Institute for Systems Research Technical Reports
Permanent URI for this collectionhttp://hdl.handle.net/1903/4376
This archive contains a collection of reports generated by the faculty and students of the Institute for Systems Research (ISR), a permanent, interdisciplinary research unit in the A. James Clark School of Engineering at the University of Maryland. ISR-based projects are conducted through partnerships with industry and government, bringing together faculty and students from multiple academic departments and colleges across the university.
Browse
Search Results
Item Improving TCP Performance over High-Bandwidth Geostationary Satellite Links(1999) Bharadwaj, Vijay G.; Baras, John S.; ISR; CSHCNThe Transmission Control Protocol (TCP) is the most widely used transportprotocol in the Internet today. The problem of poor TCP performance oversatellite networks has recently received much attention, and much work hasbeen done in characterizing the behavior of TCP and proposing methods forimprovement. Meanwhile it remains hard to upgrade the majority of legacyhost and gateway systems in the Internet that are running old and outdatedsoftware so that they can perform better in the changing networks of today.In this thesis we consider an alternative network architecture, where largeheterogeneous networks are built from small homogeneous networksinterconnected by carefully designed proxy systems. We describe the designand implementation of such a proxy and demonstrate marked performanceimprovements over both actual and simulated satellite channels. We alsodiscuss some benefits and drawbacks of using proxies in networks andexplore some tradeoffs in proxy design.
Item Statistical Parameter Learning for Belief Networks with Fixed Structure(1999) Li, Hongjun; Baras, John S.; ISR; CSHCNIn this report, we address the problem of parameter learning for belief networks with fixed structure based on empirical observations. Both complete and incomplete (data) observations are included. Given complete data, we describe the simple problem of single parameter learning for intuition and then expand to belief networks under appropriate system decomposition. If the observations are incomplete, we first estimate the "missing" observations and treat them as though they are "real" observations, based on which the parameter learning can be executed as in complete data case. We derive a uniform algorithm based on this idea for incomplete data case and present the convergence and optimality properties. Such an algorithm is suitable trivially under complete observations.Item An Introduction to Belief Networks(1999) Li, Hongjun; Baras, John S.; ISR; CSHCNBelief networks, also called Bayesian networks or probabilistic causal networks, were developed in the late 1970s to model the distributed processing in reading comprehension. Since then they have attracted much attention and have become popular within the AI probability and uncertainty community. As a natural and efficient model for humans' inferential reasoning, belief networks have emerged as the general knowledge representation scheme under uncertainty.In this report, we first introduce belief networks in the light of knowledge representation under uncertainty, then in the remainingsections we give the descriptions of the semantics, inference mechanisms and some issues related to learning belief networks, respectively. This report is not intended to be a tutorial for beginners. Rather it aims to point out some important aspects of belief networks and summarize some important algorithms.
Item An Automated, Distributed, Intelligent Fault Management System for Communication Networks(1999) Li, Hongjun; Baras, John S.; Mykoniatis, George; ISR; CSHCNIn this paper we present a Distributed Intelligent Fault Management (DIFM) system for communication networks. The overall architecture of the proposed system is based on a distributed, cooperative, multi-agent paradigm, with probabilistic networks as the framework for knowledge representation and evidence inferencing. We adopt the management by delegation paradigm for network monitoring and integrate both hard and soft faults.Item REU Report: Development of an Agent-Based Factory Shop Floor Simulation Tool(1999) Whangbo, Albert J.; Lin, Edward; Herrmann, Jeffrey; ISRManufacturing systems of the future are expected to be agile and failure-tolerant. Current simulation tools are not well equipped to model these dynamically changing systems. Agent-based simulation represents an attractive alternative to traditional simulation techniques. This project aims to develop software for agent-based factory shop floor simulation. The current version of the factory simulation software is implemented in Java. Although the program lacks important features of a decision support tool, it provides a flexible agent-based framework for modeling and testing shop floor configurations.Item Fixture-Based Design Similarity Measures for Variant Fixture Planning(1999) Balasubramanian, Sundar; Herrmann, Jeffrey W.; ISROne of the important activities in process planning is the design of fixtures to position, locate and secure the workpiece during operations such as machining, assembly and inspection. The proposed approach for variant fixture planning is an essential part of a hybrid process planning methodology.The aim is to retrieve, for a new product design, a useful fixture from a given set of existing designs and their fixtures. Thus, the variant approach exploits this existing knowledge.
