window.dataLayer = window.dataLayer || []; function gtag(){dataLayer.push(arguments);} gtag('js', new Date()); gtag('config', 'UA-129731328-1');

Container reshuffling is a major operational challenge in yard management. If the space allocation is poorly designed, the yard crane needs to move all of the containers on top of the target container to other stacks before it can retrieve the target container, and these extra moves will significantly reduce the yard crane throughput. Containers have different attributes, such as weight, size, brand, destination, and special requirements (e.g., reefers, dangerous goods, out-of-gauge cargo, tight connections or certain handling sequence). Containers with the similar attributes are classified as single container class.

To reduce reshuffling, the “homogeneity rule”, which stacks containers of the same class closely for easy retrieval, has been widely implemented in terminal operations. From a space allocation perspective, the consignment strategy, which places containers belonging to the same destination vessel close to one another, has been adopted to support yard planning in terminals such as those in Singapore.

Although many studies assume that reshuffling can be omitted for simplicity based on the above rules, the operators of one of the world’s largest container ports have commented that reshuffling still has a major impact on yard operations, even when the above rules are implemented. Intuitively, although the containers are grouped and stored together according to specific attributes, the uncertainty of vessel arrival times, the discharging and loading sequences, and the arrival and departure of vehicles will all contribute to reshuffling.

To study the impact, we developed a discrete event simulation model which simulates stacking and un-stacking operations.



For the stacking operation, the “homogeneity rule”, which closely stacks containers with the same property, is applied to simulate the actual stacking operation:

  • Among all available stacks containing the container of the same class (target container), choose the stack with the fewest containers of other classes that are on the top of the target container.
    • If there are containers of other classes stacking above the “same class container”, these containers will be reallocated to other stacks according to the same rule.
  • If the above condition is not met, find any empty stack.
  • If the above condition is not met, find the highest available stack.

During the unstacking process, the unstacking sequence is not random. Indeed, the unloading sequence is driven by a vessel stowage plan and quay crane loading sequence, which also follows the homogeneity rule. Therefore, the containers of the same class will usually be retrieved together. Thus, the following rules are applied to simulate the actual unstacking operation:

  • The yard crane unstacks the containers of the same class batch by batch. The unstacking sequence follows a sequence of container classes, which is randomly generated.

The yard crane always unstacks the container with the fewest obstructing containers (a higher stack is chosen to break the tie) and moves all obstructing containers to an available stack with the same container class as the obstructing container, or if unavailable, to the lowest stack.


The simulation result is as above. In general, each subfigure is drawn with dozens of simulation results by fixing the number of container classes, i.e., 350, 400, 450, and 500. The x-axis represents the number of ground slots, the y-axis represents the number of containers, and the z-axis symbolizes the number of unproductive moves. Each subfigure shows the same pattern of the number of unproductive moves negatively correlated to the space size and positively correlated to the number of containers. The relation is reasonable, as an increasing stacking height will incur more reshufflings. Observation of the data reveals that the number of unproductive moves is positively correlated to the number of container classes when fixing the space size and the number of containers.

This result also matches the intuitions we learned from the terminal operators: when the average stacking height is above 4, the reshuffling will get more and more. Due to the increasing reshuffling, the terminal with high operation efficiency requirement is not able to stack high.


The outcome of the reshuffling study can be applied in several research topics, for example, yard space allocation problem, operator scheduling, etc.