Raymond K. Cheung

A Two-Stage Stochastic Network Model and Solution Methods for the Dynamic Empty Container Allocation Problem

Containerized liner trades have been growing steadily since the globalization of world economies intensified in the early 1990s. However, these trades are typically imbalanced in terms of the numbers of inbound and outbound containers. As a result, the relocation of empty containers has become one of the major problems faced by liner operators. In this paper, we consider the dynamic empty container allocation problem where we need to reposition empty containers and to determine the number of leased con tainers needed to meet customers? demand over time. We formulate this problem as a two-stage stochastic network: in stage one, the parameters such as supplies, demands, and ship capacities for empty containers are deterministic; whereas in stage two, these parameters are random variables. We need to make decisions in stage one such that the total of the stage one cost and the expected stage two cost is minimized. By taking advantage of the network structure, we show how a stochastic quasi-gradient method and a stochastic hybrid approximation procedure can be applied to solve the problem. In addition, we propose some new variations of these methods that seem to work faster in practice. We conduct numerical tests to evaluate the value of the two-stage stochastic model over a rolling horizon environment and to investigate the behavior of the solution methods with different implementations.

The berth allocation problem: models and solution methods

In this paper, we consider the problem of allocating space at berth for vessels with the objective of minimizing total weighted flow time. Two mathematical formulations are considered where one is used to develop a tree search procedure while the other is used to develop a lower bound that can speed up the tree search procedure. Furthermore, a composite heuristic combining the tree search procedure and pair-wise exchange heuristic is proposed for large size problems. Finally, computational experiments are reported to evaluate the efficiency of the methods.

A multiprocessor task scheduling model for berth allocation: heuristic and worst-case analysis

We consider a scheduling problem in which the processors are arranged along a straight line, and each job requires simultaneous processing by multiple consecutive processors. We assume that the job sizes and processing times are agreeable. Our objective is to minimize the total weighted completion time of the jobs. This problem is motivated by the operation of berth allocation, which is to allocate vessels (jobs) to a berth with multiple quay cranes (processors), where a vessel may be processed by multiple consecutive cranes simultaneously. We develop a heuristic for the problem and perform worst-case analysis.

Interblock Crane Deployment in Container Terminals

We consider the problem of scheduling the movements of cranes in a container storage yard so as to minimize the total unfinished workload at the end of each time period. The problem is formulated as a mixed-integer linear program, and the computational complexity ofthe problem is analyzed. A Lagrangian decomposition solution procedure is described. A new solution approach, called the successive piecewise-linear approximation method, is also developed. Through computational experiments, we show that our proposed solution methods are both efficient and effective for large-sized problems.

Models and Algorithms for Distribution Problems with Uncertain Demands

We consider the problem of distributing goods from one or more plants through a set of warehouses in anticipation of forecasted customer demands. Two results are provided in this paper. First, we present a methodology for approximating stochastic distribution problems that are computationally tractable for problems of realistic size. Comparisons are made to standard deterministic formulations and shown to give superior results. Then, we compare logistics networks with varying degrees of redundancy represented by the number of warehouses which serve each customer. Overlapping service regions for warehouses provides additional flexibility to handle real-time demands. We quantify the expected savings that might result from such strategies.

Adaptive Labeling Algorithms for the Dynamic Assignment Problem

We consider the problem of dynamically routing a driver to cover a sequence of tasks (with no consolidation), using a complex set of driver attributes and operational rules. Our motivating application is dynamic routing and scheduling problems, which require fast response times, the ability to handle a wide range of operational concerns, and the ability to output multiple recommendations for a particular driver. A mathematical formulation is introduced that easily handles real-world operational complexities. Two new optimization-based heuristics are described, one giving faster performance and the second providing somewhat higher solution quality. Comparisons to optimal solutions are provided, which measure the quality of the solutions that our algorithms provide. Experimental tests show that our algorithms provide high quality solutions, and are fast enough to be run in real-time applications.

The transition from freight consolidation to logistics: the case of Hong Kong

Abstract This paper examines the challenges and opportunities confronting the traditional container consolidation industry in Hong Kong as it strives to modernise by adopting logistics functions. The study is based upon a series of detailed surveys of the industry undertaken during a period of adjustment. Improving value-added services is seen as the key to the survival of Hong Kong as a major transportation centre; this study indicates that making the improvements represents unique challenges to the firms actually involved. The barriers and potentials for these firms to transform themselves to third-party logistics service providers are analysed.

Shape - a Stochastic Hybrid Approximation Procedure for Two-Stage Stochastic Programs

We consider the problem of approximating the expected recourse function for two-stage stochastic programs. Our problem is motivated by applications that have special structure, such as an underlying network that allows reasonable approximations to the expected recourse function to be developed. In this paper, we show how these approximations can be improved by combining them with sample gradient information from the true recourse function. For the case of strictly convex nonlinear approximations, we prove convergence for this hybrid approximation. The method is attractive for practical reasons because it retains the structure of the approximation.

Dynamic Routing for Priority Shipments in LTL Service Networks

Currently, less-than-truckload (LTL) carriers route both the regular and the priority shipments through their service networks using some fixed route patterns, known as load plans. In this paper, we consider an alternative strategy for routing priority shipments in LTL networks. This strategy exploits the stochasticity and dynamism embedded in the routing process where the real-time information at the current terminal (such as the loading status of trailers and the arrival of shipments) and the expected travel times from its neighboring terminals to the destination are used to route shipments adaptively. We show that this strategy can be approximated by the problem of finding dynamic shortest paths over a network with random arc costs. We develop an efficient algorithm to compute the expected travel times of these paths. These times can then be used when we implement the proposed strategy. In the numerical testing with real data, we observe that the level of service for priority shipments can be improved significantly.

Multi-attribute label matching algorithms for vehicle routing problems with time windows and backhauls

We consider the class of vehicle routing problems with backhauls and time window constraints. Our motivating application is the land transportation of air-cargo freight forwarders, which requires fast solution times and the ability to handle various operational issues such as heterogeneous vehicles, multiple trips per vehicle, and penalty for early arrival at customer sites. We formulate the problem in the framework of label matching where the labels have multiple attributes representing the states of vehicles at customer locations or possible routes that vehicles may continue to cover. Two optimization-based heuristics are developed. Experimental tests on random problems and real data show that the methods can produce quality solutions quickly and are flexible in incorporating complex constraints.