Scheduling model in cross docking

Cross docking are widely adopted as an alternative to traditional warehousing in many industries. It consolidates different deliveries from suppliers into specified shipments catered for respective customers thus reducing transportation and inventory holding costs. Since cross docking minimizes the amount of product handling, perishable items such as dairy and food products are suitable for it. The sophisticated analysis of cross docking problems involves various methodologies such as Approximate and Exact algorithms. It was observed that the former performed much better for large-sized instances as compare to the latter which is meant for smaller instances. One notable methodology is Genetic Algorithm which is based on the principle of Darwin’s Natural Selection Theory to determine an optimal solution. The advantage of this method anchors on the ability to retain and explore the characteristics of good solution during iterations. Also, the algorithm can arrive at the optimal solution without searching the whole space within reasonable computational time. A cross docking network problem, based on Strategic Decision Level, was modeled to simulate the realistic cross docking model subjected to multiple dock doors, soft time window constraints and multiple products. The framework of this model was created with MATLAB using Genetic Algorithm. The key process of the algorithm includes initial population generation, mutation, crossover, elite selection and stopping. Three sets of experiments were conducted to study the robustness of the model and the performance of the important parameters. From the results, it was found out that the model produce relatively good solutions at initial population of 100, mutation probability of 20% and elite percentage of 50%. It was also concluded that as the number of deliveries, pickups, cross docks, time horizon and product types increase, the number of variables involved increases and so is the complexity. With much higher complexity, the computational time elapsed will increase tremendously. Future works and improvements can be done to reduce the stochastic nature of the Genetic Algorithm iteration process. Currently, the process has the capability to reduce 50% randomness of the iteration operation by constraints filters. Also, the model can be used to solve real life logistic problems using real data in future investigation against other methodologies.Bachelor of Engineering (Mechanical Engineering