在物流業中，配送問題可說是影響成本最重大的空間問題，近年來許多學者致力於車輛途程問題 (vehicle routing problem, VRP) 演算法的發展，以求改善物流配送的距離與時間成本，由於VRP問題是屬於非完全多項式 (NP-Complete) 的問題，故學者大多致力於啟發式演算法的開發及改良，而在VRP求解模式當中，一般大多使用二階段法，以往學者們多致力於第二階段求解的改善，對於第一階段起始解的產生，則較少以空間的角度切入研究。 本研究嘗試發展一具空間分析理論基礎的車輛配送途程的規劃模式，並針對不同的需求點空間分佈形態 (pattern)，與傳統演算方法進行配送成本之比較。實驗結果發現本研究所提出之群聚VRP演算法在需求點分佈較為群聚的條件下，可得到較傳統演算法為佳的結果，同時本研究也針需求點的空間分佈，歸納出兩種有利於群聚配送的型態。

In the last few years, many researches have studied the vehicle routing problem (VRP) in logistics studies, which is a NP-Complete problem that can only be solved heuristically. The most frequently used methodology is the “two-step method.” The first step is to generate several unimproved solutions, and the second step is to improve the results generated in the first step. Although many studies have focused on the algorithms of the second step, little is known about the influences of the result of the first step. The purpose of this research is to compare the dispatching costs calculated by the application of the VRP models with the costs obtained by traditional algorithms. The study shows that the VRP models yield better results when the demand points are clustered. We also show two clustered patterns of demand points under which the dispatching of goods is cost effective.