This paper introduces a multi-stage optimization model that identifies an optimal evacuation plan in response to spatial-temporal weather-events with track uncertainty. We establish how a multi-stage model can effectively take into account track uncertainty to generate a better evacuation plan. In particular, we study evacuation management for hazardous convective weather events (e.g., tornadoes) that cause hundreds of fatalities in the US every year. The common practice against a tornado threat is shelter-in-place. In our computational results, we quantify the benefits and costs of evacuation as a response policy in comparison to the status quo policy, i.e., shelter-in-place.

Consider a private network of geographically dispersed computers with fast and high capacity connections, and an Internet application session, such as a massive multiplayer online game, with a server and a set of clients. We refer to the former as a service overlay network (SON), and assume that it could be connected to the Internet. The problem is to decide how to configure and utilize the SON in support of this application, such that the clients׳ speed of communication with the server is within given communication performance requirements. We provide an Integer Programming formulation of this problem, and prove that it is NP-Hard. In an attempt to solve the problem within strict computational time requirements of actual applications, we develop a solution framework based on partitioning and enumerating the solution space into smaller subproblems, one or more of which contains an optimal solution. In this framework, we develop and test an optimal seeking exact, and a fast polynomial time heuristic algorithm with success. The exact algorithm sets optimally solvable sizes of the subject problem, whereas the heuristic algorithm sets the size of solvable instances in a real application.

Kernels are executable code segments and kernel fusion is a technique for combing the segments in a coherent manner to improve execution time. For the first time, we have developed a technique to fuse image processing kernels to be executed on GPGPUs for improving execution time and total throughput (amount of data processed in unit time). We have applied our techniques for feature tracking on video images captured by a high speed digital video camera where the number of frames captured varies between 600-1000 frames per second. Image processing kernels are composed of multiple simple kernels, which executes on the input image in a given sequence. A set of kernels that can be fused together forms a partition (or fused kernel). Given a set of Kernels and the data dependencies between them, it is difficult to determine the partitions of kernels such that the total performance is maximized (execution time and throughput). We have developed and implemented an optimization model to find such a partition. We also developed an algorithm to fuse multiple kernels based on their data dependencies. Additionally, to further improve performance on GPGPU systems, we have provided methods to distribute data and threads to processors. Our model was able to reduce data traffic, which resulted better performance.The performance (both execution time and throughput) of the proposed method for kernel fusing and its subsequent execution is shown to be 2 to 3 times higher than executing kernels in sequence. We have demonstrated our technique for facial feature tracking with applications to Neuroscience.

Broadcast television viewing over the Internet (IPTV) is becoming commonplace. Multicasting trees serve as an efficient mechanism to deliver streaming data as each internal node duplicates the packets it receives and sends it along to its children which eventually delivers them to the clients. Given a set of multicasting trees whose roots are servers capable of broadcasting a set of distinct channels, and a set of clients (which are not part of the multicasting trees) each with a set of requested channels, our goal is to determine for each client for each of its channel request, a node (contact node) in the appropriate multicast tree (that serves the channel). The contact nodes are determined in such a way that certain optimization constraints are taken into consideration and satisfied. We have provided Integer Programming (IP) models and heuristics to find these contact nodes in order to optimize constraints on zap time and bandwidth utilization. The proposed IP model is novel and the polynomial-time heuristic provide a fairly good solution in a short amount of time.

Multicasting is an efficient way to deliver multimedia content (streaming, for instance) to different locations in the network. While end-to-end real-time constraints are important for interactive applications, sustained availability of bandwidth is more important to the destinations for multimedia streaming. In this research, we address the problem of multi-stream multi-source multicast routing problem (MMMRP) where each data stream could have multiple sources that will serve it and each source can serve multiple data streams in a sustained manner. The goal of MMMRP is to construct a routing forest for each of the data streams and the destinations while maximizing the residual bandwidth. The residual bandwidth is the available bandwidth after all destinations have been served with their desired streams. Our problem is shown to be NP-hard and we provide an Integer Programming formulation together with an efficient heuristic algorithm (MMForests) based on widest-path algorithm. Our empirical evaluations show that our algorithm MMForests can construct the multicast routing trees both quickly and keeping the residual bandwidth close to the optimal.

