Jaeho Kang

Deriving stacking strategies for export containers with uncertain weight information

In a container terminal, export containers are usually classified into one of a few weight groups and those belonging to the same group are stored together on a same stack. The reason for this stacking by weight groups is that it becomes easy to have heavier containers be loaded onto a ship before lighter ones, which is important for the balancing of the ship. However, since the weight information available at the time of container arrival is only an estimate, containers belonging to different weight groups are often stored together on a same stack. This becomes the cause of extra moves, or re-handlings, of containers at the time of loading to fetch out the heavier containers placed under the lighter ones. In this paper, we propose a method based on a simulated annealing search to derive a good stacking strategy for containers with uncertain weight information. Simulation experiments have shown that our strategies more effectively reduce the number of re-handlings than the traditional same-weight-group-stacking strategy. Also, additional experiments have shown that further improvement can be obtained if we increase the accuracy of the weight classification by applying machine learning.

Using Cluster-Based Sampling to Select Initial Training Set for Active Learning in Text Classification

We propose a method of selecting initial training examples for active learning so that it can reach high performance faster with fewer further queries. Our method divides the unlabeled examples into clusters of similar ones and then selects from each cluster the most representative example which is the one closest to the cluster’s centroid. These representative examples are labeled by the user and become the members of the initial training set. We also promote inclusion of what we call model examples in the initial training set. Although the model examples which are in fact the centroids of the clusters are not real examples, their contribution to enhancement of classification accuracy is significant because they represent a group of similar examples so well. Experiments with various text data sets have shown that the active learner starting from the initial training set selected by our method reaches higher accuracy faster than that starting from randomly generated initial training set.

Planning for intra-block remarshalling in a container terminal

Intra-block remarshalling in a container terminal refers to the task of rearranging export containers scattered around within a block into designated target bays of the same block. Since the containers must be loaded onto a ship following a predetermined order, the rearrangement should be done in a way that containers to be loaded earlier are placed on top of those to be loaded later to avoid re-handlings. To minimize time to complete a remarshalling task, re-handlings should be avoided during remarshalling. Moreover, when multiple yard cranes are used for remarshalling, interference between the cranes should be minimized. In this paper, we present a simulated annealing approach to the efficient finding of a good intra-block remarshalling plan, which is free from re-handlings at the time of loading as well as during remarshalling.

Generating a rehandling-free intra-block remarshaling plan for an automated container yard

Intra-block remarshaling in a container terminal refers to the task of rearranging the export containers, which are usually scattered around within a block, into designated target bays within the same block. Since the containers must be loaded onto a ship following a predetermined order, the rearrangement should be performed in such a way that the containers to be loaded first are placed on top of those to be loaded later in order to avoid rehandling. To minimize the time required to complete a remarshaling task, rehandling should also be avoided during the remarshaling operations. Moreover, when multiple stacking cranes are used for the remarshaling, the interference between cranes should be minimized. This paper presents a method to efficiently search for an intra-block remarshaling plan which is free from rehandling during both the loading operation and remarshaling, and which minimizes the interference between the stacking cranes.

Determination of storage locations for incoming containers of uncertain weight

In container terminals, heavier containers are loaded onto a ship before lighter ones to keep the ship balanced. To achieve efficient loading, terminal operators usually classify incoming export containers into a few weight groups and group containers belonging to the same weight group in the same stack. However, since the weight information available at the time of the container’s arrival is only an estimate, a stack often includes containers belonging to different weight groups. This mix of weight groups necessitates extra crane works or container re-handlings during the loading process. This paper employs a simulated annealing algorithm to derive a more effective stacking strategy to determine the storage locations of incoming containers of uncertain weight. It also presents a method of using machine learning to reduce occurrences of re-handling by increasing classification accuracy. Experimental results have shown that the proposed methods effectively reduce the number of re-handlings than the traditional same-weight-group-stacking (SWGS) strategy.

Optimization of container load sequencing by a hybrid of ant colony optimization and tabu search

Many algorithms that solve optimization problems are being developed and used. However, large and complex optimization problems still exist, and it is often difficult to obtain the desired results with one of these algorithms alone. This paper applies tabu search and ant colony optimization method to the container load sequencing problem. We also propose a hybrid algorithm, which can combine the merits of these two algorithms by running them alternately. Experiments have shown that the proposed hybrid algorithm is superior to both tabu search and ant colony optimization individually.

Applying Memory-Based Learning to Indexing of Reference Ships for Case-Based Conceptual Ship Design

This paper presents a method of applying a memory-based learning (MBL) technique to automatic building of an indexing scheme for accessing reference cases during the conceptual design phase of a new ship. The conceptual ship design process begins with selecting previously designed reference ships of the same type with similar sizes and speeds. These reference ships are used for deriving an initial design of a new ship, and then the initial design is kept modified and repaired until the design reaches a level of satisfactory quality. The selection of good reference ships is essential for deriving a good initial design, and the quality of the initial design affects the efficiency and quality of the whole conceptual design process. The selection of reference ships has so far been done by design experts relying on their experience and engineering knowledge of ship design and structural mechanics. We developed an MBL method that can build an effective indexing scheme for retrieving good reference cases from a case base of previous ship designs. Empirical results show that the indexing scheme generated by MBL outperforms those by other learning methods such as the decision tree learning.