Dispatching Automated Guided Vehicles in a Container Terminal

Abstract

Automated guided vehicles (AGVs) are increasingly becoming the popular mode of container transport in seaport terminals. These unmanned vehicles are used to transfer containers between ships and storage locations on land. The efficiency of a container terminal is directly related to the amount of time each vessel spends in the port. Hence to maintain competitive advantage and increase the efficiency of the container terminal it is necessary to determine the appropriate number of AGVs to deploy, and formulate good dispatching strategies for these AGVs.Existing techniques available in the literature use a variety of heuristic methods for dispatching the automated guided vehicles. These methods have been primarily developed to work in a manufacturing context where the network structure is simple and only a small number of such vehicles are required. The situation in seaport terminals however entails a greater network complexity and also a large fleet of 80 or more automated guided vehicles. There is little experimental evidence on how these techniques perform in a container terminal environment.Furthermore, for the handful of papers on AGV dispatching in a terminal environment, to the best of our knowledge, none of the proposed methods have explicitly considered the effects of congestion to determine the appropriate number of AGVs to deploy.In this paper, we address this gap in existing research. We present a network flow formulation for the dispatching problem of the automated guided vehicles. This formulation can be efficiently solved to obtain deployment strategies for large network sizes. Furthermore, our objective is formulated in a way to minimize the waiting time of the AGVs at the berth side, reducing the possibility that the AGVs will be clustered near there. Simulation results verify that the network flow-based technique significantly increases the throughput of a container terminal. As a by-product, the simulation model can also be used to determine the appropriate number of AGVs to deploy.

Model Predictive Scheduling for Container Terminals

“…Another interesting method can be found in [2]. The authors proposed a network-flow model to solve the vehicle assignment problem.…”

“…The time constants will be loosely based on the test cases mentioned in [4], [20] and [2].τ c , the time required by the cranes to handle each container, will be 75 seconds for all cycles taking place at a QC, and 100 seconds for all cycles taking place at a stacking crane (SC).τ ∆ , the time required to transfer a container between a crane and a vehicle, will be 30 seconds for all cycles. In practice most vehicles will have a higher maximal speed than 3 m/s.…”

“…This algorithm, called a greedy algorithm, can be found in literature in e.g. [2,4]. The greedy algorithm will be used as a benchmark for the performance of the MILP solution found earlier.…”

“…The time at which a crane finishes retrieving/storing a container f 1 Finish time of crane processing before the vehicle arrives f 2 Finish time of crane processing after the vehicle arrives G A graph representing a square max-plus matrix I Max-plus identity matrix J Cost function of the MPS problem J throughput Terminal throughput in containers per second k Event counter of an (S)MPL system k load (k) The cycle in which the container which is to be unloaded in cycle k was loaded onto the vehicle k preceding (k) The preceding cycle at the crane for cycle k k load (t)…”

“…The competitiveness of container seaports is mainly determined by the time in port for ships (transshipment time), and the rates for loading and discharging [1]. The importance of the transshipment times can be illustrated by the operating cost of container vessels, which, according to industry estimates, can easily exceed US$ 30,000 per day for large size container vessels [2]. Therefore this thesis will focus on reducing transshipment times.…”