Department of Electrical and Systems Engineering

ESE Senior Design, 2010-2011
Ken Laker, Raymond Watrous, Peter Scott

ESE Home Page > ESE Undergraduate Labs > Senior Design > 2010-2011 Abstracts


Street Solutions

Dhruv Batura,Rylan Collins, Aditya Kaji, Eric Lamb

Advisor:  Peter Scott & Walter Sobkiw


In countless applications, there is a need to monitor and respond to problems within a system. Monitoring can take on the form of an aggregation of individual reports or data points, and that aggregation provides insight into the current status of the overall system. This insight can then determine the necessary response. In most American cities, aesthetic, maintenance, and sanitation problems are reported by citizens on an individual basis and are then prioritized and addressed.

The purpose of the StreetSolutions system is to engage the public in the monitoring of problems (sanitation, neglected properties, illegal dumping, etc.) in small (approximately 100,000 residents or less) urban areas. Essentially, StreetSolutions works by compiling reports from individuals. These reports are generated either from a mobile phone or online. These reports are then made visible on a blog-style website with a map, pictures, and descriptions. The eventual aim is that city authorities could use a system like this to see which problems the public cares most about. Most importantly, by making the reports public, the process becomes more transparent while also increasing the city’s accountability to its citizens.

View: Poster / Movie

AutoPlug: Open Architecture for Automotive Services

Authors: Ross Boczar, Jason Suapengco, and Gabriel Torres

Advisor:  Rahul Mangharam


The modern automobile features more software components and electronics than ever before.Unfortunately, many of these components (known as ECUs - Electronic Control Units) are “black boxes”—their internal states are difficult to determine. Thus, there is a need to remotely diagnose, update and certify automotive software for efficient warranty and safety management.
AutoPlug is an automotive ECU testbed to develop mechanisms and protocols for remote diagnosis, programming and testing of future vehicles. The goal of AutoPlug is to allow car manufacturers to better identify, characterize, and predict automobile software errors.
AutoPlug consists of three layers: the vehicle simulation layer, the ECU network layer, and the middleware layer. The vehicle simulation layer, implemented in the open-source racing engine TORCS, provides realistic data to the ECU network. The ECU network layer, which models individual car systems, such as steering and anti-lock brakes, receives this data and applies typical control algorithms. The middleware layer controls the interconnections of the ECU network and is responsible for acting as a gateway (via a user interface) between the manufacturer and the ECU network. This layer provides the interface for re-flashing software, diagnosing software errors, and gathering vehicle statistics.
This system is able to demonstrate scenarios where a software error is observed in simulation, quantitatively diagnosed, and then corrected by upgrading the software through the AutoPlug portal.  Additionally, an AutoPlug Android application show the potential for a large software upgrade rollout by the manufacturer, authenticated by the user via smartphone.

View: Poster / Movie

AccuEnergy - Forecasting System for Penn’s Electricity Consumption

Authors: Madhur Agarwal, Jesse Beyroutey, Grace Gui, and Katie Joo

Advisor:  Andrew Huemmler


Given the volatility in today’s energy markets, large consumers need to carefully manage their electricity spending. The University of Pennsylvania (Penn) is one of the largest electricity consumers in Philadelphia. Penn’s Facilities and Real Estate Services (FRES) is in charge of energy management, and in an effort to reduce costs, has decided to become an electric load-serving entity (LSE).
Functioning as an LSE will allow FRES to purchase electricity on the open market, eliminating the need for intermediaries. This move will newly expose Penn to the risk of electricity price fluctuations on the local market, while providing the opportunity to leverage better internal information about campus electricity consumption.
In order to help FRES manage its risk appropriately, AccuEnergy is a forecasting system that accurately predicts Penn’s electricity consumption, allowing FRES to better understand the campus’ electricity demand throughout the year. The AccuEnergy system predicts demand using a time-series forecasting model that accounts for recent consumption and a forward temperature and relative humidity forecast. By modeling the particulars of Penn’s 2007-2011 consumption patterns, a nonlinear predictive model has been built that allows FRES to determine its optimal electricity purchasing strategy for yearly budget and day-ahead purchasing purposes.
With the use of AccuEnergy, FRES can minimize Penn’s total cost of electricity consumption and realize the financial gains made possible by being an LSE. In turn, this allows facility administrators to maintain their buildings more efficiently and economically.

