|
|
|
Berkeley Expert Systems Technology: Selected Seminars Andrew Favor, Jennifer Mangold, Stefanie Robinson Berkeley Expert Systemss and Technology a_favor at berkeley dot edu 6102 Etcheverry 2008-02-11, 11:00 Engineering Pathways K-12 Evaluation The Engineering Pathway website, engineeringpathway.com, is a portal to high‐quality teaching and learning resources in engineering, applied science and math, computer science/information technology, and engineering technology and is designed for use by K‐12 and university educators and students. Engineering Pathway is a combined effort of NEEDS.org and TeachEngineering.com to provide a more complete digital library of engineering education materials for all education levels. The goals of the study were: • Needs assessment of K‐12 Teachers • Assess usability of Engineering Pathway website • Assess usability of TeachEngineering website Four methods were used to conduct this study. The methods were chosen to best achieve the goals of the study. A competitive analysis was done to identify the key competitors of Engineering Pathway and to compare them to the website. A heuristic evaluation was done to make an initial assessment of the usability of the Engineering Pathway website. In order to assess the needs of K‐12 teachers’, interviews were conducted. Usability testing was employed to evaluate the usability of the Engineering Pathway and TeachEngineering websites, in order to assess the users overall experience and impressions of the websites. Sohyeong Kim Berkeley Expert Systems Technology sohyeong at berkeley dot edu 6102 Etcheverry Hall 2007-12-07, 12pm An Interdisciplinary Approach of Innovation Portfolio Assessment: Individual innovation projects traversing the funnel go through a process of evaluating their commercial and technological feasibility. A lot of companies have been evaluating a project's feasibility by using an innovation portfolio in order to make sure each of them is in sync with overall system design. By interviewing Innovation Management department at DaimlerCrysler, the evaluation process has been done seperately by each department. This process sometimes does produce unsatisfactory outcome because the final portfolio doesn't fit with the vehicle marketing positioning in the marketplace. I am proposing an interdisciplinary approach to evaluate the overall innovation portfolio based on the case study I conducted with the Innovation Management Department of DaimlerChrysler when they reviewed the S-Class for 2013 model. Jonathan Rheaume UC Berkeley Mechanical Engineering jrheaume at me dot berkeley dot edu 6102 Etcheverry Hall 2007-11-30, 12pm Solid State, Single Cell MICRO2 Sensor (Monitoring In Cylinder Reduction of Oxygen) The goal of my research is to make diesel automobiles go farther on a tank of fuel and to do so with fewer emissions. I am focusing on a sensor to monitor oxygen variation inside engines. For my PhD, I am designing, fabricating, and will test a wide band oxygen sensor that operates in the combustion zone of a lean burn internal combustion engine. The signal that it generates from monitoring the oxygen flux is expected to facilitate improved control of fuel injectors leading to higher efficiency fuel conversion in lean burn engines. The signal will also provide on board diagnostic capability by indicating discrepancies across all cylinders. The ultimate goal is to locate an oxygen sensor inside each cylinder of a lean burn engine. Nobody has ever tried to locate an oxygen sensor inside the piston of an internal combustion engine. Also, no one has ever used gel-casting to fabricate an O2 sensor. In the talk on Nov. 30, I intend to provide an overview of the current state of the art in oxygen sensors and explain the design and fabrication of my own device. In addition, I will present the results of recent modeling that provides illumination on the design and operation of the device. Erik Kolb Berkeley Expert Systems Technology erik.kolb at googlemail dot com 5108 Etcheverry Hall 2007-11-10, 12pm Effect of Coupling between Emergency Department and Inpatient Unit on the Overcrowding in Emergency Department Emergency Department (ED) overcrowding has become a common problem in the United States as well as other developed nations (Australia, Canada, United Kingdom, Singapore, …), threatening the safety of patients who rely on timely emergency treatment. A dramatic consequence of ED overcrowding is the rising amount of patients, who are diverted to far away hospitals, leading to higher mortality rates. Here, we use discrete event simulation to clearly characterize the relationship between Inpatient Unit (IU) utilization and ED crowding: We find that the sensitivity of ED overcrowding with respect to IU utilization depends on the degree of coupling between the two units. We prove that a major cause of ED overcrowding is the volume of patients admitted to the IU. Our findings agree well with previous observations/studies. We suggest using the IU utilization and the ED-IU coupling factor to compare EDs’ capability for delivering timely emergency medical care. Our findings have implications in guiding a hospital’s effort to optimize their system, which could reduce the mortality rate. Hans-Jurgen Sebastian University of Technology RWTH Aachen sebastian at or dot rwth-aachen dot de 6102 Etcheverry Hall 2007-11-02, 12pm Probabilistic and Possibilistic Uncertainties in Engineering Design Engineering Desgin is a creation of new structures and systems. It covers both the engineering analysis and the engineering design synthesis view of the whole process. Uncertainty is envolved in the overall engineering design process. Keno Urquiza Nuclear Thermal Hydraulics Lab (UCB NE Department) keno at me dot berkeley dot edu 6102 Etcheverry Hall 2007-10-26, 12pm Thermo modeling of intermediate heat exchanger Nuclear power technology is poised to safely supply the world's electrical energy demands on the scale currently generated using combustion of fossil fuels while avoiding the greenhouse emissions of fossil fuel dependant methods. The advanced high temperature reactor is a new generation reactor using molten salts to transfer heat from a nuclear reaction to a closed loop multiple reheat helium brayton cycle for power generation. Unlike other reactor designs the AHTR's molten salt loops isolate the core form the power generation. This approach also avoids the inherent complications associated with phase change phenomena as well as high pressure near the reactor; this in turn, increases safety and reduces capital costs. This multiple reheat helium brayton cycle can provide upwards of 50% thermodynamic efficiency and would be significantly cheaper than comparable steam turbine cycles. The proposed system calls for heat to be delivered between 800°C and 1000°C. The elevated temperature requires the implementation of highly compact heat exchanger that retains its strength well above the prescribed temperatures. The research project entails designing and modeling (thermally and mechanically) a compact heat exchanger that can effectively transfer heat from a closed molten salt loop to a closed high pressure helium loop. The compact heat exchanger must be designed to minimize pump work and maximize heat transfer while safely dealing with the high temperature salt, thermal transients, and the operating pressures across the heat exchanger. The heat exchanger geometry is too complex to analyze globally (thermal and mechanical) at the flow channel level, with commercial codes. In this presentation a new modeling approach will be introduced titled, CHEETAH (Compact Heat Exchanger Explicit Thermal And Hydraulics) code. Lora Oehlberg Berkeley Institute of Design lora at berkeley dot edu 6102 Etcheverry Hall 2007-10-12, 12pm Ubiquitous Sustainability: Technologies for Green Values Products that fulfill their functional purpose soon after purchase may become disposable and thus contribute to unsustainable consumption practices. Alternatively, products surrounded by rich stories can become more meaningful to their owners and thus more likely to be kept around beyond their initial use. We propose using sensing and online publishing technologies to capture and share stories from the entire product life cycle, encouraging consumers to become more involved with the production, use, and post-consumer stories of the products they purchase. Blaise Mtopi Laboratoire d'Automatique de Besancon mtopi at yahoo dot com 6102 Etcheverry Hall 2007-10-05, 12pm Methodological Approach to Modular Design: Diversity Modelling in Product Families During the last fifteen years, some essential characteristics of the market have changed. In order to remain competitive, companies must meet consumer's needs and expectations. The fast reactivity and adaptation of firms to this changing environment is strongly related to an advanced control of their internal mechanisms. They compete in five dimensions: price, quality, flexibility, delivery and service. In addition to cost and quality criteria, a tendency to decrease the life cycle of product and increase product diversity is observed. In response to these market changes, companies develop mass customization strategies. Mass customization strategy is designed to support mutually contradicting priorities of price and variety wished by the consumer. Aiming at the computerization, and eventual automation, of product family design, we defines a product family as a structured system to create variety of products with involves the shared base product, but also encompasses customization modules, standard designs, and primary patterns of variety to generate custom designs. The graph language is developed to specify the design space of customizing the base product through manipulating particular modules is modelled by rewriting the starting graph the starting graph using a series of productions according to the control diagram. The graph grammars present both advantages of a mathematical formulation and a graphical