Expected Cost Analysis of Online Facility Assignment on Regular Polygons

Abstract: This paper analyzes the online facility assignment problem in a geometric setting where facilities with unit capacity are positioned at the vertices of a regular $n$-gon. Customers arrive sequentially at uniformly random positions along the edges. They must be assigned immediately to the nearest available facility, with ties broken by coin toss. The sequential nature and unknown future arrivals require a probabilistic analysis of the expected assignment cost. Our main contribution is a recursive characterization of the expected cost: for any occupancy state $S$, the expected remaining cost $V(S)$ equals the average over all edge positions of the immediate assignment cost plus the expected future cost after assignment. We prove that this integral equation can calculate a solution and provide the expected value for small $n$ ($n = 3, 4, 5$). For larger values of $n$ and expected cost, we develop efficient numerical methods, including a discretized dynamic programming approach and Monte Carlo simulation. The work establishes a fundamental probabilistic approach for online assignment in polygonal environments.