View: Poster / Movie

Penn Cap and Trade

Authors: Hyung Soo Byun, Eve Ying Lee, Junxu Lye

Advisor: Andrew Huemmler, Peter Scott  


In 2007, the University of Pennsylvania instituted the Penn Climate Action Plan (PCAP) with the goal of reducing Penn’s carbon emissions by 80% (over 2007 levels) by 2050. However, no tangible progress towards the goal has been made as the University’s cost centers (e.g. SEAS, Wharton, SAS) are not sufficiently incentivized and have no systematic means of tracking their carbon emissions reductions. “Penn Cap and Trade” proposes the use of a University-wide Emissions Trading System (ETS) as a means of achieving PCAP’s carbon emissions reduction goal. The central administrator of the system will be Facilities and Real Estate Services (FRES), while the cost centers will be the players in the system trading carbon credits. The “Cap” represents the mandated amount of carbon emissions reduction each period. The eventual reduction target of 80% in 2050 will be achieved by adjusting the “Cap” in stepwise increments of 2% each year over a 40 year period. Cost centers can achieve their target reductions by buying carbon credits or implementing carbon-saving projects. In the system, one carbon credit is equivalent to one ton of carbon emissions reduction. If a cost center exceeds its target by implementing projects, it will have excess credits to sell. Conversely, a cost center that implements fewer projects will have to buy credits.
The model provides a unique user interface for each cost center through which they input their project selection preferences and project data. The model then aggregates this information across the University and produces outputs such as the price of a carbon credit for each period, projects implemented, and the total cost of achieving the target reductions for each cost center and the overall University.
This model proposes four main benefits over the status quo. First, the model serves as a tracking tool that allows its users to keep track of their progress towards PCAP’s goals. Second, the model serves as an evaluative tool to assist cost centers in their decision making process as to whether they should implement a carbon reduction project or buy carbon credits. Third, the model allows cost centers to clearly examine the financial consequences of suboptimal project selection preferences and pushes them to re-think such preferences. Fourth, the model proposes a redistribution of wealth in which a cost center could potentially fund another cost center’s more efficient carbon savings project through the buying of credits. The model’s versatility allows for the potential realization of these benefits in institutions in and beyond Penn’s campus. 

View: Poster / Movie

Urban Energy Consumers as Solar Energy Producer’s

Authors: Soh Nagano, Kristin K Yamauchi, Amanda Zwarenstein, Pallavi Yerramilli

Advisor:  Roch Guerin


Renewable energy policy is quickly coming to the forefront of political debate as countries attempt to reap the benefits of investing in alternative energies. However, very little is understood of the complex interactions between the consumers, utility companies, and governments that define the system, limiting the ability of policymakers to implement effective incentives and utilities to understand the impact of these incentives on their business. This study focuses on the adoption of solar energy technology in cities, and explores how certain drivers, such as retail rate, tax subsidies, and Solar Renewable Energy Credits (SRECs) affect commercial and residential investment to aid the development of better strategies towards renewable energy deployments.
A research-intensive methodology was taken in approaching this issue to gain a better understanding of the decision processes that govern the actions of the different stakeholders. Findings were then directed towards the development of a model that simulates the auction process used by utility companies to purchase SRECS, shedding light on how government subsidies, tax credit, and the demographic of the target population affect the prices bid, and in turn affect resulting adoption, SREC price, and retail rate of electricity.
By conducting sensitivity and scenarios analyses, we observed a gradual decline in SREC prices, an increase in average retail rates, heavy commercial user adoption, and indication that the decline in solar panel prices and other effects of increased investment in solar energy will slowly incentivize more consumers to invest in this technology without relying so heavily on the government.