representation. Its choice is justified because in put in light the complex linkages between modules and family components through the variants derivation. The development and the application of the model on an industrial case study demonstrate the potential of the graph grammars. We contribute to thinking on complexity in design through the proposal of several matrix transformations. From a structural analysis of products families, we propose decomposition matrix linking functional needs with generic components. We show how to structure technological solutions associated to the functions in diagonal unit matrix. Every unit is looked as a module of products family and allows grouping the technological solutions. This representation allows translating the change from the functional area to the organic area. View Slides Chris Yuan Department of Mechanical Engineering chrisyuan at berkeley dot edu 6102 Etcheverry Hall 2007-09-28, 12pm Life Cycle Impact Assessment for Decision Support in Green Automotive Manufacturing Automotive manufacturing is a major industry in the world, which generates a lot of emissions to the environment. Green automotive manufacturing requires life cycle thinking to avoid shifting environmental problems from one life cycle stage to another, or from one environmental medium to another. This presentation covers the fundamental principles of Life Cycle Assessment (LCA), and analysis procedures of Life Cycle Impact Assessment (LCIA) in real practices. The state-of-the-art LCA tools and LCIA models are analyzed and presented with their advantages and disadvantages. Finally, a research plan is proposed to integrate fate and transport analysis of pollutants into LCIA framework, to develop a completely new LCIA tool to address the problems of current LCIA models. Corie Cobb Department of Mechanical Engineering ccobb at me dot berkeley dot edu 6102 Etcheverry Hall 2007-09-21, 12pm Case-Based Reasoning and Object-Oriented Data Structures Exploit Biological Analogs to Generate Virtual Evolutionary Linkages Multiobjective Genetic Algorithms (MOGA) and Case-based Reasoning (CBR) have proven successful in the design of MEMS (Microelectromechanical Systems) suspension systems. Object-oriented data structures of primitive and complex genetic algorithm (GA) elements have been developed to restrict genetic operations to produce feasible design combinations as required by physical limitations or practical constraints. Thus, virtual linkage between genes and chromosomes are coded into the properties of pre-defined GA objects. A new design problem requires selecting the right primitive elements, associated data structures, and linkages that promise to produce the best gene pool for new functional requirements. In this paper, biomimetics is proposed as a means to examine and classify functional requirements so that case-based reasoning algorithms can be used to map design requirements to promising initial conceptual designs and appropriate GA primitives. The concept is demonstrated using micro-mechanical resonators. Andy Dong University of Sydney a.dong at arch dot usyd dot edu dot au 6102 Etcheverry Hall 2007-09-14, 12pm Language technologies in engineering design This talk covers the development of language technologies to support engineering design. We start by theoretically and philosophically examining what it means to design through words and how it is that words can be seen as designing. Based on this, we develop a theory for the language of design. Using this theory, we look at various computational linguistic tools that help us understand how words design and how, at the same time, this understanding allows us to deploy these tools toward supporting engineering design. We will cover three computational linguistic methods, latent semantic analysis, lexical chain analysis, and sentiment analysis. Chris Yuan UC Berkeley chrisyuan at berkeley dot edu 6102 Etcheverry Hall 2007-08-07, 2pm Green Manufacturing Johanna Mathieu Department of Mechanical Engineering and LBNL jmathieu at berkeley dot edu 6102 Etcheverry Hall 2007-08-03, 2pm Treating drinking water for arsenic in Bangladesh About 30 million people in Bangladesh drink groundwater contaminated with arsenic. Although the WHO's recommended maximum limit for arsenic in drinking water is 10 ppb, the arsenic levels in Bangladesh, in some cases, exceed 1000 ppb. About 20 million people in Bangladesh are already showing signs of arsenic poisoning, in what is rightly called the largest case of mass poisoning in history. Johanna Mathieu and Tasnuva Khan (4th year, Economics) just returned from their second trip to Bangladesh to test a new arsenic removal technology, ARUBA: Arsenic Removal Using Bottom Ash, developed by Dr. Ashok Gadgil at LBNL. On her first trip to Bangladesh in March, Johanna and Susan Amrose confirmed ARUBA's ability to remove arsenic from groundwater. The goal of this trip was to perform a number of experiments to better understand ARUBA's chemical properties. Johanna will discuss the project, results of the trips, and research directions for the future. For more information visit: arsenic.lbl.gov. Jessica Granderson Department of Mechanical Engineering jgrander at me dot berkeley dot edu 6102 Etcheverry Hall 2007-07-24, 4pm Darfur Cook Stoves With large refugee camps, little wood and increasing desertification, Darfur women are facing high risks on their regular searches for wood. More efficient kitchen stoves that burn less wood help reduce the dangerous search for wood, reduce smoke-related diseases and help the local environment to recover. Jessica Granderson will be joining a project in Darfur to help ramp up production of a new design of cook stoves developed by Dr. Ashok Gadgil and his team at Lawrence Berkeley National Laboratory. Early trials with refugee families have proved an overwhelming success and Jessica will help with the shift from small batch manufacture to mass production to meet the substantial demand. The project was recently covered in the Newsweek article: The Flames of Hope Jessica Granderson Department of Mechanical Engineering jgrander at me dot berkeley dot edu 6102 Etcheverry Hall 2007-06-05, 4pm Human-centered sensor-based Bayesian control: Increased energy efficiency and user satisfaction in commercial lighting Esha Datta BEST lab esh at berkeley dot edu 6102 Etcheverry Hall 2007-05-01, 4pm Mobile Learning and Digital Libraries: Designing for Pre-Engineering Education at the Elementary School Level With the recent advancement of mobile technologies, such as smart phones, cameras and PDAs (Personal Digital Assistants), many opportunities arise for informal education. In particular, the use of mobile technologies to access digital libraries opens up doors for providing unique learning experiences outside of the classroom. This paper presents the design and implementation of mobile learning applications. From the data collected during implementation, we provide recommendations for digital library infrastructures and mobile device features that will facilitate informal learning. We first conducted a user needs analysis of students, educators, and parents in order to understand the challenges involved in mobile learning. Ultimately, a list of twelve needs was produced which informed the design of a lesson plan centered around using mobile devices and digital libraries to teach the concept of simple machines. This lesson was implemented during a workshop conducted with elementary school students. From this experience, we were able to provide recommendations for the technology that would enhance mobile learning experiences. Sohyeong Kim Stanford Center for Design Research sohyeong at gmail dot com 6102 Etcheverry Hall 2007-03-20, 4pm Design thinking vs. Analytical thinking Through my research, I hope to seek the answer to how we can improve product planning through design thinking. To do this, I have defined three sub-questions: (1) What is design thinking? (2) What are the current approaches for product planning? (3) What does design thinking offer to product planning? I would like to review the previous studies related to design thinking and product planning. This would entail interviewing product managers, market researchers, designers, and engineers in product development departments of companies, and posing questions related to how they generate and implement the new product ideas. These interviews will allow me to confirm my research questions and to give me insights on real world problems. I believe that the insights gained from my research will serve to enhance understanding of product planning with different approaches, and to increase the potential innovations of new products to be successful in industry. This research will contribute to both engineering and business fields. Priya Sreedharan Department of Mechanical Engineering psreedharan at lbl dot gov 6102 Etcheverry Hall 2007-01-06, 4pm Bayesian based analysis and design of sensor systems for real-time detection of high-risk indoor contaminant releases The deliberate or accidental release of toxic contaminants that reach indoor spaces can be hazardous to the health of building occupants. To respond effectively, one must quickly detect that a release occurred, and also characterize the release in terms of unobserved parameters such as the release source and strength. Designing a robust sensor system that will achieve this objective is challenging for several reasons: the movement of contaminants is complex and inherently uncertain, sensor information is limited and imperfect, and real-time operation imposes computational limitations. Selecting an optimal number of sensors, their placement, characteristics, and sensor interpretation algorithms is not straightforward. This research is based on the premise that to optimize a sensor system design, it is necessary to consider how the sensors will perform as a system. I use a Bayes Monte Carlo framework to design and analyze real-time sensor systems. The Bayes Monte Carlo framework