Novel Trip Agglomeration Methods for Efficient Extraction of Urban Mobility Patterns

Abstract

Mobility patterns in an urban area can be defined as the trip making behavior of an urban population. Traditionally, the origin-destination matrix representation of travel demand, where trip ends are agglomerated toward zone centroids that are decided a priori, has historically been used to identify trip making behavior. In this paper, different agglomeration methods are explored to extract the trip making behavior and their performances are analyzed. First, a variant of the zone-based agglomeration method is proposed, in which zones are optimally located rather than having their locations determined beforehand. Then a trip-based agglomeration method is proposed, where each trip is represented as an ordered pair of origin and destination in the form of a line segment and agglomeration of these line segments is performed. The proposed line-based agglomeration method serves a two-fold purpose, (a) the proposed trip-based agglomeration method helps in identifying the corridors carrying the majority of the flow in a single step, as opposed to trip-end based agglomeration methods where several post-processing steps may be required to identify the corridors, and (b) this method performs better than the existing trip-end based agglomeration methods in terms of the number of corridors that are required to cover the given trips. Efficient algorithms are also developed to solve the proposed trip-based agglomeration method, their performance on real-world trip datasets is tested and finally, the properties of the proposed algorithms are explored.