[Paper Review] A greedy approach for increased vehicle utilization in ridesharing networks

Aqsa Ashraf Makhdomi, Iqra Altaf Gillani|arXiv (Cornell University)|Apr 2, 2023

Transportation and Mobility Innovations4 citations

TL;DR

This paper proposes a greedy, k-hop sliding window-based route recommendation system for ridesharing platforms that leverages predicted origin-destination pairs to maximize vehicle utilization and reduce fleet size. By modeling the route recommendation problem as a submodular optimization within a localized search window, the method achieves near-optimal performance (within 1.3× of optimal) and reduces vehicle count by efficiently pairing passengers with overlapping routes, demonstrating real-time feasibility and environmental benefits on NYC and DC datasets.

ABSTRACT

In recent years, ridesharing platforms have become a prominent mode of transportation for the residents of urban areas. As a fundamental problem, route recommendation for these platforms is vital for their sustenance. The works done in this direction have recommended routes with higher passenger demand. Despite the existing works, statistics have suggested that these services cause increased greenhouse emissions compared to private vehicles as they roam around in search of riders. This analysis provides finer details regarding the functionality of ridesharing systems and it reveals that in the face of their boom, they have not utilized the vehicle capacity efficiently. We propose to overcome the above limitations and recommend routes that will fetch multiple passengers simultaneously which will result in increased vehicle utilization and thereby decrease the effect of these systems on the environment. As route recommendation is NP-hard, we propose a k-hop-based sliding window approximation algorithm that reduces the search space from entire road network to a window. We further demonstrate that maximizing expected demand is submodular and greedy algorithms can be used to optimize our objective function within a window. We evaluate our proposed model on real-world datasets and experimental results demonstrate superior performance by our proposed model.

Motivation & Objective

• To address the underutilization of vehicle capacity in ridesharing platforms, which contributes to increased greenhouse gas emissions and inefficiency.
• To improve route recommendations by incorporating predicted origin and destination of passenger requests, rather than relying solely on demand density.
• To reduce the number of vehicles required to serve all passenger requests by optimizing for shared rides with overlapping routes.
• To develop a scalable, real-time solution that overcomes the NP-hard complexity of optimal route recommendation.

Proposed method

• Proposes a k-hop-based sliding window to reduce the search space from the entire road network to a localized region, improving computational efficiency.
• Models the route recommendation problem as maximizing expected passenger requests within the window, which is proven to be submodular, enabling use of greedy algorithms with approximation guarantees.
• Applies a backward greedy approach to estimate the minimum fleet size required by transforming the origin-destination graph into a vehicle count graph and solving a minimum path cover problem.
• Uses historical taxi datasets from New York City and Washington DC to simulate and validate the model under real-world demand patterns.
• Employs a dynamic window sliding mechanism that moves forward in the direction of expected passenger demand to maintain relevance and efficiency.
• Integrates both route recommendation and fleet size estimation into a unified system to support sustainable and cost-effective ridesharing operations.

Experimental results

Research questions

• RQ1 Can incorporating predicted origin and destination of passenger requests into route recommendations significantly improve vehicle utilization in ridesharing platforms?
• RQ2 Is the problem of recommending optimal routes based on origin-destination pairs NP-hard, and if so, can it be efficiently approximated?
• RQ3 How effective is a k-hop sliding window approach in reducing the search space while maintaining high-quality route recommendations?
• RQ4 To what extent can a greedy algorithm with submodular optimization improve vehicle utilization and reduce fleet size compared to existing baselines?
• RQ5 Can the proposed system scale to real-time deployment with low latency, and how does it perform across different time periods of the day?

Key findings

• The proposed model achieves an approximation ratio of 1.3× the optimal vehicle utilization on a 6×6 km test area, indicating strong performance close to the theoretical optimum.
• The model reduces fleet size significantly compared to existing baselines by efficiently pairing passengers with overlapping routes, leading to fewer vehicles on the road.
• The system responds to route queries in approximately 1.1 seconds, demonstrating real-time feasibility for deployment in production ridesharing platforms.
• Fleet size dynamically adjusts throughout the day, peaking during morning and evening rush hours (7–11 and 16–20) and dropping during late-night hours (2–7), aligning with real-world demand patterns.
• The backward greedy approach for fleet size estimation effectively identifies the minimum number of vehicles needed to serve all requests, with results validated against actual demand data from NYC and DC.
• The model performs best during periods of high demand dispersion, with approximation ratios close to 1.0 during off-peak hours, indicating robustness and scalability.