Songshan Guo

An investigation of IDA* algorithms for the container relocation problem

The container relocation problem, where containers that are stored in bays are retrieved in a fixed sequence, is a crucial port operation. Existing approaches using branch and bound algorithms are only able to optimally solve small cases in a practical time frame. In this paper, we investigate iterative deepening A* algorithms (rather than branch and bound) using new lower bound measures and heuristics, and show that this approach is able to solve much larger instances of the problem in a time frame that is suitable for practical application.

Two natural heuristics for 3D packing with practical loading constraints

In this paper, we describe two heuristics for the Single Vehicle Loading Problem (SVLP), which can handle practical constraints that are frequently encountered in the freight transportation industry, such as the servicing order of clients; item fragility; and the stability of the goods. The two heuristics, Deepest-Bottom-Left-Fill and Maximum Touching Area, are 3D extensions of natural heuristics that have previously only been applied to 2D packing problems. We employ these heuristics as part of a two-phase tabu search algorithm for the Three-Dimensional Loading Capacitated Vehicle Routing Problem (3L-CVRP), where the task is to serve all customers using a homogeneous fleet of vehicles at minimum traveling cost. The resultant algorithm produces mostly superior solutions to existing approaches, and appears to scale better with problem size.

A New Hybrid Genetic Algorithm for the Robust Graph Coloring Problem

The RGCP (Robust Graph Coloring problem) is a new variant of the traditional graph coloring problem. It has numerous practical applications in real world like timetabling and crew scheduling. The traditional graph coloring problem focuses on minimizing the number of colors used in the graph. RGCP focuses on the robustness of the coloring so that the coloring is able to handle uncertainty that often occurs in the real world. By that, we mean that given a fixed number of colors we would like to color the graph so that adjacent vertices are assigned different colors with the consideration of the possible appearance of the missing edges. In this paper, we present a new hybrid genetic algorithm (GA), which embeds two kinds of local search algorithms – enumerative search and random search within the GA framework. In addition, we integrate a partition based encoding scheme with a specialized crossover operator into our GA method. We also propose an adaptive scheme that alternates between two local search methods to increase performance. Experimental results show that our new algorithm outperforms the best published results in terms of the quality of solutions and the computing time needed.

A tabu search algorithm for the safe transportation of hazardous materials

In this work, we study the problem of the safe transportation of hazardous materials, which is an important operational problem and has been studied extensively in the literature. We outline previous work and propose a new model that is a variant of the vehicle routing problem with time windows (VRPTW). The objective is to find a schedule to guarantee the safety of all vehicles. We propose a tabu search (TS) heuristic with a dynamic penalty mechanism to obtain good solutions. A realistic data generation mechanism is also presented and the elaborate computational results show the strengths of our algorithms.

Improved GRASP with Tabu search for vehicle routing with both time window and limited number of vehicles

One of the difficulties of using Artificial Neural Networks (ANNs) to estimate atmospheric temperature is the large number of potential input variables available. In this study, four different feature extraction methods were used to reduce the input vector to train four networks to estimate temperature at different atmospheric levels. The four techniques used were: genetic algorithms (GA), coefficient of determination (CoD), mutual information (MI) and simple neural analysis (SNA). The results demonstrate that of the four methods used for this data set, mutual information and simple neural analysis can generate networks that have a smaller input parameter set, while still maintaining a high degree of accuracy.

Space defragmentation heuristic for 2D and 3D bin packing problems

One of main difficulties of multi-dimensional packing problems is the fragmentation of free space into several unusable small parts after a few items are packed. This study proposes a defragmentation technique to combine the fragmented space into a continuous usable space, which potentially allows the packing of additional items. We illustrate the effectiveness of this technique on the two-and three-dimensional Bin Packing Problems. In conjunction with a bin shuffling strategy for incremental improvement, our resultant algorithm outperforms all leading meta-heuristic approaches.

A p-Robust Capacitated Network Design Model with Facility Disruptions

This paper studies a strategic supply chain management problem of designing robust networks which perform well under both normal condition and disruptions. A mix-integer programming model which incorporates p-robust measure is presented. The objective is to minimize the total nominal cost, while setting upperbounds on relative regrets in disruption scenarios. A GA-based hybrid metaheuristic algorithm is proposed and tested. Computational results demonstrate that system robustness can be substantially improved with little increase in cost. Our solution is also less conservative compared with common robustness measures.

Branch and bound algorithm for a single vehicle routing problem with toll-by-weight scheme

Most of previous studies on vehicle routing problems assume that traversal cost of each edge is simply equivalent to a constant number. Unfortunately, the models of this kind can not be applied in China because toll per kilometer of Chinese expressways varies with vehicles weight. Motivated by rapidly increasing market of expressway transportation in China, we address a new and special vehicle routing problem that takes a single vehicle and its weight into account. To solve this problem practically, we provide a branch-and-bound algorithm with a welldesigned lower bound. This algorithm can deal with any toll scheme in which toll per unit distance monotonically increases with weight. Computational results show that test instances with up to 42 vertices can be solved in reasonable computing time.

A new neighborhood based on improvement graph for robust graph coloring problem

In this paper, we propose a new neighborhood structure based on the improvement graph for solving the Robust Graph Coloring Problem, an interesting extension of classical graph coloring Different from the traditional neighborhood where the color of only one vertex is modified, the new neighborhood involves several vertices In addition, the questions of how to select the modified vertices and how to modify them are modelled by an improvement graph and solved by a Dynamic Programming method The experimental results clearly show that our new improvement graph based k-exchange cycle neighborhood improves the accuracy significantly, especially for large scale heuristic search.

A Branch-and-Bound Algorithm for the Talent Scheduling Problem

The talent scheduling problem is a simplified version of the real-world film shooting problem, which aims to determine a shooting sequence so as to minimize the total cost of the actors involved. We devise a branch-and-bound algorithm to solve the problem. A novel lower bound function is employed to help eliminate the non-promising search nodes. Extensive experiments over the benchmark instances suggest that our branch-and-bound algorithm performs better than the currently best exact algorithm for the talent scheduling problem.