is a hybrid statistical physics algorithm that incorporates both knowledge about the contaminant physical transport, and uncertainties associated with the model and sensors. Using this framework I describe three investigations that seek to understand how information can be better used to optimize a Bayesian based sensor system. I investigate how individual sensor performance, contaminant transport processes, and heterogeneous sensor information relate to overall system performance. These studies are performed using experimental tracer data. Certain results, such as the trade-off between sensor speed and accuracy may not have been inferred without a systems level analyses. Where possible, the results are explained with physical intuition. The overall findings and ideas developed may help in the design of sensor systems for more complex building systems. Jack Tisdale Department of Mechanical Engineering jtisdale at berkeley dot edu 6102 Etcheverry Hall 2006-12-13, tba Stochastic Analysis of a Greedy Algorithm for Unmanned Aerial Vehicle (UAV) Routing This work seeks to characterize the performance of a constant velocity, bounded turn-rate UAV visiting a set of targets uniformly distributed in a specified region. Specifically, we generate an upper bound on the expected time a target must wait before being visited; development of a lower bound is also discussed. Simulation results are found to conform to the theoretical bounds developed. Corie Cobb Department of Mechanical Engineering ccobb at me dot berkeley dot edu 6102 Etcheverry Hall 2006-12-06, 4pm MEMS Design Synthesis: Integrating Case-based Reasoning and Multi-objective Genetic Algorithms A case-based reasoning (CBR) knowledge base has been incorporated into a Micro-Electro-Mechanical Systems (MEMS) design tool that uses a multi-objective genetic algorithm (MOGA) to synthesize and optimize conceptual designs. CBR utilizes previously successful MEMS designs and sub-assemblies as building blocks stored in an indexed case library, which serves as the knowledge base for the synthesis process. Designs in the case library are represented in a parameterized object-oriented format, incorporating MEMS domain knowledge into the design synthesis loop as well as restrictions for the genetic operations of mutation and crossover for MOGA optimization. Reasoning tools locate cases in the design library with solved problems similar to the current design problem and suggest promising conceptual designs which have the potential to be starting design populations for a MOGA evolutionary optimization process, to further generate more MEMS designs concepts. Surface micro-machined resonators are used as an example to introduce this integrated MEMS design synthesis process. The results of testing on resonator case studies demonstrate how the combination of CBR and MOGA synthesis tools can help increase the number of optimal design concepts presented to MEMS designers. Andy Carle and Professor Michael Clancy UC Berkeley EECS acarle at cs dot berkeley dot edu 6102 Etcheverry Hall 2006-11-29, 4pm Pedagogical Patterns as a Foundation for Computer-Mediated Curriculum Design in UC-WISE Pedagogical patterns are abstractions of best practices in pedagogy and educational theory. We will discuss various aspects of pedagogical patterns in the context of the UC-WISE project:
Andy and Mike's PACT and UCWISE presentation Ryan Shelby Department of Mechanical Engineering and New Technologies Engineering Division, Lawrence Livermore National Laboratory ryan_shelby at berkeley dot edu 6102 Etcheverry Hall 2006-11-21, 3:30pm Thermal Endurance of a Cryogenic Pressure Vessel for a Hydrogen Fueled Toyota Prius Onboard hydrogen storage technology is critical component in the realization of a hydrogen-powered economy. Currently, there are four technologies suitable for vehicular hydrogen storage: compressed gas, metal hydride materials, carbon-based materials, and cryogenic liquid. In this study, a thermodynamic model of an insulated cryogenic pressure vessel was created and the liquid nitrogen loss of the vessel due to boiloff was documented. From the thermodynamic analysis, it was determined that the vessel experienced a 5.5 Watt heat leak over the course of the thermal endurance test and that radiation was responsible for ~90% of the heat leak. Given this heat leak, it is predicted that the pressure vessel will be sufficient to store 10.7 kg of (L)H2 for 10 days before boiloff begins. Wendy Ju Stanford University wendyju at stanford dot edu 6102 Etcheverry Hall 2006-11-15, 4pm The Design of Implicit Interactions The infiltration of computer technologies into everyday life has brought the problems of traditional interaction design to a head. As Neil Gershenfeld observes, "There's a very real sense in which the things around us are infringing on a new kind of right that has not needed protection until now. We're spending more and more time responding to the demands of machines." It is clear that we cannot interact with computer devices in a car the way that we do with our desktop computer-but how should we interact with the device while we are driving? We all recognize that the computer need to be less needy of our attention when we are in a meeting, or on the phone, but how do we design these less-needy interfaces? In this presentation, I outline implicit interactions as an emerging area of applied design research that investigates the design of implicit interactions, which occur without the behest or awareness of the user. Implicit interactions allow computers to be more proactive and less distracting through the use of the implicit signals inherent in physical interaction. I will be presenting a framework for implicit interactions, as well as a proposed methodology for implicit interaction design, and on-going research which applies implicit interactions to the design of digital whiteboards, automatic doors, and automotive navigation. Vicente Borja UNAM - National Autonomous University of Mexico 6102 Etcheverry Hall 2006-11-01, 4pm Engineering design at the School of Engineering, UNAM The UNAM (National Autonomous University of Mexico) is the larger and must important university in Mexico. Its relevance is based on its research and cultural facilities and activities and on their undergraduated and postgraduated programs. The School of Engineering of the UNAM, initiated when Mexico was still own by Spain. The School has around 1000 students each year. The research on engineering design in the school is based on its Manufacturing and Design Center. Important figures and relevant facts related to the UNAM and to its School of Engineering will be presented. The Center's research areas and projects will be introduced. Allison Ryan Department of Mechanical Engineering allison at vehicle dot me dot berkeley dot edu 6102 Etcheverry Hall 2006-10-25, 4pm Modeling and Control of Fixed-Wing Aircraft in Longitudinal Flight A non-linear model is developed for a fixed-wing aircraft based on simplified aerodynamic forces. The model represents longitudinal flight by modeling the pitch and altitude dynamics, but neglecting yaw and roll. This type of model is useful for altitude and stability control, and could easily be extended to a full six degrees of freedom. Due to the simplified model, a multi-input multi-output control strategy is selected to be robust to modeling error. A sliding mode controller is designed to regulate pitch angle and airspeed, and shown by simulation to provide the desired performance in spite of modeling error. Finally, a direction for future research (unrelated to the previous control problem) is proposed which makes use of distributed inference and control strategies for teams of unmanned aircraft in mobile sensing applications. Jonathan Hey Berkeley Institute of Design jono at berkeley dot edu Berkeley Institute of Design, 354/360 Hearst Mining Building 2006-10-25, 12pm The Needfinding approach to User-Centered Design User-centered design aims to put the user at the center of the design process. However, as many have highlighted, listening and watching users won't necessarily help you come up with a new breakthrough product. People can't always explain their behavior (or their purchasing decisions) and if pressed will make up a plausible story anyway. Not only that, but people can't tell you what they want. As Henry Ford glibly put it: "If I'd have asked people what they wanted they would have said a faster horse." So how can you find out what folks really want and find a way to give it to them? In this talk I'll outline the broad framework and process of Needfinding as a way to extract insights from user research and turn those insights into action. We'll walk through the four main phases of the needfinding process: 1. Observation - Learning about people 2. building Frameworks - Making sense of what you see 3. Imperatives - Turning insight into action 4. Solutions - Making it real Xue Chen UC Berkeley xuechen at newton dot berkeley dot edu 6102 Etcheverry Hall 2006-10-11, 4pm Demand Response Enabled Thermostat Strategy The State of California is moving toward managing electricity use during periods of shortages, as caused by curtailment of supply or high usage. Measures to reduce the demand for electricity during such shortages are termed 'demand response' (DR). Demand response measures include dynamic rates and some mechanism for utility customers to respond to those rates. DR-enabled thermostat is a kind of device which implements DR-enable technique. It collects information by sensor network and controls the heating ventilation and air conditioning (HVAC) system. DR-enabled thermostat strategy is developed to optimize energy use versus utility price in residential building to respond to the dynamic rates automatically. Residential comfort model is also considered. Strategy uses hierarchical control structure and layered learning. Decision making process for optimization locates in goal seeking layer in control hierarchy. Event based state transition