Digital television systems have a clear disadvantage relative to analog systems in users' quality of experience, most notably in the time required to change channels, or zap time. The goal of this research is to improve the performance of a multicasting IPTV network, both in user experience and in resource consumption. We formulate the problem of assigning IPTV clients to servers as an integer programming model, in variants which minimize channel-change times, overall network capacity consumption, or both. This problem is shown to be computationally hard, and the performance of the models is tested on problems of different sizes. Polynomial-time heuristics are presented which address a relaxed version of the problem, and the performance of these heuristics is measured.

Recent advances in multimedia software and hardware technologies and the availability of high-speed Internet service have been instrumental for growth in the online gaming industry. Multiple servers distributed across the network are commonly used to provide the desired quality-of-service (QoS) for the network game in order to achieve a higher quality-of-experience (QoE) to the players (clients). Each player in this distributed multi-player gaming environment connects to a particular server and it distributes each of the actions to all other players through the servers they are connected to. We imagine the server network to be an overlay network, wherein the latency on a link between two servers is the latency of the Internet path connecting them. We assume that we are given an overlay network of servers with link latencies and a set of players each with a different latency to each of the servers. Now our goal is to develop algorithms that perform the following actions in such a way that delay related QoS constraints are satisfied: (a) choose a subnetwork of the server network (server network selection) and (b) assign each player to a server in the subnetwork (client-assignment). More specifically, the QoS constraints that we address in this paper are a bound on the maximum delay in propagating a player's move to all other players (delay bound) and a bound on the maximum difference in the arrival times of a player's move at all other players (delay-variation bound). We have provided polynomial-time heuristics to determine a minimal cardinality server network and the corresponding client-assignment that satisfy both delay bound and that minimize delay-variation, if such a solution exists. We have considered cases in which the server network follows two communication models: client-server (CS) and peer-to-peer (P2P). Our extensive empirical studies indicate that our heuristic uses significantly less run-time in achieving the tightest delay variation for a given end-- - to-end delay bound while choosing a minimal number of servers.

Improving latency is the key to a successful online game-playing experience. With the use of multiple servers along with a well-provisioned network it is possible to reduce the latency. Given a network of servers, game clients, and a desired delay bound, we have designed algorithms to determine the subnetwork of servers whose cardinality is minimal. We have considered the cases wherein the subnetwork architecture is a client-server and a peer-to-peer. We have also provided exhaustive empirical evaluations of our algorithms and compared their performance with the optimum. Experimental results show that our polynomial-time algorithms could find good solutions quickly.

The paratransit vehicle scheduling problem involves scheduling a fleet of specially equipped vehicles for serving transportation needs of disabled and elderly people. It is also referred to as a dial-a-ride problem, and it can be classified as a multi-depot pickup and delivery problem with time windows and side constraints. The column generation approach constitutes one of the effective methodologies to solve this problem. However, in order to cope with the complexity of the problem, it is commonly applied in two consecutive stages: clustering and routing. First, clusters of customers are formed who will receive service together; next, vehicle routing decisions are made subject to the clustering decisions. This paper develops a nested column generation method, which integrates clustering and routing decisions, thus extending the applicability of the column generation approach in the context of the subject problem. The proposed method is implemented and applied to solve a problem faced by the transit authority of a mid size US city. A detailed account of implementation experience is provided. The computational results, based on actual data, indicate that problem instances with up to 680 requests and 48 vehicles can be solved within 2% of optimality under mild assumptions, and a 12% performance improvement over a well-established manual planning system can be achieved.

An algebraic modelling language (AML) is a domain-specific computer programming language for describing and solving mathematical programming models. We propose extending AMLs so that solution algorithms that are based on iteratively manipulating, modifying and solving a model are supported at a high abstraction level. We specifically focus on stochastic programming models with random parameters formulated as discrete scenarios and mathematical decomposition algorithms, which are commonly applied to solve such models. We identify the necessary language constructs and develop a design based on the open-source modelling software APLEpy. The proposed design, although specifically addressing decomposition algorithms, proves useful for implementing heuristic solution algorithms as well. The object-oriented nature of the design enables the algorithms that are coded with the proposed extensions to work with any other model that satisfies the assumptions of the initial model. This flexible and robust design helps inexperienced modellers to easily apply an advanced solution algorithm, and experienced modellers to build sophisticated algorithms quickly within the same development environment that is used to describe the model under consideration.

