Optimal Design, Evaluation, and Analysis of AGV Transportation Systems Based on Various Transportation Demands Satoshi Hoshino and Jun Ota Dept. of Precision Engineering, School of Engineering The University of Tokyo Bunkyo-ku, Tokyo 113-8656, JAPAN {hosino, ota}@prince.pe.u-tokyo.ac.jp Akiko Shinozaki and Hideki Hashimoto Mitsubishi Heavy Industries, LTD. Sagamihara-shi, Kanagawa 229-1193, Japan Abstract— In this paper, two designs for optimal Automated Guided Vehicle (AGV) transportation systems are presented. One is vertical and the other, horizontal. For these sys- tems, the hybrid design methodology proposed here is used. Therefore, we describe how to model and formulate these transportation systems and derive the design parameters, that is, the optimal design solutions. We next evaluate these two transportation systems based on the various transportation requirements, that is, the demands from a port authority. For this purpose, we compare the systems based on the total costs in constructing them. Finally, the evaluation and analytical results are provided, and the most convenient system is presented based on the validity of each system for the given demand. Index Terms— AGV transportation system, optimal and hybrid design methodology, evaluation, analysis. I. I NTRODUCTION The explosive growth in recent years in the volumes of freight has resulted in heavier workloads at seaports. Port authorities are being urged to implement advanced technologies to accommodate the increasing number of container ships. In this regard, the implementation of an Automated Guided Vehicle (AGV) transportation system would effectively automate a port container terminal. The advantages of implementing such a system are as follows: it would save the port terminal space; it would be more cost effective; and it would increase the efficiency. Literature is available to support these ideas [1] [2]. Although there are some conceivable types of AGV transportation systems, in this paper, we address two typical and prevalent AGV transportation systems for comparison, which are illustrated in Fig.1. We first design the systems optimally within the limited operation rules guiding mobile machines, that is, autonomous agents; then, the most convenient system is presented after evaluating the validity of the optimally designed systems. In the general design process, various design parameters for the various transportation demands need to be derived fast and optimally. Additionally, the agent’s behavior in the system needs to be analyzed. Moreover, the most convenient system should be identified on the basis of care- ful evaluation if there are some transportation systems as candidates. To meet these requirements, some challenging points are as follows: 1. How can a system be designed that will solve the combinatorial optimization problems inherent in a heterogeneous multi-agent transportation system fast and optimally? 2. How can a system be evaluated and analyzed to ascertain that it is valid and effective to meet the demands? Two methodologies are being proposed for the first challenge point above. One is based on a numerical model, and the other, on a simulation. On the other hand, to solve the second challenge point, the systems are compared and evaluated with the use of a deterministic design. With regard to a numerical design methodology, Abe et al. [3] [4] proposed the use of an open queuing network model. However, it is impossible to consider the gap between the numerical model and the behavior of an actual agent, for example, the existence of congestion in the system. On the other hand, Chiba et al. [5] proposed an integrated design methodology that incorporates a Genetic Algorithm (GA) in a simulation-based approach. However, several hours, that is, a significant amount of compu- tational time, are needed in the search for the optimal design parameters. Additionally, with each changed design parameter, it is impossible to analyze the global system behavior numerically by this methodology. To solve these problems, Hoshino et al. [6] have proposed a hybrid design methodology that uses a closed-cyclic queuing network model and a simulation-based optimization method. In this study, to overcome the first challenge point, we apply the proposed methodology for the optimal design of the system because of the following advantages: • It is possible to design the system optimally within a few minutes. • It is possible to consider and numerically analyze the gap and the global system behavior. In their research relating to the comparison and eval- uation of transportation systems, Chin et al. [7] have evaluated various transportation systems based on the con- struction cost. However, the optimality of the system is not considered because every design parameter is decided