structure and strategies in each state are developed. The interfaces with the other components especially layered learning are also introduced. Simple simulation result shows the feasibility of the hierarchical structure. From engineering point of view, the communication structure of the whole system will also be introduced. Several simulation tools and real house testing have been used for hardware test and software test. Jaspal Sandhu Berkeley Expert Systems Technology Lab 6102 Etcheverry Hall 2006-10-04, 4pm Human-Centered Design of ICT for Development: A Participant Observation Framework for Understanding User Needs The opportunities for introducing information and communication technologies (ICT) into the development context are framed in a history of interventions with widely varying levels of success. In order to increase the probability of success of such interventions, a human-centered approach accounting for local needs is clearly required. Design ethnography, which provides an approach for understanding these needs, has been applied extensively in the fields of human-computer interaction and product design. Although there is some debate as to the merits of using ethnography strictly for design, using ethnographic methods to uncover user needs is a well-established practice, particularly in industry as needs provide a concrete mechanism to move from customer data to design. Aside from exposing articulated needs, a unique advantage of these methods is that they can expose latent or unarticulated needs. Participant observation is a primary method used in design ethnography, but despite this, there is no systematic understanding of the role of participant observation in eliciting user needs. Given the time and resource constraints accompanying product development, such an understanding is urgently needed. While research has been conducted on obtaining customer needs from interviews and focus groups providing guidelines for numbers of participants and number and type of data analysts no such research exists for design ethnography. Participant observation is fundamentally distinct from the methods examined by Griffin and Hauser, so it is unlikely that the same findings will apply. Only Rosenthal and Capper have addressed the issue of resource-effectiveness of design ethnography, but their analysis is not focused on user needs. ICT for development (ICTD) presents unique challenges to design ethnography, since the researchers are often unfamiliar with the context and since the objective often involves technology deployment with users who have little or no ICT experience. As such, a systematic analysis of participant observation for understanding user needs in a rural, development project will be of direct value to the ICTD literature. More broadly, the operationalization of participant observation in terms of user needs will contribute to the design research literature. The research questions will be answered in the context of a single, experimental case: an Asian Development Bank project to develop ICT for rural health services in Mongolia. Specifically, this project will focus on the possible use of ICT to improve continuing education for "bagiin bag emch", rural health workers. Jono Hey Berkeley Expert System Technology Lab jono at berkeley dot edu 6102 Etcheverry Hall 2006-09-07, 11:00am Student Learning Experience with User Needs in Multidisciplinary New Product Development Teams New Product Development (NPD) classes based around problem-based learning provide an effective vehicle for authentic learning and realistic design experiences. Self-reflection provides one mechanism to expose learning accomplishments to the instructor and help the students develop self-assessment skills for professional practice. As one component of a self-assessment learning process this paper describes a structured lessons learned exercise that focuses on what the students themselves found to be the salient learning experiences when engaging in a multidisciplinary product development process. We present an analysis of over 1200 lessons learned collected from three years of a graduate multidisciplinary NPD class with students from engineering, information studies, architecture, business and industrial design. Students highlighted skills for working in multidisciplinary teams as their most important learning experience, whereas the area of greatest impact in the NPD process was the gathering and analysis of customer and user needs. By breaking down these sections we draw insights into students' learning experiences and provide implications for design education in terms of skill sets required and recommended course emphases. Corie Cobb Berkeley Expert System Technology Lab ccobb at me dot berkeley dot edu 6102 Etcheverry Hall 2006-09-07, 11:30am Case-Based Reasoning for the Design of Micro-electro-mechanical Systems Although Micro-Electro-Mechanical Systems (MEMS) are forming the basis for a rapidly growing industry and fields of research, many MEMS designers still rely on back-of-the-envelope calculations due to a lack of efficient computer-aided design (CAD) tools that can assist with the initial stages of design exploration. This paper introduces case-based reasoning (CBR) techniques to the design of MEMS, as part of a larger MEMS synthesis framework currently under development at UC Berkeley. Having the ability to draw upon past design knowledge is advantageous to the MEMS designer, allowing reuse and modification of previous successful designs to help deal with the complexities of a new design problem. CBR utilizes past successful MEMS designs and sub-assemblies as building blocks stored in an indexed library. Reasoning tools find cases in the library with solved problems similar to the current design problem in order to propose promising conceptual designs. This paper discusses case representation and case library design as well as the results of case retrieval studies, focusing on MEMS resonant structures. The paper recommends strategies for integrating the MEMS case library with evolutionary computation when parameter optimization over the retrieved conceptual designs is not sufficient or there are gaps of knowledge in the case library. Charlie Huizenga Center for the Built Environment huizenga at berkeley dot edu 6102 Etcheverry Hall 2006-08-14, 1:30pm Technology Exchange on Energy Efficient Lighting Control Charlie Huizenga is a Research Specialist and Lecturer in the Department of Architecture at UC Berkeley. Charlie leads CBE's building quality benchmarking and occupant feedback projects. This seminar is a discussion and exchange of technologies and experiences on the energy efficient lighting control project funded by the Energy Innovation Small Grant (EISG) of Public Interest Energy Research (PIER) program. To achieve lighting electricity conservation, Charlie developed an affordable and easy-to-install wireless module that interfaces the "smart dust motes" with the ballast of the electirc lights. The module is controlled by a handheld remote controller, and can perform bi-level switching lighting control. Corie Cobb Berkeley Expert System Technology Lab 6102 Etcheverry Hall 2006-03-21, 2:30pm Case-Based Reasoning for the Design of Micro-electro-mechanical This talk discusses preliminary research in applying case-based reasoning (CBR) techniques to Micro-electro-mechanical Systems (MEMS) design as part of a larger MEMS synthesis architecture currently under development at UC Berkeley. Having the ability to draw upon past design knowledge is advantageous to the MEMS designer, allowing reuse and modification of previous successful designs to help deal with the complexities of solving a current problem. Case-based reasoning will utilize past successful MEMS designs and sub-assemblies as building blocks stored in an indexed library. Reasoning tools will find cases in the library with solved problems similar to the current design problem and propose an approximate solution. Igor Verner Department of Education in Technology and Science Technion - Israel of Institute of Technology 3110 Etcheverry Hall 2006-01-24, 4pm Robotics Curriculum: A Syllabus, Program, Process and Practice This talk explores and evaluates different opportunities of learning through designing, building, and operating robot systems. The features of robotics curricula and principles of their design and follow-up are discussed with emphasis on the need of keeping the balance of the two goals of robot projects: (1) creating a working robot prototype, and (2) systematic learning science and engineering subjects. We consider the Technion course "Teaching Methods in Design and Manufacturing" in which teacher education students develop various robots and learn to teach robotics to school pupils and mentor their projects. A number of robot projects developed in connection with the course and through the collaboration of students and pupils are presented. We discuss the methods of assessment of robot projects experienced by our Technion teams at the International Robot Olympiads 2004 and 2005. Shankaran Sitarama Oracle sshankar at me dot berkeley dot edu 6102 Etcheverry Hall 2006-01-23, 10:15am Computational Pattern Analysis of Design Team Interactions This research study integrates computational linguistic and social network analyses to the communication of new product development design teams in order to characterize interaction patterns and team dynamics, as well as identify the signatures of successful teams. Traditional communication and interaction analyses using social networks pose several challenges when applied to design teams due to the complex nature of communication and interaction that takes place in small multidisciplinary teams. Conventional interaction analyses, including social networks, focus on the transactional nature and symbolic behavior of group communication and do not include the "shared meaning" context of group communication. The research presented introduces a framework and methodology that