Chan, Yupo (University of Arkansas at Little Rock)

Yupo Chan
University of Arkansas at Little Rock
Little Rock, AR
USA 72204-1099

Phone: 501-569-8926
Fax: 501-569-8698
Email: yxchan@ualr.edu
URL: www.ualr.edu/yxchan

Topics

A Game-Theoretic Model for Secure International Communications
We developed a new defensive model for secure global voice communications. To insure the security, it uses an n-person, zero-sum, cooperative and non-cooperative game to optimize the revenue among service-provider coalitions. The cooperative game allows a coalition of network service providers (NSP) to be formed. It is based not only on their revenues from traffic, but also from incentive payments from a coalition leader. The non- cooperative game guards against adversarial tampering or attacks. In particular, we worry about these two revenues after an adversary attacks an NSP and renders it nonfunctional: coalition revenue and network provider revenue. We optimize these revenues by hardening NSPs and improving their respective revenues with federal incentive payments. A multi- criteria optimization problem was developed to establish the strategic competition between the coalition defender and attacker. Irrespective of the amount of incentive payments, an applicable hardening and tampering strategy can be obtained. It was shown that the option to harden NSPs has measurable value whether an incentive is provided to form the secure coalition. In addition, the adversary's tampering strategy is revealed in the shadow prices associated with the non-cooperative-game constraints. Intermediatet of Defense. Results are in part validated against Monte-Carlo simulation. (Intermediate)

A COMBINED INVENTORY AND DELIVERYMODEL FOR REPAIRABLE ITEMS
This paper considers a network composed of multiple depots that face uncertain demand for repairable items. It models the joint problem of determining how many units to repair and hold in inventory at each depot and how many to ship to other depots so as to minimize system-wide inventory, shortage and delivery costs in a single period. The fleet of vehicles are all stationed in the main depot and each depot has a certain holding and repair capacity. The formulation is a "multi-commodity" extension of Federgruen & Zipkin's combined vehicle-routing and inventory-allocation model (1984). The additional complexity is that each depot needs to decide how many units to repair and can send to any others-i.e., lateral re-supply. Being a nonlinear mixed integer program, the problem is solved using generalized Benders' decomposition. (Intermediate)

MULTI-CRITERIA VEHICLE-ROUTING IN A THREE-DIMENSIONAL NETWORK:
ElementaryIntermediaA special asymmetric vehicle-routing problem (VRP) is addressed here in three?dimensional space. One wishes to make several stops in an ingress path and return via a different egress path. Three criteria are considered in this ingress-egress vehicle-routing problem: distance, passive risk, and active risk. Routing is to be done in the shortest distance, and exposing the vehicle to the least risk. Risk is defined as passive if only exposure is involved. It is active if harm is inflicted because of the exposure. A multi-objective dynamic-programming (DP) algorithm is proposed to generate a set of non-dominated solutions without assuming a particular form for preference or value functions. Filtering is used finally to select from non-dominated solutions in a case study. This algorithm is compared with regular DP. The conventional DP procedure can be shown to generate paths in real time. When combined with sound judgment, it can yield non-dominated solution one at a time. It holds promise for day-to-day operational use. On the other hand, the multi-objective DP, when combined with filtering, has been shown to yield better quality solutions. Extensive computational experiences also point toward the preference of grid to hexagon tessellation in representing the three-dimensional space. (Intermediate)

SINGLE-COMMODITY MULTICRITERIA STOCHASTIC-NETWORKS: IMPROVING RELIABILITY VS. THROUGHPUT
There are three objectives in this research. First we measure the reliability of large single-commodity stochastic-networks. This is accomplished through an application of a factoring program developed by Page and Perry (1989). Second we develop a reliability-improvement model given that a practical reliability-expression does not exist. This is modelled by the Jain and Gopal (1990) heuristic through a linear improvement model. Finally, we examine the tradeoff between maximizing expected-flow and reliability. This is accomplished by generating bounds on the efficient frontier using an approximate multicriteria optimization approach. In this simplified approach, both expected flow and reliability can be measured practically and subsequent improvements made, providing insights into the operations of stochastic networks. Extensive computational experiences have been gained through experiments with three large-scale communication networks from the U S Department of Defense. Results are in part validated against Monte-Carlo simulation. Intermediate expected flow and reliability can be measured practically and subsequent improvements made, providing insights into the operations of stochastic networks. Extensive computational experiences have been gained through experiments with three large-scale communication networks from the U S Department of Defense. Results are in part validated against Monte-Carlo simulation. (Intermediate)