However, since calculating each fixture's feasibility and then determining the necessary modifications for infeasible fixtures would require too much effort, the approach searches quickly for the most promising fixtures based on a surrogate design similarity measure. Then, it evaluates the definitive usefulness metric for those promising fixtures and identifies the best one for the new design.
Item Structural Matrix Computations with Units: Data Structures, Algorithms, and Scripting Language Design(1999) Austin, Mark; Lin, Wane-Jang; Chen, Xiaoguang; ISRDespite the well-known benefits of physical units, matrices,and matrix algebra in engineering computations,most engineering analysis packages are essentially dimensionless.They simply hold the engineer responsible for selecting a set ofengineering units and making sure their use is consistent.While this practice may be satisfactory for the solution ofself-contained and well-established problem-solving procedures,where the structure of the output is well known and understood,identifying and correcting unintentional errors in the solution ofnew and innovative computations can be significantly easierwhen units are an integral part of the computation procedure.This report begins with a description of thedata structures and algorithms needed torepresent and manipulate physical quantity variables,and matrices of physical quantities.
The second half of this report focuses on the implementation of Aladdin,a new computational environment for matrix and finite element calculations.Aladdin employs a novel combination of system programming languages,scripting language concepts, and stack machine technology.The result is a high-level scripting language that offers enhancedtype checking for expressions and assignments,problem-oriented scaling of variables, automatic conversion of systems of units, and program control structures for the solution of engineering problems.
Functionality of the Aladdin stack machine is illustratedby working step by step through the parsing and execution ofa simple statement involving units.The capabilities of Aladdin are demonstrated through thedeflection analysis of a cantilever beam.
Item Simulation-Based Algorithms for Average Cost Markov Decision Processes(1999) He, Ying; Fu, Michael C.; Marcus, Steven I.; Fu, Michael C.; Marcus, Steven I.; ISRIn this paper, we give a summary of recent development of simulation-based algorithmsfor average cost MDP problems, which are different from those for discounted cost problems or shortest pathproblems. We introduce both simulation-based policy iteration algorithms and simulation-based value iterationalgorithms for average cost problems, and give the pros and cons of each algorithm.Item Reducing Manufacturing Cycle Time during Product Design(1999) Herrmann, Jeffrey W.; Chincholkar, Mandar; ISRThis paper describes an approach that can reduce manufacturing cycle time during product design. Design for production (DFP) determines how manufacturing a new product design affects the performance of the manufacturing system. This includes design guidelines, capacity analysis, and estimating manufacturing cycle times. Performing these tasks early in the product development process can reduce product development time. Previous researchers have developed various DFP methods for different problem settings. This paper discusses the relevant literature and classifies these methods. The paper presents a systematic DFP approach and a manufacturing system model that can be used to estimate the manufacturing cycle time of a new product. This approach gives feedback that can be used to eliminate cycle time problems. This paper focuses on products that are produced in one facility. We present an example that illustrates the approach and discuss a more general approach for other multiple-facility settings.Item Using Neural Networks to Generate Design Similarity Measures(1999) Balasubramanian, Sundar; Herrmann, Jeffrey W.; Herrmann, Jeffrey W.; ISRThis paper describes a neural network-based design similarity measure for a variant fixture planning approach. The goal is to retrieve, for a new product design, a useful fixture from a given set of existing designs and their fixtures. However, since calculating each fixture feasibility and then determining the necessary modifications for infeasible fixtures would require too much effort, the approach searches quickly for the most promising fixtures. The proposed approach uses a design similarity measure to find existing designs that are likely to have useful fixtures. The use of neural networks to generate design similarity measures is explored.This paper describes the back-propagation algorithm for network learning and highlights some of the implementation details involved. The neural network-based design similarity measure is compared against other measures that are based on a single design attribute.