attempts to include the shared meaning of communication in interaction analyses. The basis of the framework is the synthesis of the computational linguistics approach of Latent Semantic Analysis (LSA) and Social Network Analysis (SNA). LSA has been successfully shown in the literature to capture the semantics of text documents and is used in this framework to measure the content of communication and evaluate the semantic similarity and thus the shared meaning in the context of communication. This is then synthesized with the SNA techniques of computing the tie-strengths between actors based on the content of their communication in the team. Traditional metrics of SNA computed for the design teams serve as the basis for team interaction pattern classification. Visualization of the social network thus obtained using multi-dimensional scaling (MDS) for the network node placement and clearly distinguishes the interaction patterns. The intra-team communication of multi-disciplinary design teams in a graduate-level New Product Development (NPD) design course is the subjects of this study. The study includes approximately 20 teams with 4-5 members each and a total of about 4000 email communications. LSA is performed on the team communication to ascertain semantic content of emails. Social network analysis is then applied on the teams with the semantic content, semantic similarity and shared meaning as the basis for tie-strength computations. Social network metrics of network balance, structure and centrality measures provide insight into the interaction patterns of communication of the NPD designers. The design team assessments by instructors, peer-evaluations and external judge ratings serve to characterize the signatures of successful teams. Chiung-Hui Chiu Tainan University, Taiwan cchui at newton dot berkeley dot edu 6102 Etcheverry Hall 2006-01-07, 2pm Collaborative Concept Mapping System This talk will cover materials from Professor Chiu's recent publications including "Exploring how primary school students function in computer supported collaborative learning", "An interface design for a structured computer mediated communication tool for elementary school students: pull-down versus explicit menus", "Evaluating system-based strategies for managing conflict in collaborative concept mapping"; "The impact of difficulty in group task on computer supported collaborative learning", "Exploring the patterns of computer mediated synchronous collaboration by elementary school students", and "Individual differences of elementary school students in synchronous computer mediated small group learning". Eleanor Lee Lawrence Berkeley National Laboratory 6102 Etcheverry Hall 2005-10-03, 3pm Recent on-goings in Windows and Daylighting R&D for Commercial Buildings Eleanor Lee, Scientist/Architect with the Windows and Daylighting Group (Building Technologies Program, Environmental Technologies Division) of LBNL will discuss recent research and development in windows and daylighting for commercial buldings. This will include a detailed discussion of her group's activities, including a 4500 ft2 test facility in Berkeley and a collaborative project at the New York Times Building (NYC). Experimental results, future research, and possible collaboration will be discussed. view slides Jonathan Hey Berkeley Expert System Technology Lab 6102 Etcheverry Hall 2005-08-05, 2:30pm Effective Framing in Design Designers of new innovative products face a complex interaction of social and technical constraints and hidden user needs and perceptions. In order to work with this uncertain and ambiguous information, designers frame the situation, providing structure by selecting the most relevant features of a situation. Framing provides structure to a situation from a particular viewpoint of an actor or designer, both highlighting and hiding different elements or features. Appropriate framing, or problem-setting, of a design situation plays a significant role in the success of the design outcome. This dissertation proposes to better characterize the role that framing plays in design, and the process by which a design team is able to negotiate a shared design frame. Using this new understanding, I will develop tools that facilitate and encourage effective framing within product development and design teams, to increase their likelihood of successful and innovative design outcomes. Jonathan Hey Berkeley Expert System Technology Lab 6102 Etcheverry Hall 2005-07-12, 2:30pm Research into Creative Design This talk discusses my personal path into researching creative design. I discuss experience with techniques for systematic innovation, such as TRIZ, together with new product development experiences using User-centered design. I explain common definitions for creativity from the creativity research literature together with reasons why researching, and in particular, measuring, creativity in design is a challenging thing to do. I continue to discuss the use of metaphors in design and the metaphors we have for the design process. I finish by discussing the role that framing plays in the design process with some examples of problem framing from new product development classes. Jonathan Hey UC Berkeley Department of Mechanical Engineering 6102 Etcheverry Hall 2005-05-17, 2pm Triangulation of Indicators of Successful Design Teams This paper reports on research conducted on design teams at UC Berkeley over several years at the undergraduate and graduate levels. The paper provides a triangulation of indicators of successful design teams drawn from different research methods. The research sources include questionnaires, team documents, email communication, individual design journals, faculty evaluations, and ratings from external design judges. Computational linguistic algorithms are used to analyze the text documents with a focus on latent semantic analysis and semantic coherence. Sketches are analyzed using a comprehensive list of metrics, including Shah's "variety" measure for quantifying the breadth of the solution space explored during the generation process. A synthesis of the results provides interesting and counterintuitive indicators for predicting the success of student design teams. This analysis, in turn, provides insight into learning how the student design teams negotiate and learn the design process and can assist educators in improving the teaching of design. Jialong Wu UC Berkeley Department of Mechanical Engineering jialong at newton dot berkeley dot edu 6102 Etcheverry Hall 2005-05-13, 2pm Unified Language System for Engineering Design (ULSED): A Framework and Automation Tools for Better Design Information Retrieval In the current fast-paced global market, it is critical for companies to have shorter and more effective product design processes. Information management plays an important role in assisting the design teams to achieve that. However, with the heterogeneous nature of the information sources design teams draw upon, traditional information management systems are not able to provide required information effectively. One major reason is that different information sources use different terminologies. The National Library of Medicine (NLM) recognized this problem and initiated the Unified Medical Language System (UMLS) project that aims at providing a platform for medical information exchange and facilitating electronic biomedical information retrieval from multiple sources. This research expands the UMLS model and proposes the Unified Language System for Engineering Design (ULSED) to allow more effective access and exchange of engineering design information. A knowledge repository of the concepts in engineering design is needed in ULSED. This research introduces a method for extracting keyphrase from engineering design documents to suggest terms to represent key concepts. The keyphrase extraction task is modelled as a multi-objective optimization problem and solved with a genetic algorithm. One of the objectives is to maximize the total statistical significance of a set of phrases, which is based on the clumping properties of content-bearing phrases proposed by Bookstein. Another objective is to minimize the number of keyphrases while maintaining the integrity of the phrase set. Results from a set of conference proceedings show promising results when evaluated by a group of experts in engineering design. Another knowledge source in ULSED is a semantic network to connect concepts. This dissertation applies Latent Semantic Analysis (LSA) to discover the semantic relationships between keyphrases. A prototype of the query expansion mechanism was implemented in an engineering educational digital library to assist users in improving their search queries. Federating search over and exchanging information between multiple repositories are important in order to achieve effective collaborations between organizations. One obstacle is that these repositories often use different data models. This dissertation develops a framework for translating data models and implements a web search service for federating multiple repositories. Marisela Avalos UC Berkeley Dept of Mechanical Engineering mavalos at me dot berkeley dot edu 6102 Etcheverry Hall 2005-05-06, 1pm Technology Assessment of Batteries for Implantable Cardioverter Defibrillators Batteries have proven to be a limiting factor in the developments of new electronic equipment and devices in consumer, space, military, and medical applications. Research on any one area of focus for improvement of battery technology takes over a decade to develop for actual use. This is an even longer and arduous process for medical applications due to the health risks involved. In the medical industry, it is also an ethical and financial risk to invest in new battery developments. Yet, this investment is important because it can give a medical device company the lead in the market. An assessment of battery technology and trends in new developments is helpful in determining in what technology