View: Poster / Movie

R.A.V.EN – Remote Autonomous Vehicle Explorer Network

Authors: William H Etter, Paul D Martin

AdvisorRahul Mangharam


As robotic platforms and unmanned aerial vehicles (UAVs) increase in sophistication and complexity, the ability to determine the spatial orientation and placement of the platform in real time (also known as localization) becomes an important issue. What is a relatively simple task for a human operator to complete becomes a daunting process for an autonomous platform.
Current methods to achieve localization in UAV systems require computation-intensive sensor systems on the platforms themselves or pre-installed in the location of interest.  This not only increases the system cost, but also decreases the overall applicable range of the platform.  In addition, no system to control multiple units simultaneously for search and rescue (SAR) exists. This prevents more capable and advanced platforms from being used in situations where they could provide the most benefit.
The system presented uses an on-board near-infrared camera to track beacons on user-controlled quadrotor units.  By fusing this data with orientation data that is broadcasted over a wireless network from leader to follower, an accurate position relative to other quadrotors can be determined.  Provided with this accurate position relative to the leader, a slave quadrotor can autonomously follow based on the amplitude of its error in three-dimensional space.
The resulting system demonstrates a relative localization scheme with centimeter accuracy at under 3 meters distance and less accurate localization information at larger distances.  Under normal conditions, a second follower quadrotor can respond appropriately to its error displacements and track the movements of the lead quadrotor.

View: Poster / Movie

Intelligent Information Systems for Technical Trading Analysis

Authors: Ranganathan Ramachandran, Michael Y Zhao, Hicham Laalej

Advisor: Prof. Zhao

Traders buy and sell financial instruments in order to make a profit. Technical analysis is a set of tools some traders use to predict future price movements from past price and volume trends. By analyzing charts, traders identify technical indicators, or specific market trend patterns, which have associated future market movements.
The pretrade analysis that traders conduct is a lengthy procedure, during which every trader has to go through various charts, data, and information before executing a trade. The amount of time required before executing a trade limits the frequency of trades and the ability to trade on a shorter time-frame.
By building increasingly comprehensive pattern recognition functions based on historical data, we have created a model which recognizes the appearance of technical trading indicators and evaluates a security’s past performance. Based on long- and short-term price and volume trends, our model identifies different conditions for technical indicators which it applies to real-time data.
An improved prediction over current technical analysis methods is made possible by the automated evaluation of more data points per period and the analysis of secondary characteristics of indicators. Additionally, the automation of our model enables identification of technical indicators in a fraction of the time previously required. These improvements make it possible for traders to perform technical analysis more accurately and faster than they could before.

View: Poster / Movie

Economic Network Model : Forecast of Cascading Shocks

Authors: Robert Feigenberg, Jungsun (Ella) Kim, Carolina Lee, Daniela Savoia

Advisor:  Ali Jadbabaie & Alejandro Ribeiro

The United States economy can be modeled as an interconnected network made up of various sectors. The connections between sectors determine the network structure and the influence levels of each sector on the rest of the economy. The sector relationships are determined by the use of one sector’s output as another sector’s input.
When a shock, or a large change in expected output, occurs in a given sector, the effect of the disturbance is experienced throughout the entire economy. It is a useful tool to forecast the magnitude and spread of the effects of a shock over time.
The approach for forecasting the shock effects in an economy is to determine the network structure and connections between sectors. A mathematical model is used to measure the random noise in the economy and differentiate it from a larger shock in the economy. Using this model, the effects of an individual shock are forecasted and simulated over time.
The model’s accuracy is verified with data on past shocks in the US economy. The model visually displays the network structure in a diagram and a table, highlighting the central and most influential sectors in the economy. The forecasted effects of a shock are simulated over time and demonstrated graphically. The final product design is an application that allows users to input the origin and magnitude of a shock and analyze its forecasted effects throughout the economy.

