Space-time clustering-based method to optimize shareability in real-time ride-sharing.

Area of Science:

• Operations Research
• Artificial Intelligence
• Transportation Science

Background:

• Real-time ride-sharing services face significant computational challenges in optimizing trip matching, especially with high request volumes.
• Existing clustering methods for ride-sharing requests often overlook crucial factors, limiting the efficiency of identifying shareable trips.

Purpose of the Study:

• To develop an AI-based approach for optimizing real-time ride-sharing by enhancing solution quality and reducing computation time.
• To introduce a novel clustering method that effectively identifies and groups the most shareable ride-sharing requests.

Main Methods:

• A new shareability function was defined to consider all sharing states between trip pairs, improving upon traditional spatio-temporal clustering.
• An AI-based clustering technique was employed to group highly shareable trips, followed by a heuristic framework for intra-cluster matching.
• The method was validated using a large dataset of half-million ride-sharing requests in Lyon, France.

Main Results:

• The proposed method successfully provides high-quality solutions within a short computation time, even for large-scale ride-sharing problems.
• The novel clustering approach significantly improves the identification of shareable trips compared to previous methods.
• The algorithm demonstrated efficiency in optimizing ride-sharing operations in a real-world urban network.

Conclusions:

• The AI-driven clustering method offers an effective solution for the complex optimization problems in real-time ride-sharing.
• The developed shareability function and heuristic framework enhance the efficiency and quality of ride-sharing matching.
• The proposed methodology is adaptable for various mobility services, including car-sharing and bike-sharing.