In designing container terminals, the terminal management has to consider the choice for interrelated AGV and ASC dispatching rules. In this paper, we therefore examine the joint decision problem of dispatching containers to AGVs and selecting ASCs. Savings of approximately 20 per cent in the number of AGVs are achievable, if we apply the nearest-AGVfirst rule in combination with the cyclic ASC rule. Twin-load AGVs can also be used to obtain significant savings in unloading times and number of AGVs required.
The most critical planning and control problems arise at the deepsea side to ensure short berth times of the ships. In this paper, we study the interrelated planning and control problems of dispatching unloaded containers to automated guided vehicles (AGVs) and the dispatching of these containers to automated stacking cranes (ASCs) operating in the stack. We perform a simulation study and use several performance criteria such as the unloading times of the ship, the number of automated guided vehicles required and the utilization of these vehicles to examine dispatching rules. The input data used in this simulation study are obtained from interviews with logistics managers of Europe Combined Terminals at the port of Rotterdam and Vis and Harika .
We consider an automated container terminal in which the quay cranes (QCs) and ASCs are connected by a multiple lane loop layout. This loop is a fixed sequence of pick-up and delivery points at QCs and ASCs. AGVs travel over these fixed guide paths to handle all transportation requests. The moment an QC starts unloading a container, an AGV is selected from the parking area, according to a certain dispatching rule.
In this study, we will compare the performance of the following QC-initiated dispatching rules (i.e. an AGV is selected by an QC from a set of idle AGVs to transport a container – see ):
• Nearest-vehicle-first: A free AGV at the smallest distance is dispatched to the QC needing an AGV.
• Farthest-vehicle-first: A free AGV at the farthest distance is dispatched to the new QC.
• Random vehicle rule: The new pick-up container is randomly dispatched to any available AGV regardless the location of the AGV and the QC.
• Cyclic rule: Select the first available AGV beginning with the successor of the last AGV selected (to balance the workloadamong all AGVs in the system).
• Preferred order rule: Select the available AGV with the lowest unit number to transport the container. The assigned AGV travels to the QC and waits for the container.
The QC positions the container on the AGV. It might also occur that the crane needs to wait for an AGV to arrive. In that case, the QC needs to wait before it can unload a new container. After receiving the container, the AGV starts transporting the container to the stack. A variable number of AGVs is available to transport containers from the ship to the stack.
The number of vehicles required to minimize the unloading times of the ship will be used as a performance measure in the experiments. The AGV with its container will already be dispatched to an ASC with empty storage locations. In this way, we avoid the possibility that the AGV might have to travel to an ASC with no free space for a container. In this paper, we will study the performance of the following ASC selection rules:
• Cyclic rule: Select the first available ASC with empty space beginning with the successor of the last ASC selected (to balance the workload among all ASCs in the system.)
• Random rule: Randomly select an ASC with empty space.
• Farthest/nearest ASC first: Select the ASC (with empty space) at the farthest/smallest distance and with no more than v AGVs on their way to this ASC. The optimal value of v will be determined in the various experiments.
After arrival at the stack, the container needs to be stored in the stack. If the ASC is still handling another container, the AGV needs to wait. The ASC lifts the container off the AGV and the empty AGV travels back to the parking area at the QCs. After arrival of the AGV at the pick-up and delivery point of the assigned block, the ASC lifts the container and stores it in the stack.
First, we select one of the rows in the block according to a uniform distribution, in which each row has an equal probability to be chosen. Thereafter, a storage location is randomly selected from the empty locations in the selected row. Figure 1 summarizes the conceptual model. All processes have been implemented in Arena Rockwell Software.
We will compare the dispatching rules with the following performance measures:
• Total cycle time required to unload all containers off the ship and store them in the stack
• Minimum number of AGVs required to achieve a minimal total cycle time
• Average utilization of the AGVs
We generate 100 replications of each experiment to obtain a high level of accuracy (relative error smaller than two per cent with a probability of 95 per cent) in the results.
In the simulation study, we assume that 2,500 containers need to be unloaded off the ship by four QCs. Due to space restrictions at both sides of the ship, containers are not equally distributed over the ship. Usually, 15 per cent of the containers are designated for the leftmost QC, 35 per cent for each of the two QCs in the middle and 15 per cent for the rightmost QC.