View: Poster / Movie


Authors: Shahid I Bosan, Dara M Elass, Divya Singhal, Aarti D Kochhar

Advisor:  Peter Scott & Walter Sobkiw


As fossil fuel supplies are depleting at increasing rates, wind energy is gaining popularity as a clean and renewable power source in the United States. However, choosing a turbine appropriate for a given location and needs is a lengthy and expensive process. Furthermore, lack of knowledge and resources can result in an inefficient selection. Recent examples of poor turbine choices have resulted in lack of power generation and high costs.
Greengineering aims to simplify the process of selecting an appropriate turbine to meet the client’s specifications. The focus is on urban areas in the United States to allow for individuals and small institutions to install turbines. The model runs in MATLAB and has an easy-to-use graphical user interface. To additionally simplify the Greengineering experience, the information required to run the model is readily available to the user and requires minimal research.
The model includes a database of turbines and wind data for numerous cities in the United States. After the decision maker has chosen a location, the model runs by filtering out turbines that would be unacceptable based on budget, building weight capacity, available area and power generation requirement. The output of the model gives the client the best turbine(s) for the given preferences, along with accompanying statistics to facilitate the final choice.

View: Poster / Movie

Electronic Pitch Trainer

AuthorsJonathan Hodrick, James Silverstein, Alexander Slocum

Advisor:  Daniel Lee


A baseball pitch has many properties that vary from pitch-to-pitch. Some of the more apparent properties are the release time, the speed, spin, and break of the ball, and the overall trajectory of the ball. The behavior of the ball path follows Newtonian projectile physics relating to velocity, trajectory, and angular velocity. The difficulty lies in the fact that not all of these attributes can be easily scrutinized. Because of the high speed at which pitches are thrown (~60 – 100 mph), it is difficult to see the variations between the spin and trajectory of a ball with the naked eye. Therefore some malformed pitches that could be attributed to the spin of the ball may not be accurately judged and corrected. Our system aims to remedy this problem.
By using object recognition software to isolate the baseball from the background during a pitch, the location of the ball is found at each frame of the recording. That information is then processed so that pertinent information, such as speed, trajectory, break, and strike zone location are relayed to the pitcher.
In addition, the statistics of each pitch are saved, and can be accessed by the pitcher in order to see how certain pitches are trending in terms of speed, break, and accuracy. The coach can also access the information in order to see how his team is trending as a whole, and isolate which pitchers have been performing well recently.

View: Poster / Movie

Detection of Pneumonia by Computer Aided Auscultation

Authors: Rachel S Egan, Ningcheng Li, Jin Ser 

Advisor: Raymond Watrous

Auscultation has long been informative about a patient’s health condition, especially in pulmonary settings. Many lung diseases, such as pneumonia and COPD, are closely correlated with specific acoustic characteristics. Recently, the advent of chest X-rays and High Resolution Computer Tomography (HRCT) scans have allowed doctors greater analysis of lung physiology and reduce the human error listening to abnormal respiratory sounds. However, these methods are expensive for patients, consume valuable hospital time and resources, and rarely serve rural or poor communities and/or countries. The time needed to conduct these tests also put patients with severe cases of pneumonia at high risks.
A cheap and fast measure for analyzing lung physiology would be to use a stethoscope to listen to patients’ lung sounds with reduced human error. Thus we propose a solution that collects lung sound recordings from an electronic stethoscope and uses digital signal processing techniques to identify pneumonia symptoms.
According to the IRB-approved clinical protocol which is currently being executed at the Hospital of University of Pennsylvania, four different listening procedures, each indicative of an aspect of pneumonia, are performed on a patient: pleural rub, peripherally heard bronchial sound, egophony and whisper pectoriloquy. The sounds are collected by a 3M Littmann Bluetooth stethoscope and analyzed by digital signal processing algorithms. A user-friendly interface is implemented for the auscultation procedure and the presentation of results. Healthcare personnel can then use the findings in conjunction with other aspects of medical diagnosis to form an appropriate and accurate medical diagnosis.

