A dynamic data driven optimization system for yard crane management

• Main Author: Guo, Xi.
• Other Authors: Huang Shell Ying
• Format: Theses and Dissertations
• Language: English
• Published: 2012
• Subjects: DRNTU::Engineering::Computer science and engineering::Computing methodologies::Simulation and modeling; DRNTU::Engineering::Computer science and engineering::Computer applications::Physical sciences and engineering
• Online Access: http://hdl.handle.net/10356/48098
• Institution: Nanyang Technological University

Breakdown of trade barriers and the trend of globalization have greatly increased the importance of marine transportation systems and container terminals, which serve as hubs and are crucial links of the marine transportation network. Nowadays, the number of container transshipment is burgeoning. There are increased competition between seaports and higher demands on container terminal logistics and operation management systems. The objective of the study is to generate new optimization methods and techniques for managing yard crane (YC) operations based on predicted information. Potential of fusing and employing real time tracking data of moving assets could generate predicted vehicle arrival information at the yard side, which would support new optimization methods to improve YC operation efficiency. A hierarchical scheme for YC operation management is proposed. The hierarchical scheme aims to minimize the overall average vehicle job waiting time through flexible re-distributions of YCs to cope with dynamically changing job arrival patterns over time in the yard. The hierarchical scheme is organized into three levels. It handles the YC dispatching problem in Level 3 and handles the YC deployment problem in Level 2 and Level 1. A DDDOS (Dynamic Data Driven Optimization System) framework is presented for the understanding, controlling, synchronizing and optimizing of the YC hierarchical scheme. For the YC dispatching problem in Level 3 of the hierarchical scheme, a modified A* search algorithm with an admissible heuristic is proposed to find optimal solutions. To overcome the large memory usage limitation of the A* search, a RBA* algorithm is further proposed which combines the advantages of the A*search heuristic and a backtracking algorithm with prioritized search order. The experiments under noise show that the proposed algorithms perform well even when predictions of arrivals are not 100% accurate. The efficiency of the algorithms suggests that even when there is heavy noise disturbance like large unexpected change in job arrivals, the dispatching sequence may be re-generated in a computational efficient manner without delaying the YC operations in rolling planning windows.