The Simpler The Better: An Indexing Approach for Shared-Route Planning Queries

Ridesharing services have gained global popularity as a convenient, economic, and sustainable transportation mode in recent years. One fundamental challenge in these services is planning the shared-routes (i.e., sequences of origins and destinations) among the passengers for the vehicles, such that the platform's total revenue is maximized. Though many methods can solve this problem, their effectiveness is still far from optimal on either empirical study (e.g., over 31% lower total revenue than our approach) or theoretical study (e.g., arbitrarily bad or impractical theoretical guarantee). In this paper, we study the shared-route planning queries in ridesharing services and focus on designing efficient algorithms with good approximation guarantees. Particularly, our idea is to iteratively search the most profitable route among the unassigned requests for each vehicle, which is simpler than the existing methods. Unexpectedly, we prove this simple method has an approximation ratio of 0.5 to the optimal result. Moreover, we also design an index called additive tree to improve the efficiency and apply randomization to improve the approximation guarantee. Finally, experimental results on two real datasets demonstrate that our additive-tree-based approach outperforms the state-of-the-art algorithms by obtaining up to 31.4%--127.4% higher total revenue.