When Hashing Met Matching: Efficient Spatio-Temporal Search for Ridesharing

TL;DR

This paper introduces a novel spatio-temporal search algorithm based on locality sensitive hashing to efficiently find near-optimal ridesharing matches in real-time urban datasets, outperforming existing heuristics.

Contribution

It models ridesharing matching as a near-neighbor search problem and develops a new algorithm based on MIPS theory, demonstrating improved efficiency and scalability.

Findings

Outperforms state-of-the-art heuristic methods in large datasets

Provides a scalable algorithm with theoretical time and space complexity

Proven practical applicability through experiments on NY taxi data

Abstract

Carpooling, or sharing a ride with other passengers, holds immense potential for urban transportation. Ridesharing platforms enable such sharing of rides using real-time data. Finding ride matches in real-time at urban scale is a difficult combinatorial optimization task and mostly heuristic approaches are applied. In this work, we mathematically model the problem as that of finding near-neighbors and devise a novel efficient spatio-temporal search algorithm based on the theory of locality sensitive hashing for Maximum Inner Product Search (MIPS). The proposed algorithm can find $k$ near-optimal potential matches for every ride from a pool of $n$ rides in time $O(n^{1 + \rho} (k + \log n) \log k)$ and space $O(n^{1 + \rho} \log k)$ for a small $\rho < 1$. Our algorithm can be extended in several useful and interesting ways increasing its practical appeal. Experiments with large NY yellow taxi trip datasets show that our algorithm consistently outperforms state-of-the-art heuristic methods thereby proving its practical applicability.