View: Poster / Movie

hermes: The OBDFree Project

Authors: James Michael Ottavi-Brannon, Evan Hyde

Advisor: Dr, Oleg Sokolsky 

As of 2008, the United States has over 255 million registered vehicles traveling over its roadways. Although automobiles provide an effective means of moving people and possessions from one location to another, they are complicated, error-prone systems. In order to increase the safety and lifespan of cars, auto manufactures have equipped them with error lights and warning messages to inform the operator of system malfunctions. However, these measures are of very little use to the average driver, providing little information about the source of the error and even less information about cost-effective solutions.
To improve the car ownership experience, individuals should be able to cheaply diagnose and resolve basic vehicle problems without having to default to a vehicle specialist. hermes: The OBD Free Project connects vehicle owners with crowd-sourced solutions to common vehicle problems. An iPhone application, wirelessly connected to a vehicle’s OBDII systems, receives real-time diagnostic information about the car’s overall health and well-being. In the event that there is a vehicle malfunction and the “Check Engine” light is illuminated, the application will query a Wikipedia-like database and provide the driver with a description and crowd-sourced solution to the problem. In addition, the system also communicates common vehicle performance metrics such as instantaneous/average MPG, instantaneous/average RPM and remaining fuel levels.
By providing higher levels of consumer access to vehicle health and well-being, customers will be able to make more cost effective decisions regarding vehicle maintenance and claim more ownership of their automobiles.

View: Poster / Movie


Sustainable Rainforest Solutions: Agent-Based Modeling of the Amazon Rainforest

Authors: Amanda E Smith, Erich R Sorger, Jane P Kim, Ivan Levcovitz

Advisor:   Barry Silverman

Spanning over 1.3 million square miles, the Brazilian Amazon contains the largest rainforest in the world. The Amazon Rainforest is home to 600,000 indigenous people, and produces over 5,000 different types of fruits and ingredients for 25 percent of the world’s pharmaceuticals. In the midst of global warming, the Amazon Rainforest plays a vital role in mitigating the rate of carbon emission by absorbing 4.8 billion tonnes of carbon dioxide per year and generating 20 percent of the world’s oxygen.
Over the last fifty years, much of the rainforest has been depleted due to agriculture and cattle farming. Changes in policy regarding the rainforest and the environment must be enforced in order to ensure the sustainability of the rainforest in the future. With about 60 percent of the Amazon located in Brazil, country-enforced policies will greatly influence the longevity of the rainforest.
To test the effects of different Brazilian government policies, an Agent-Based Modeling tool representing the Amazon Rainforest was designed to simulate the dynamic interactions between the Brazilian government, cattle ranchers, farmers, logging companies, alternative industries, non-governmental organizations, and international institutions. The model allows users to forecast the results associated with the implementation of a variety of policies, assessing both their effect on the country’s deforestation and economy. Using this model, we found that the policy that most optimally reduces the rate of deforestation, while maintaining Brazil’s economic growth, consists of a combination of increased tax rates, enhanced enforcement, and improved funding for non-governmental organizations.

View: Poster / Movie

Statistical Arbitrage Trading Model

Authors: Xiang-Li Lim, Lu Tian, Xiaobin Zhang  

Advisor:  Alejandro Ribeiro, Tony Smith

An important tool for financial traders in this technology age is effective models that can systematically assist them in making decisions. Having an effective model that can capture volatility of asset prices and generate systematic sell and buy signals will be beneficial for traders and analysts.

As many traders do not have academic or professional backgrounds in engineering, mathematics or statistics, developing a quantitative model that can capture volatility can be challenging. These target groups are also exposed to human errors, emotional distractions and over-reliance on intuitions but do not have the resources to develop a decision tool to assist them.

