Yard crane scheduling with job arrival time uncertainties

Abstract

Operation control in container terminals is always of great interest in the area of Computational Logistics and Operational Research. Yard Crane operations is an important step in the handling of containers and greatly impacts the efficiency of an entire terminal. During Yard Crane scheduling, the most important objective is to minimize the total tardiness of container jobs, in order to reduce the vessel turnaround time. In this project, we have addressed two problems. The first is to take uncertainties of job arrival times into consideration during Yard Crane scheduling. Uncertainties in job arrival times are always ignored in previous optimization algorithms. However, operations are highly unpredictable in container terminal operations. We adopted apparent tardiness cost-based dispatching rule with ready time and separable setup (ATCRSS) to solve single Yard Crane scheduling problem with job arrival time uncertainties. The second problem is to determine a more accurate model of dispatching rules’ performance. Such models will provide predictions of job completion times with better accuracy, which will support the planning of down-stream operations. To solve the first problem, we modeled the job arrival time using normal distributions and analyzed the impact of uncertainties in the original ATCRSS algorithm. Subsequently, we proposed ten methods of modifying the ATCRSS dispatching rule to handle job arrival uncertainties. Our experimental results showed that the impact of arrival time uncertainties was non-ignorable. However, the ten modified ATCRSS rules could not demonstrate statistically important improvements in the performance. Further studies are needed in the future. To solve the second problem, we compared vehicle waiting times, vehicle staying times and Yard Crane gantry times among ATCRSS, First-Come First-Serve (FCFS) and Deadline-First-Serve (DFS) dispatching rules. A suitable average inter-arrival time of container jobs was chosen to provide a stable environment. Exponential distributions of the performance indicators were hypothesized but were statistically rejected using the chi-square test. However, box plots and Q-Q plots have demonstrated good graphical models of summarizing the dispatching rules’ performance.