Aiming to address the issue that current ridesharing solutions mostly focus on serving online passenger requests and overlook offline ones, leading to underutilization of taxi resources, this paper proposed a probability-based routing ridesharing optimization algorithm that relied on mining historical travel trajectory data. This algorithm calculated a regional probability transition matrix from historical spatiotemporal data of passenger requests and utilized this matrix to optimize the recommended ridesharing routes for taxis. It presented a solution that integrated historical data with the ridesharing problem, effectively increasing the taxi's passenger capacity. By ensuring the offline passenger acceptance rate and online passenger tolerance, the algorithm employed a relaxed-time metric to predict and evaluate the passenger tolerance time for the entire path in O(n) time, and used the Dijkstra algorithm to plan routes through detour areas, minimizing the detour distance for taxis. This paper validated the effectiveness of the algorithm by using the real-world DiDi GAIA dataset. Experimental results demonstrate that the proposed algorithm outperforms the baseline algorithm by 12% in terms of the number of service requests served.

[2023-08-11] Accepted Paper /doc/2023.06.0248/v1

[2024-02-05] Printed Article /doc/2023.06.0248/v2