Logo image
組合最佳化問題之平行混合搜尋法
Dissertation

組合最佳化問題之平行混合搜尋法

陳威志
Doctor of Philosophy (PHD), 國立清華大學, 工業工程與工程管理學系
2011

Abstract

組合最佳化問題 啟發式搜尋法 分支界線法 塔布搜尋法 模擬退火法 平行演算法 混合式演算法 供應鏈設計 combinatorial optimization problem heuristic search branch-and-bound tabu search simulated annealing parallel algorithm hybrid algorithm supply chain design
Obtaining optimal solutions of combinatorial optimization problems is computationally intractable. These problems are known as NP-hard and cannot be solved optimally within a reasonable amount of time. Satisfying with good solution obtained by heuristic search methods within an acceptable execution time is the efficient way in practice. Even using heuristic search, obtaining a good quality solution within a reasonable computing time for large scale and evaluating the quality of the obtained solutions have still difficulties. The hybridization and parallelism of heuristics search offer the possibility for enhancing the efficiency of the search. In first place, a case study of heuristic search techniques for solving a real-world combinatorial optimization problem is discussed in this study. Two straightforward and efficient approaches based on simulated annealing and tabu search are implemented to solve the integer nonlinear programming model which proposed in the case study. In second place, we aim the hybridization of different methods by parallel computing. We present a parallel hybrid heuristic search that combines branch-and-bound method and tabu search algorithm by cooperative multi-search scheme to integrate the benefits of exact methods and metaheuristics. These two algorithms perform searches in parallel and cooperate by asynchronously exchanging information. We use a master-slave model to reduce the complexity of communication and enhance the performance of data exchange. A branch-and-bound process is used as the master process to control the exchange of information and the termination of computation. Several tabu search processes are executed simultaneously as the slave processes, and are cooperative by asynchronously exchanging information of the best solutions found and new initial solutions with the master process of branch-and-bound. According to the computation experiments of solving traveling salesman problems, the proposed heterogeneous parallel search algorithm outperforms a conventional parallel branch-and-bound method and a conventional parallel tabu search. The results also show the proposed heterogeneous parallel search algorithm achieves linear accelerations when we use more processors to accelerate searching time.

Metrics

1 Record Views

Details

Logo image