MEASURING THE DELAY OF PACKET-SWITCHING NETWORKS
The fast pace of telecommunications dictates the need to access current information with some level of confidence. Automated information systems must be able to anticipate the time required for processing under favorable and adverse conditions. This research presents the minimum and maximum delay encountered in a packet-switching network. The lower bound can be estimated at a given probability for the path with independent and identically distributed processors using the Central Limit Theorem. Similarly, the upper-bound estimating technique describes the maximum delay as a given probability. It is computed for the path with M/M/1 queues. These two techniques were applied to two typical test networks. They appropriately bounded the representative sojourn times. Without elaborate simulations, decision makers and network designers can use this technique to gain insight into a variety of communication and design concerns. (Intermediate)

Location, Transport and Land-Use
This talk identifies the underlying principles that govern siting, community development, and product/service delivery. Included are procedures to perform: site location, land-use planning, location-routing, competitive allocation of products & services, and spatial forecasting. It suggests solution techniques for emergency-response to natural and manmade hazards, environmental planning, infrastructure management, intelligent transportation systems, real-estate development, satellite remote-sensing, supply-chain management, and urban land-use plans. (Intermediate)

Incident Management on Highway Networks
In an Advanced Traveler Information System (ATIS), we study how to map a driver’s interest to real-time routing decisions. Accounting for en-route de-lays and alternate routing, ATIS networks exhibit non-FIFO behavior—drivers who depart earlier may not arrive ahead of those who depart later. Given a time-dependent network with full travel-time information, we model such dynamic routing decisions that include waiting en-route for an incident to clear. We employ a wait-time search algorithm to account for the best delays en-route. The algo-rithm elicits the bottlenecks in the network and obtains the optimal wait-times that would be favorable to a driver in achieving the fastest travel-time. We further model the driver’s risk-aversion behavior using a stochastic metric in the algo-rithm, by simultaneously considering risk and fastest travel time in path choice. Our defined routing policy takes decision at every node based on the current and future network states to determine the optimal wait-time and the next hop-node. Empirical results are obtained from a Central Arkansas Highway Network. The computational efficiency of the proposed algorithm has also been assessed. It is shown to be operationally acceptable for real-time applications. (Intermediate)

Background

• Ph.D. MIT (Operations Research)
• Master of Science MIT (Transportation systems/Economics)
• Bachelor of Science MIT (Civil Engineering)

Dr. Yupo Chan is Professor & Founding Chair of the Systems Engineering Department at the University of Arkansas at Little Rock. He received his BS degree in civil engineering, MS degree in transportation systems/economics and Ph.D. degree in operations research, all from the Massachusetts Institute of Technology. Dr. Chan has written numerous journal articles and books. The most recent ones include Location, Transport and Land-use: Modeling Spatial-Temporal Information, Springer-Verlag, Forthcoming, 750 pages (with Web-based software) and Location Theory and Decision Analysis, ITP/South-Western, 2000, 533 pages (with Computer Software disk).

His research interests include telecommunication systems, transportation systems, networks and combinatorial optimization, multi-criteria decision-making, spatial-temporal information, econometrics and technology assessment. He has previously held faculty positions at the Air Force Institute of Technology, Washington State University, State University of New York, Stony Brook and Pennsylvania State University. Dr. Chan is a Fellow of the American Society of Civil Engineers and has won numerous awards for his work, including the Harland Bartholomew Award of the American Society of Civil Engineers, the Koopman Prize of the Operations Research Society of America and a Congressional Fellowship with the Office of Technology Assessment.