Textile manufacturing consists of yarn production, fabric formation, and finishing and dyeing stages. The subject of this paper is the yarn production planning problem, although the approach is directly applicable to the fabric production planning problem due to similarities in the respective models. Our experience at an international textile manufacturer indicates that demand uncertainty is a major challenge in developing yarn production plans. We develop a stochastic programming model that explicitly includes uncertainty in the form of discrete demand scenarios. This results in a large-scale mixed integer model that is difficult to solve with off-the-shelf commercial solvers. We develop a two-step preprocessing algorithm that improves the linear programming relaxation of the model and reduces its size, consequently improving the computational requirements. We illustrate the benefits of a stochastic programming approach over a deterministic model and share our initial application experience.

Operations research is an interdisciplinary science that develops scientific methods for modeling and solving decision-making problems. Arguably, the two most established and commonly practiced OR methodologies are mathematical programming (also called optimization) and discrete-event system simulation (called simulation). Mathematical programming describes a decision problem as decision variables and as mathematical functions of variables that capture the objective and problem constraints. For dynamic systems that are too complex to allow exact, closed-form solutions in continuous time, simulation can identify and describe events of interest and model their occurrences at discrete points in time. Python can glue together independent software environments to improve the software development cycle for sophisticated operations-research applications, which integrate different modeling methodologies.

We describe a transportation problem we studied at the trucking division of a textile manufacturer that owns and operates a large number of manufacturing facilities in all stages of the manufacturing supply chain. The problem involves scheduling of pickup and delivery of daily inventory movement between plants. We develop integer programming models that are based on the current workflow of the schedulers and that capture and optimize crucial aspects of the problem. We validate the models and demonstrate their use with actual data. The results suggest that an effective computerized decision-support system can be built around these models.

We study incentive issues that arise in semiconductor capacity planning and allocation. Motivated by our experience at a major U. S. semiconductor manufacturer, we model the capacity‐allocation problem in a game‐theoretic setting as follows: each product manager (PM) is responsible for a certain product line, while privately owning demand information through regular interaction with the customers. Capacity‐allocation is carried out by the corporate headquarters (HQ), which allocates manufacturing capacity to product lines based on demand information reported by the PMs. We show that PMs have an incentive to manipulate demand information to increase their expected allocation, and that a carefully designed coordination mechanism is essential for HQ to implement the optimal allocation. To this end, we design an incentive scheme through bonus payments and participation charges that elicits private demand information from the PMs. We show that the mechanism achieves budget‐balance and voluntary‐participation requirements simultaneously. The results provide important insights into the treatment of misaligned incentives in the context of semiconductor capacity‐allocation.

This article surveys a new generation of analytical tools for capacity planning and management, especially in high-tech industries such as semiconductors, electronics and bio-techs. The objectives of the article are to (1) identify fundamental theory driving current research in capacity management, (2) review emerging models in operations research, game theory, and economics that address strategic, tactical and operational decision models for high-tech capacity management, and (3) take an in-depth look at capacity-optimization models developed in the specific context of semiconductor manufacturing. The goal of this survey is to go beyond typical production-planning and capacity-management literature and to examine research that can potentially broaden capacity-planning research. For instance, we explore the role of option theory and real options in modeling capacity decisions. We not only examine capacity-planning problems from the perspective of a particular firm, but also the interaction of capacity investment among supply chain partners. Not only are these issues increasingly important in the fast-changing high-tech environment, they draw on new tools from different disciplines and pose significant intellectual challenges. We also examine papers that represent the multifaceted nature of high-tech capacity planning, integrating capacity decisions with issues related to contracting, coordination, sourcing, and capacity configurations.

We study strategic capacity planning in the semiconductor industry. Working with a major US semiconductor manufacturer on the configuration of their worldwide production facilities, we identify two unique characteristics of this problem as follows: (1) wafer demands and manufacturing capacity are both main sources of uncertainty, and (2) capacity planning must consider the distinct viewpoints from marketing and manufacturing. We formulate a multi-stage stochastic program with demand and capacity uncertainties. To reconcile the marketing and manufacturing perspectives, we consider a decomposition of the planning problem resembling decentralized decision-making. We develop recourse approximation schemes representing different decentralization schemes, which vary in information requirements and complexity. We show that it is possible to arrive at near optimal solutions (within 6.5%) with information decentralization while using a fraction (16.2%) of the computer time.