a company must invest. This study seeks to evaluate the current battery technology in medical devices and propose a future area of focus for future developments. Since the spectrum of medical devices and their battery requirements is so large, I have chosen to focus on one of the most power demanding devices. This is the implantable cardioverter defibrillator, also known as an ICD. Its functionality in treating abnormal heart beating conditions is assessed to determine the battery requirements for this device. The characteristics of a battery are largely determined by its electrochemical materials. For this reason, the current state of the art materials and those under development are assessed. Trends in electrochemical changes in battery developments are noted and used to project the next best area of focus for new ICD battery developments and investments. Ying Zhang UC Berkeley Civil Systems Program yzh at berkeley dot edu 6102 Etcheverry Hall 2005-04-08, Hierarchical Synthesis and Optimization of MEMS Jaspal Sandhu UC Berkeley Department of Mechanical Engineering 6102 Etcheverry Hall 2005-04-01, 1pm Enabling Access to Medical Technologies: the Aurolab Model Aurolab is one of the only non-profit organizations in the world that produces medical devices or pharmaceuticals. This socially-driven organization produces ophthalmic technologies more cost effectively than any other comparable manufacturer, delivering their products to over 80 countries and owning 10% of the global market for intraocular lenses (IOLs, the synthetic lens implanted during cataract surgery). Moreover, their success has made affordable to those who could not previously afford it, a superior surgical procedure (ECCE) for cataracts. The factors of success span organizational, financial, and technical aspects of Aurolab. The organizational structure is characterized by strong leadership and a unified vision; partnerships with key NGOs have aided in technology transfer; and a close working relationship with AECS has provided a critical resource for product development. Also critical are disease prevalence, disease characteristics, and importance of technology to the cost of surgery. This study utilized primary and secondary data collection techniques. Primary data sources consisted of interviews and site visits, while secondary data collection involved analysis of epidemiological data, financial records, literature reviews, and regulatory issues. Onsite work was conducted during June 2004 in Tamil Nadu, India at Aurolab and the Aravind Eye Care System. Additionally, onsite work in the United States was conducted in early 2004 with Seva Foundation (Berkeley, California) and Project Impact (Berkeley). Joint work with Aman Bhandari and Mahad Ibrahim, UC Berkeley School of Public Health and UC Berkeley School of Information Management and Systems. Eric Eslinger SESAME PhD Candidate, UC Berkeley 6102 Etcheverry Hall 2004-12-03, 2pm Student self-assessment in an interactive learning environment: technological tools for scaffolding and understanding self-assessment practices Metacognitive skills are a crucial component of a successful learning career. We define metacognition as the ability to plan, monitor progress toward a goal, reflect on the quality of work and process, and revise the work or plan accordingly. By explicitly addressing certain metacognitive practices in classrooms, researchers have observed improved learning outcomes in both science and mathematical problem solving. Although these efforts were successful, they were also limited in the range of skills that could be addressed at one time and the methods used to address them due to the static nature inherent in traditional pencil-and-paper format. We wished to address these skills in a more dynamic, continuous representation such as that afforded by a computerized learning environment. This paper outlines such an environment and describes pedagogical activities afforded by the system. The ThinkerTools group developed and tested a software scaffold for inquiry projects in a middle-school classroom. By analyzing student use of the software tool, three forms of self-assessment activity were noted: integrated, task and project self-assessment. Each assessment form was related to the degree of interleaving between assessment and work the students engaged in as they developed their inquiry products. I argue that the integrated forms of assessment are more beneficial to student learning, and show that there is a significant relationship between active self-assessment forms and measures of student achievement and product quality. Through the use of case studies including video analysis, I address specific student self-assessment activity that utilized the software as well as self-assessment that took place outside of the software. A model of student self-assessment activity was created, highlighting aspects of activity that afford more productive self-assessment episodes. Shuang Song UC Berkeley Dept of Mechanical Engineering shuang at me dot berkeley dot edu Tien Room, 6th floor Etcheverry Hall 2004-12-02, 1230pm Shared Understanding, Sketching, and Information Seeking and Sharing Behavior in the New Product Design Process Research on new product design is challenging because of the complex orchestration of craft, human cognition, and social dynamics of the design process. This dissertation introduces methodologies for understanding the new product design process, reporting on research involving three components: information seeking and sharing behavior, sketching, and shared understanding of evolving product designs. The research demonstrates that an extension of Latent Semantic Analysis (LSA), a computational linguistic method, is a promising tool for modeling and evaluating the cognitive and psychosocial behavior of both individual designers and design teams. Further insight is gained by integrating usage tracking, usage analysis and questionnaires to study designers' behaviors in seeking and sharing information. The dissertation also describes a protocol for analyzing design sketches to extract an understanding of design activities. Raffi Kamalian UC Berkeley Dept of Mechanical Engineering raffi at me dot berkeley dot edu 521 Cory Hall, Hogan Room 2004-11-24, 12pm Evolutionary Synthesis of MEMS An evolutionary synthesis framework suitable for Microelectrical Mechanical System (MEMS) design is presented. Genetic algorithms and other synthesis approaches are used to design surface micromachined MEMS using flexural suspensions and electrostatic actuation. Synthesis is based on the use of reduced order modeling to simulate design performance. The applicability of our approach, the design encoding, objectives and constraints are discussed for several MEMS examples, including resonating masses, accelerometers and gyroscopes. Human interactive evolution is applied to MEMS to improve synthesis performance; user studies show an increase in output quality through the use of human interaction compared to our non-interactive synthesis tool. We demonstrate validation of our approach through fabrication and characterization; we were able to successfully generate MEMS devices with measured performance that matched simulation. The results of the characterization are studied to further improve our method for more accurate synthesis. Committee members: Alice Agogino; Kristofer S. J. Pister; Albert P. Pisano. Mirjana Spasojevic HP Labs 3110 Etcheverry Hall 2004-11-19, 2pm How and Why People Use Camera Phones This talk presents an in-depth study into how people use their camera phones. The total of 34 participants were recruited in the US and UK, with approximately a third being teenagers. Using a combined method of interviews and grounded discussions around a sample of actual photos, the study examined people's intentions at the time of capture, subsequent patterns of use, and desires for future technology. The result is a 6-part taxonomy describing the way images are captured both for sharing and personal use, and for affective and functional use. While majority of images are captured with both sharing and affective intentions, a substantial percentage of images support exclusive individual or functional purposes. For the images that are shared, this happens in a variety of ways, from showing images on the phone during face-to-face encounters to other means of sending or posting. The implications of these findings for future products and services are discussed. This is a joint work with Tim Kindberg, Abi Sellen and Rowanne Fleck. View public version of paper. Yao-Jung Wen UC Berkeley Dept of Mechanical Engineering rio at me dot berkeley dot edu 6102 Etcheverry Hall 2004-11-19, 4pm Fuzzy Validation and Fusion for Sensor Networks "Smart Dust" is proposed as a futuristic dust-sized sensing and communication unit based on MEMS technology. Millimeter-scale "motes" are available today as prototypes and can be configured with a variety of sensors in high density distributed sensor networks. While mote sensors are miniature in volume, their power, communication range and memory is limited, and reliability and accuracy vary. Moreover, the sensors integrated on mote sensor boards are not calibrated. Characterization of motes to establish the connection between sensor readings and physical phenomena is essential before pursuing them for "Intelligent daylighting control systems for commercial buildings", and other targeted applications. This presentation covers the characterization of the onboard photoconductors, thermisters and accelerometers, a fuzzy approach of sensor validation and fusion for sensor networks, and the architecture and implementation of mote-based actuation. Viral Shah University of California, Santa Barbara viral at cs dot ucsb 6102 Etcheverry Hall 2004-09-17, 2pm Using Matlab*P The speaker will discuss and demonstrate (on the Millenium Cluster) Matlab*P, a parallel version of Matlab that is well suited to large computation problems like genetic algorithms. Matthew