to estimate a reasonable percentage/number of trips among commuters in single-occupancy vehicles (SOV) that can carpool together, considering spatiotemporal constraints of their trips. The framework consists of two major procedures: (1) trip clustering; and (2) trip optimization. The framework tackles the problems associated with using large amounts of data (for example, the Atlanta travel demand model predicts more than 19 million vehicle trips per day) by following “split-apply-combine” procedures. A number of tricks and technologies (e.g., pre-computing, databases, concurrency, etc.) are employed to make the mass computing tasks solvable in a personal laptop in a reasonable time. Two different methods are established to solve the carpooling optimization problem. One method is based on the bipartite algorithm, while the other uses integer linear programming. The linear programming method estimates both the systemic optimal performance in terms of saving the most vehicular travel mileage, while the bipartite-based algorithm estimates one Pareto optimal performance of such system that pairs the greatest number of carpool members (i.e., maximum number of travelers that can use the system) given acceptable (defined by the user) reroute cost and travel delays. The performance of these two methods are carefully compared. A set of experiments are run to evaluate the carpooling potentials among single-occupancy vehicles based on the output of activity-based model’s (ARC ABM) home-to-work single-occupancy vehicle (SOV) trips that can be paired together towards designated regional employment centers. The experiment showed that under strict assumptions, an upper bound of around 13.6% of such trips can be carpooled together. The distribution of these trips over space, time, and travel network are thoroughly discussed. The results are promising in terms of finding carpooling and decreasing total vehicle mileage. Moreover, the framework is flexible enough with the potential to act as a simulation testbed, to optimize vehicular operations, and to match potential carpool partners in real-time.

Degree

thesis:*
Level thesis:degree_level
Masters
Department dc:contributor.department
Civil and Environmental Engineering
Grantor dc:publisher
Georgia Institute of Technology

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Liu, Diyi
Advisor dc:contributor.advisor
  • Guensler, Randall L.
Committee members dc:contributor.committeemember
  • Rodgers, Michael O.
  • Liu, Haobing

Subjects

dc:subject × 4

Rights

Language dc:language.iso
en_US

Identifiers

dc:identifier.*
Handle dc:identifier.uri
http://hdl.handle.net/1853/66193

Chain of custody

source
Harvested from
Georgia Tech
Base URL
repository.gatech.edu/server/oai/request
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
citation

Liu, Diyi. EVALUATING THE SUSTAINABILITY IMPACTS OF INTELLIGENT CARPOOLING SYSTEMS FOR SOV COMMUTERS IN THE ATLANTA REGION. Masters thesis, Georgia Institute of Technology, http://hdl.handle.net/1853/66193