車輛途程問題(VRP)屬於NP-Hard問題,而此類型之問題在找尋最佳 解之情況下,必須花費十分冗長的運算時間;因此,為了要迅速得到與最 佳解差異不大的近似解,一般都會採用啟發法來求解。本研究將同時考慮 任務節點及任務網枝,並於車輛行走時有總時間之限制,而任務節點及任 務網枝有個別時窗之限制;如此之考慮條件,可使得本研究更加符合實際 車輛途程之應用。 本研究之啟發式解算,可分為兩個階段:初始解及尋優改善。初始解 以「最鄰近法」及「掃描法」求得,尋優改善以「塔布搜尋法」求得,如 此便可得到一個的近似解;在尋優改善時,並提出數種交換法(如:swap 、2-opt* exchange及insert),針對任務來進行交換改善之比較。最後 本研究之啟發解與「門檻值接受法」所求得之結果,並且以「幹枝界限法 」求得一個下限解做為基準相互比較,以瞭解本研究之實際成效。
The vehicle routing problem (VRP) is a kind of NP-Hard problem.It takes a lot of operational time to find optimal solutions of VRP. All known algorithms for solving these problems require a number of computations which grow exponentially with the size of the problem. Therefore most of the research into solving vehicle routing problems has been concerned with finding heuristics that provide good solutions to the VRP. In this paper we consider the Multi-Vehicle Routing Problem with Time Windows (VRPTW) which is a generalization of the VRP. In this paper, the heuristic method is composed of two phases: initial solution and improvement algorithm. The initial solution can be obtained by "nearest neighbor procedure" and "scan algorithm". The improvement algorithm is obtained by "Tabu research". Therefore, we can obtain an approximate solution. Besides, we propose many exchange methods (such as swap, 2-opt* exchange and insert) to deal with different missions and compare their difference. We compare with the results of "Threshold Accepting" and the heuristic solution of this paper, and use "Branch and Bound" to obtain a lower bound of the optimum value as a comparing basis. Therefore, we can realize the effects of this paper.