Dubberly UC Berkeley Dept of Mechanical Engineering mattdubb at newton dot berkeley dot edu 6102 Etcheverry Hall 2003-12-05, 1pm The Life-cycle Assessment of Intelligent Lighting System using Distributed Mote Network The Intelligent Lighting System using a Distributed Mote Network, proposed by the UC Berkeley's BEST Lab, was designed to be an efficient lighting solution for commercial buildings. Studies show that this lighting system will decrease buildings lighting electricity consumption by as much as 45%. In addition, economic analysis showed the system installation and retrofit cost could be recovered in approximately three and a half years. Knowing these facts, the proposed intelligent lighting system appears to be an excellent system for saving money and the environment. It is possible, however, that this statement is not entirely true. A previous analysis of this lighting system neglected to examine the external costs related to the environment. Since reduced environmental impact is one of the major selling points of this system, an in-depth environment impact study is required. A Life-cycle Assessment (LCA) was used to compare the current conventional lighting system to the proposed intelligent lighting system. The LCA provides an answer to the question, Is the proposed lighting system better for the environment than the existing lighting systems? The results of the LCA revealed that the impacts of the proposed intelligent lighting system ranged from 17 to 344 times smaller than that of conventional lighting systems for the different environmental impact categories. The LCA also determined the system components that contributed the most to the system impact were the mote printed circuit board, mote integrated circuit, lithium battery, ballast housing paint, and the silicon steel and copper in the ballast transformer and inductor. Adrian Agogino University of Texas, Austin agogino at ece dot utexas dot edu 1165 Etcheverry Hall 2003-11-24, 2pm Design and Control of Large Collections of Learning Agents The intelligent control of multiple autonomous agents is an important yet difficult task. Previous methods used to address this problem have proved to be either too brittle, too hard to use, or not scalable to large systems. The Collective Intelligence project at NASA/Ames provides an elegant, machine learning approach to address these problems. This approach mathematically defines some essential properties that a reward system should have to promote coordinated behavior among reinforcement learners. My talk will focus on creating additional key properties and algorithms within the mathematics of the Framework of Collectives. The additions will allow agents to learn quickly in more complex systems. Also they will let agents learn with less knowledge of their environment. These additions will allow the framework to be applied more easily, to a much larger domain of multi-agent problems. Charlie Ortiz SRI International 1165 Etcheverry Hall 2003-11-10, postponed Distributed Sensor Networks as Multiagent Systems In this lecture, I discuss agent-based approaches to realtime distributed resource allocation. Within such approaches, resources are modeled as agents that negotiate with each other while exchanging resource requirements and task information. I focus on two approaches and systems in particular. Dynamic Mediation is an approach which supports negotiations between agents in the context of a changing problem and resource situation and in which tasks can interact in positive and negative ways. The Distributed Dispatcher Manager (DDM) is a system that supports the hierarchical organization and management of massive agent-based systems, on the order of thousands of agents and tasks. This represents joint work with Harvard and Bar-Ilan Universities and is described in more detail in the recent volume published by Kluwer Academic Publishing entitled "Distributed Sensor Networks," by Lesser, Ortiz and Tambe. Dr. Charlie Ortiz is Program Manager of the Teambotics Program in the AI Center at SRI International. His research interests are in commonsense reasoning (rational action, theories of causation, and counterfactual reasoning) and collaborative systems (multiagent systems, distributed robotics, and collabrative interfaces). website. Raffi Kamalian UC Berkeley Dept of Mechanical Engineering raffi at me dot berkeley dot edu FANUC room (6th floor Etcheverry Hall) 2003-11-08, 1230pm MEMS Synthesis Using Interactive Evolutionary Computation In cooperation with the Kyushu Institute of Design in Japan we have developed an Interactive Evolutionary Computation (IEC) component to our MEMS synthesis software. Allowing a human designer to select and further evolve designs generated by our existing evolutionary multiobjective optimization techniques. This allows a user to include human judgement and design experience in addition to the simulation software's assessment of designs performance, thus avoiding currently non-simulatable design issues such as stress concentrations, dynamic interference, etc. Results from user testing is presented to show the effectiveness of IEC for MEMS synthesis. Masao Arakawa Kagawa University, Reliability-based Information Systems Engineering (RISE) arakawa at eng dot kagawa-u dot ac dot jp FANUC room (6th floor Etcheverry Hall) 2003-11-08, 130pm Development of Genetic Range Genetic Algorithms Usually, Genetic Algorithms have two types of expression: G-type (genotype) and P-type (phenotype). G-types imitate chromosomes, and are usually expressed as strings of binary bit.s P-Type are usually design variables. In conventional GAs, there is fixed setting inbetween them. However, Genetic Range Genetic Algorithms thinks that the setting inbetween two expression as range and inherit its range. Therefore, it has four offsprings out of parents. Whenever the population number of each range becomes small, it means there is no possibility for prosperity. Therefore, those ranges become extinction, and produce new ranges instead. There are several strategies for given new ranges. Both single objective optimization case and multi-objective optimization cases are considered. Beside that, in order to reduce the number of function calls, we introduce radial basis function neural networks (RBF). We have also created an environment to solve any kind of optimization problem through an EXCEL user interface. I will explain a little bit about RBF and demonstrate, the software. After that, I will show you three examples: design of a coil spring, syncroton and bridge. Kai Goebel General Electric Global Research Center 1165 Etcheverry Hall 2003-11-03, 2pm Fuzzy Influence Diagrams General Electric has used soft computing techniques in a variety of fielded casebased reasoning systems. In doing so we were able to leverage the tolerance for imprecision and uncertainty which is intrinsic to soft computing techniques. The payoff of this conjunctive use of soft computing and case-based reasoning techniques is a more accurate and robust solution than a solution derived from the use of any single technique alone. The fielded systems for medical equipment diagnostics and residential property valuation used fuzzy membership functions for greater selection accuracy through their noise tolerance and for the ability to determine a confidence in a result. A plastic color matching system used fuzzy logic to combine several selection criteria which allowed its users to detect potential problems during the case selection phase while previously these problems were not detected until the color match was complete. Finally, adaptive fuzzy clusters were used for fault classification of aircraft engine sensor data. The advantages of this system were its ability to deal with extremely noisy sensors and to adjust to unpredictable slow drift. Piero P. Bonissone General Electric Global Research Center bonissone at research dot ge dot com 1165 Etcheverry Hall 2003-10-29, 2pm Soft Computing and Meta-heuristics: Using Knowledge and Reasoning to Control Search and Vice-versa Meta-heuristics are heuristic procedures used to tune, control, guide, allocate computational resources or reason about object-level problem solvers in order to improve their quality, performance, or efficiency. Offline meta-heuristics define the best structural and/or parametric configurations for the object-level model, while on-line heuristics generate run-time corrections for the behavior of the same object-level solvers. Soft Computing is a framework in which we encode domain knowledge to develop such meta-heuristics. We explore the use of meta-heuristics in three application areas: a) control; b) optimization; and c) classification. In the context of control problems, we describe the use of evolutionary algorithms to perform offline parametric tuning of fuzzy ontrollers, and the use of fuzzy supervisory controllers to perform on-line mode-selection and output interpolation. In the area of optimization, we illustrate the application of fuzzy controllers to manage the transition from exploration to exploitation of evolutionary algorithms that solve the opimization problem. In the context of discrete classification problems, we have leveraged evolutionary algorithms to tune knowledge-based classifiers and maximize their coverage and accuracy. Lotfi Zadeh UC Berkeley Dept of Electrical Engineering and Computer Science 1165 Etcheverry Hall 2003-10-20, 2pm From Search Engines to Question-Answering Systems - The Need For New Tools Search engines, with Google at the top, have many remarkable capabilities. But what is not among them is the deduction capabilitythe capability to synthesize an answer to a query by drawing on bodies of information which are resident in various parts of the knowledge base. It is this capability that differentiates a question-answering system, Q/A system for short, from a search engine. Dealing with world knowledge needs new tools. A new tool which is suggested for this purpose is the fuzzy-logic-based method of