| Summary: | 碩士 === 明志科技大學 === 工業管理研究所 === 97 === The Generalized Assignment Problem (GAP) combines limited resource capacity and multi-dimensional knapsack problems to construct a combinatorial optimization problem; and it has extensive application in the areas of resource scheduling, factory selection, vehicle routing problem, etc. Capacity utilization has been used as an important indicator for effective allocation of limited resources to ensure that the total cost of all possible combinations can be minimized. The GAP assumes that each job is assign to one and only one agent (machine), but each agent may simultaneously perform multiple jobs, and each agent has its own capacity limitations. The capacity and cost of each agent vary according to the type of job performed. Since the GAP is classified as NP-complete, the solving time needed for a GAP would expand exponentially as the problem expanding its scope. This study developed a hybrid genetic heuristic algorithm to solve the GAP within a reasonable number of iterations to find a near-optimal solution but with significant savings in computation time. To avoid blind random search within a vast solution space, the solution space is prescreened by applying the select smallest capacity method to obtain an initial feasible solutions. Most of the traditional genetic algorithm significantly changes the value of the fitness function of mating chromosomes when using the one-point crossover operator, this study developed an improved mating-type single-point rule that incorporates heuristic algorithm which would rapidly satisfy the fitness function. The developed algorithm is tested on different sizes of randomly generated problems, and the results show that the solution would achieve convergence within a reasonable number of generations.
|
|---|