Semiconductor capacity planning is a cross-functional decision that requires coordination between the marketing and manufacturing divisions. We examine the main issues of a decentralized coordination scheme in a setting observed at a major US semiconductor manufacturer: marketing managers reserve capacity from manufacturing based on product demands, while attempting to maximize profit; manufacturing managers allocate capacity to competing marketing managers so as to minimize operating costs while ensuring efficient resource utilization. This cross-functional planning problem has two important characteristics: (i) both demands and capacity are subject to uncertainty; and (ii) all decision entities own private information while being self-interested. To study the issues of coordination we first formulate the local marketing and the manufacturing decision problem as separate stochastic programs. We then formulate a centralized stochastic programming model (JCA), which maximizes the firm's overall profit. JCA establishes a theoretical benchmark for performance, but is only achievable when all planning information is public. If local decision entities are to keep their planning information private, we submit that the best achievable coordination corresponds to an alternative stochastic model (DCA). We analyze the relationship and the theoretical gap between (JCA) and (DCA), thereby establishing the price of decentralization. Next, we examine two mechanisms that coordinate the marketing and manufacturing decisions to achieve (DCA) using different degrees of information exchange. Using insights from the Auxiliary Problem Principle (APP), we show that under both coordination mechanisms the divisional proposals converge to the global optimal solution of (DCA). We illustrate the theoretic insights using numerical examples as well as a real world case.

This paper studies the scheduling/rescheduling problem in a multi-resource FMS environment. Several reactive scheduling policies are proposed to address the effects of machine breakdowns and processing time variations. Both off-line and on-line scheduling methods are tested under a variety of experimental conditions. The performance of the system is measured for mean tardiness and makespan criteria. The relationships between scheduling frequency and other scheduling factors are investigated. The results indicated that a periodic response with an appropriate period length would be sufficient to cope with interruptions. It was also observed that machine breakdowns have more significant impact on the system performance than processing time variations. In addition, dispatching rules were found to be more robust to interruptions than the optimum-seeking off-line scheduling algorithm. A comprehensive bibliography is also included in the paper.

We study strategic capacity planning in the semiconductor industry. Working with a major US semiconductor manufacturer on the strategic configuration of their worldwide production capacities, we identify two unique characteristics of this problem as follows: (1) wafer demands and manufacturing capacity are both main sources of uncertainty, and (2) capacity planning must consider two distinct viewpoints: a product perspective concerning marketing and strategic demand management, and a process standpoint involving manufacturing, yield, and technology configuration. These two unique characteristics change, in a fundamental way, how strategic capacity planning problem should be approached. To describe this complex problem, we first formulate a multi-stage stochastic program with recourses where demand and capacity uncertainties are incorporated via a scenario structure. To reconcile the marketing and manufacturing perspectives to the problem, we consider a decomposition of the planning problem resembling decentralized decision-making involving the headquarter, the marketing manager, and the manufacturing manager. To study various trade-offs under this decentralized structure, we develop recourse approximation schemes simulating different decentralization strategies. These schemes vary in information requirements and complexity, while providing insight on the value of information in this environment. We conduct extensive experiments to analyze the characteristics of decisions under different levels of uncertainties, and assess the value of alternative schemes from the standpoint of computational requirements and solution quality. The results indicate that it is possible to arrive at near optimal solutions (within 6.5%) with information decentralization while using a fraction (less than 16.2%) of the computer time.

This paper presents a new algorithm for the flexible manufacturing system (FMS) scheduling problem. The proposed algorithm is a heuristic based on filtered beam search. It considers finite buffer capacity, routing and sequence flexibilities and generates machine and automated guided vehicle (AGV) schedules for a given scheduling period. A new deadlock resolution mechanism is also developed as an integral part of the proposed algorithm. The performance of the algorithm is compared with several machine and AGV dispatching rules using mean flow time, mean tardiness and makespan criteria. It is also used to examine the effects of scheduling factors (i.e., machine and AGV load levels, routing and sequence flexibilities, etc.) on the system performance. The results indicate that the proposed scheduling algorithm yields considerable improvements in system performance over dispatching rules under a wide variety of experimental conditions.