computing with words and perceptions (CWP), with the understanding that perceptions are described in a natural language. A concept which plays a key role in CWP is that of Precisiated Natural Language (PNL). It is this language that is the centerpiece of our approach to reasoning and decision-making with world knowledge. Jessie Hey Southampton University Library Member of the Intelligence, Agents, Multimedia Group in the School of Electronics South Hall 107 2003-10-03, 3-5pm Academic Scholarship and the Deep (or Invisible) Web Academics constantly need to keep up to date with the latest work for their research and their teaching. However, resource discovery has become a complex task in a hybrid world of paper and digital libraries. Various techniques have been tried to make this easier. When Google becomes a search engine of choice many valuable resources lie behind a barrier that we think of as the invisible web. We describe experiments with a global information gathering agent, combining agent technologies and information management skills, to make visible these hybrid resources. We then discuss an alternative approach stimulated by the Open Archives Initiative in which academic e-Print archives become harvested by global search services. *Biography:* Dr. Jessie Hey has worked in information management for many years at the interface of computers and users and has taught courses on Human Computer Interaction and many workshops on the electronic/digital library. She has a Physics degree from Oxford, a postgraduate Certificate of Education, and Diploma in Library and Information Studies. She is a Chartered Librarian (MCLIP) and Member of the ACM. Besides spells at Caltech in Pasadena and CERN in Geneva, she worked in the UK in a variety of posts in higher, further and primary education. She was Manager of Technical Information Services at IBM's UK Research Labs for some years where she also set up an interactive learning centre. At Southampton University Library she supported Engineering, Mathematics and Physical Science users and is now working on the UK funded TARDis e-Prints repository project. A member of the Intelligence, Agents, Multimedia Group in the School of Electronics and Computer Science, she has previously worked on digital and 'hybrid' library projects such as ERCOMS and MALIBU. She completed a PhD in Resource Discovery In Digital Libraries last year and maintains a wide interest in digital libraries and scholarly communication issues as we move towards the Semantic Web. Personal webpage: http://www.ecs.soton.ac.uk/~jmnh/ Related URLs: http://tardis.eprints.org/ and http://www.ecs.soton.ac.uk/ "Building quality assurance into metadata creation: an analysis based on the learning objects and e-prints communities of practice" Barton, Currier, and Hey will be presented at DC-2003 on 29th September. Michael Wetter LBNL, Simulation Research Group mwetter at lbnl dot gov 6102 Etcheverry Hall 2003-06-02, 1pm Nonlinear programming for engineering design optimization Nonlinear programming (NLP) algorithms can assist the designer in solving complex engineering design optimization problems. In many cases, the performance is measured by a cost function that is smooth in the design parameters. However, if the cost function is evaluated by a computer code that contains adaptive solvers, such as for the iterative solution of nonlinear equations or for the step-size control of integration methods, then it becomes replaced with an approximation that fails to be even continuous. Consequently, standard NLP algorithms can fail far from a solution. We will present a classification of different engineering optimization problems and discuss the applicability of various optimization algorithms. We compare algorithms that use Computational Intelligence methods, derivative free deterministic algorithms and gradient-based algorithms. We present a precision control scheme that can be exploited to obtain a significant reduction in computing time, and smoothing methods for discontinuous cost functions. We will compare the performance of different algorithms in optimizing a building's envelope, air-conditioning system and daylighting control, and present an application of optimization with adaptive precision simulations. Hesham Mohamed Kamel CS Department, UC Berkeley 6102 Etcheverry Hall 2003-05-19, 1pm The Integrated Communication 2 Draw When sighted persons draw, they continually adjust their input based on visual feedback, an essential part of the drawing process. Existing drawing tools for blind users give inadequate contextual feedback on the state of the drawing, leaving blind users unable to comprehend and successfully produce graphical information. In animation, feedback becomes even more crucial. Although visually impaired people are able understand the meaning of what is portrayed in an animation, today there is also no accessible method for them to create animation. This dissertation describes the design, implementation, and evaluation of a dynamic drawing and animation tool for the blind, call the Integrated Communication 2 Draw (IC2D), that overcomes these problems and combines a simple grid-based navigational interface, keyboard input, and auditory feedback in a 2D graphics environment. When drawing, one of the most important operations is point location and relocation. One must be able to conceptually relate one point to the whole image or another point in the image. For visually impaired users, some guideline or frame of reference is needed. This same frame of reference must also offer a mechanism to specify the angle of a line. One solution that addresses the described needs is a grid. IC2D is based on a 3-3 grid that mimics the layout of the telephone keypad, a familiar interface for most blind users. The grid introduces nine fixed regions with unique center points, where one is the top left corner and nine is the bottom right corner of the drawing area. The grid can be used recursively inside each of the nine regions to allow finer point selections. Rather than force visually impaired users to adapt to a limited method for creating and accessing graphical information formatted for sighted users, IC2D overcomes the disadvantages of current tools and implements methods more intuitive to visually impaired users, allowing them to create and explore recognizable drawings, and graphically communicate with sighted people. Several evaluations of IC2D show that visually impaired users can employ 2D drawing applications with the accuracy of sighted users. Lori Lorigo Cornell University 6102 Etcheverry Hall 2003-04-14, 1pm Design of an Interactive Digital Library of Formal Algorithmic Knowledge Working for nearly a century, mathematicians, logicians and computer scientists have achieved the practical means to formalize and mechanically verify vast amounts of mathematical knowledge. That formal knowledge and the tools used to create it have been used to improve the reliability of software and our ability to program it more efficiently. Storing this formal material on-line in digital libraries will further increase its value. In this talk I present the first prototype digital library, called a Formal Digital Library (FDL) built at Cornell, Cal Tech and Wyoming, under an active MURI Grant from the Office of Naval Research (ONR). This library contains definitions, theorems, theories, proof methods, and articles about topics in computational mathematics. The basic library operations will be described, as well as capabilities to check proof dependencies and share material from differing theories. I will discuss design challenges when dealing with mathematical content, include examples of critical identification and search functionalities developed, and pose open questions for further development. In addition, I will mention other work being done at Cornell including NSF's National Science Digital Library, an application of machine learning to ranking functions, and a model of what's being called "burstiness" which reflects peaks of content or links in internet log files. Lori Lorigo has been a Programmer Analyst at Cornell University since 1997, working with CIS Dean Robert Constable in the Formal Methods Research group. She completed her Master's of Engineering in Computer Science at Cornell in January 2001, and B.S. in Computer Science and Operations Research and Industrial Engineering in 1997. She has published several papers in the field of automated reasoning, with applications to verification, embedded system design, and education. Shuang Song, Andy Dong, and Alice M Agogino UC Berkeley Dept of Mechanical Engineering 6102 Etcheverry Hall 2003-04-14, 1pm Time Variation of Design "Story Telling" in Engineering Design Teams Engineering design practice can be viewed as a process of "story telling," where the "stories" generated explore various aspects of the design process and the designed artefact. This paper focuses on the study of the design process by using computational linguistics to analyze time variant patterns of "story telling" in eight multidisciplinary student design teams at the University of California at Berkeley, USA. We examine the oral and written histories left by the designers through their documentation, presentation material, and e-mail communication. A formal, quantitative methodology for time variant analysis of "story telling" in engineering design teams based on metrics of semantic coherence is established. Results from the analysis suggest a positive correlation between design outcomes and patterns of the average semantic coherence over time as well as with variation in semantic coherence between design stages. This research provides empirical evidence of the phenomena of changing levels of coherence in "story telling" in design and the scope of design concepts explored by design teams. Further, we show that both the e-mail data set and the documentation data sets contain approximately the same amount of "story telling" capacity. | ||||||||||||||||||||||||||||
