A machine learning-based approach to time-dependent shortest path queries

• Main Author: Wei, Yumou
• Other Authors: Xiao Xiaokui
• Format: Final Year Project
• Language: English
• Published: 2017
• Subjects: DRNTU::Engineering::Computer science and engineering
• Online Access: http://hdl.handle.net/10356/70473
• Institution: Nanyang Technological University

Road traffic is known to be time-dependent. The travel time of a road varies at different times of the day. Many algorithms have been proposed for finding a shortest path in a time-dependent road network. In this project, I explored an alternative approach that leveraged on GPS trajectories collected from thousands of taxis. Each GPS trajectory was mapped to a set of real road segments. An abstract landmark graph was built to represent the city’s road network and a machine learning-based approach was proposed to estimate the travel time of each edge. The estimates made by this approach were compared against real-time estimates made by existing online mapping services to evaluate its accuracy. A modified Dijkstra’s algorithm was presented to calculate a shortest path in a time-dependent landmark graph, based on the travel time estimates.