Optimized allocation of straddle carriers to reduce overall delays at multimodal container terminals

Abstract

Increased global container trade and strong competition are boosting the need for efficient container terminals. Using internal transport resources efficiently reduces transfer times and operating costs. This article addresses the allocation of straddle carriers at a tactical level at intermodal container terminals via optimization and simulation. Our objective is to reduce overall delays at the terminal—with special focus on delays of inland transport modes (rail, road and waterway). We introduce a notation to describe different service strategies and delay criteria and analyze the complexity of different service strategies. A generic mixed integer linear program—that can be easily adapted to different service strategies—models the allocation problem based on a network flow representation of the terminal. We conduct a case study for a container terminal at the Grand Port Maritime de Marseille to evaluate the proposed allocation: we adapt the optimization model to the specific terminal, implement a discrete event simulation model and conduct experiments on actual data. Results show that the allocation proposed by the optimization model reduces delays at the terminal and performs well in a stochastic environment.

Appendix: Estimation of truck arrivals

This section presents a mechanism to forecast the number of export containers arriving by truck on a daily basis. This mechanism was developed within the ESPRIT project whose objective was to improve operations within the terminal based on reliable forecasts on the the number of container movements. We developed and tested an adapted version of the forecast mechanism presented by Gambardella et al. (1996).

Historic data reports for each export container arriving per truck, when it arrived at the terminal, on which vessel it was loaded and the planned departure of the vessel. Based on this data, we determine the average percentage on the total number of containers per vessel that arrive each day before the planned departure of the vessel (Fig. 6). To compute the number of export containers for a given day, we use information on announced arrivals of vessels and estimates on their volumes. For each vessel arriving in the next 14 days, we multiply the percentage with the volume of the vessel. The forecast equals the sum over all vessels. To determine the number of trucks per hour, we determine the average distribution of the workload over a working day (Fig. 7) and multiply it with the forecasted volume for the entire day.

Fig. 6

Distribution of container arrivals with regard to the planned arrival of a vessel

Fig. 7

Distribution of arriving containers over the day

We evaluated the quality of the forecast mechanism for a terminal at the Grand Port Maritime de Marseille. A prototype computed the estimates for the next two to three days based on historic data and announced vessels. Figure 8 presents the daily gap between the forecast and the realized entries of export container via trucks. The tests started in March 2011. We used data from March and April to calibrate the model and started forecasting in May. For most of the forecasts (around 85 %) the gap is smaller than 20 %, around 55 % of the forecasts do not exceed the gap of 10 % required by the terminal. All gaps above 40 % are due to postponed arrivals of vessels or by perturbations within the terminal.

Fig. 8

Gap between the forecast and the realized entries of export container via trucks

The prototype may be improved by including more information, e.g., the day of the week, already arrived containers, time and not only day of a vessels arrival. The terminal operator can also use his specific knowledge on the situation at the terminal to adjust the forecast. The same mechanism may also be used to forecast daily volumes for import containers.