New method improves how to assign requests to facilities
Researchers developed an algorithm to better manage capacity when requests arrive one by one.
The Online Transportation problem involves assigning requests to facilities as they arrive. Each request must be assigned to one of facilities immediately and irrevocably. The goal is to minimize the total distance between requests and their assigned facilities.
This problem is a generalization of the Online Metric Matching problem. In that version, every facility has a capacity of only one. A lower bound for all deterministic algorithms in this field is .
Kalyanasundaram and Pruhs (1998) conjectured that the ratio could also work for online transportation. Recently, Seth, Das, and Raghvendra (2026) found a ratio of approximately using a Robustified Greedy algorithm.
This new work provides an algorithm with a competitive ratio of , where . This improves the leading coefficient in the general-metric bound. The method builds on work by Harada and Itoh (2025) and Arndt, Moseley, Pruhs, and Uetz (2026).
The algorithm uses a virtual configuration to simulate car movements. In this simulation, cars may be evicted and continue searching for space. This process allows the algorithm to decide which real request goes to which garage without changing previous decisions.
The algorithm organizes its search using a spanning tree of the garage locations. It uses a phase to shift geometric search scales. This helps the system decide when to skip certain searches and optimize thresholds for including edges.
A competitive ratio compares the cost of an online algorithm to an optimal offline solution. A metric space is a set of locations with a distance function. An irrevocable assignment is a decision that cannot be changed once it is made.
Key facts
- The new algorithm achieves a competitive ratio of with .
- The method uses priority-based virtual depth-first searches over known facilities.
- The algorithm employs a phase to shift geometric search scales to optimize thresholds.