To provide a solution for these problems, we have developed a statistical arbitrage model that is affordable, robust and reliable.  We have employed a statistical approach using mean-reversion techniques to evaluate volatility of equity prices. Using a two-engine system, the model takes in the input of equity prices and transaction volume while generates systematic signals for users to “sell” and “buy” an asset. The model then computes a market-neutral strategy to minimize the general-market exposure that may underline the trade’s performance.

The model is back-test against a market-neutral fund. Simulated results have shown that the model gives a significant return (mean) with lower volatility and risk (standard deviation) as compared to the benchmark model. The results using over 15ETFs and 150 random stocks across 15 industries are promising and can be employed as a confirming-tool when a financial institution is considering of making a trade.

View: Poster / Movie

Phased Speaker Array for Directed Sound Control

Authors: Matthew S Byrne, Miguel A Gonzalez, Nickolaus P Woodruff, Andrew P Townley  

Advisor:  Dr. Nader Engheta & Mr. Brian Edwards

The performance of the “average” audio speaker coupled with the underlying physics of acoustic waves does not provide the necessary cancellation of noise in order to direct and focus sound spatially. This creates unwanted noise pollution that can be disruptive for people nearby. The current solution for isolating this leakage of sound is through the use of headphones or sound domes; although practical, these solutions can become cumbersome for the user. Research has been done in sound directivity through non-linear and ultrasonic techniques; however nothing has been developed for mainstream use.
We propose to tackle this problem by implementing a linear beam forming acoustic array with time delayed inputs produced through a microcontroller. When phased with respect to distance from the focal point and speaker spacing, the time delays of the outputs creates patterns of interference such that sound can be focused constructively at a certain point in space, but destructively interferes elsewhere. A 15 dB sound level difference between the focus and elsewhere is desired to achieve a difference between comfortable audio and something that isn’t distracting. This implementation provides a cost efficient alternative to current market products.

View: Poster / Movie

Wind’s Up!

Authors:  Michael Harker, Lorna Ng’eno & William Safrin

Advisor:   Peter Scott & Walt Sobkiw


Although offshore wind is a viable means to help meet renewable energy production goals, there are many hurdles to overcome before the construction of a large offshore wind project is approved. Uncertainties surround the costs of offshore power systems, the actual production a wind farm will have, and the ecological impacts from installation and life cycle processes. There exists no user-controlled model to determine the feasibility of a wind farm, providing financial analysis in an educational manner.
“Wind’s Up” is a computer application that simulates the life cycle of a user specified wind farm located on the New Jersey outer continental shelf. This model provides a stable basis for educating users on the benefits of wind farm. Users select wind farm location, turbine specifications and desired energy production. The model provides construction and life cycle analysis. The application includes data about specific wind turbines, various offshore locations, government policies, and environmental studies.
The negative opinions on wind farms held by some environmentalists and coastal communities often overshadow the financial and environmental benefits associated with offshore wind. Through the thorough financial and environmental analysis performed by this model, the user can easily understand these benefits. The application is scalable to other areas with offshore wind potential and the results of the various scenarios performed through this application not only serve as an educational tool for the public, but allow for a comprehensive evaluation of the future energy portfolio of New Jersey

View: Poster / Movie

Robotic Control of Kinematic Chain Arm-Now Do-it-Yourself (ROCK CANDY)

Authors: Juan-Antonio Macasieb, Benjamin N Plotnick, Christopher S Setian, Samuel N Oldak

Advisor:  Shai Revzen              



Robotic arms have been an integral part of science, engineering, and technology innovations and industry for over 50 years.  They have been used in industrial automation and hazardous environments such as space exploration.  In addition to these high-tech industries, secondary education organizations and hobbyists have also shown interest in obtaining robotic arms.  However, due to the high cost of commercial robotic arms, these comparatively low-budget markets have difficulty affording them. Current low-cost solutions do exist, but they typically have highly inaccurate position feedback, making precise control very difficult.

RObotic Control of a Kinematic Chain Arm: Now Do it Yourself (ROCK CANDY) is a low-cost, highly accurate robotic arm system.  The system consists of two cameras that track markers on the end of any given robotic arm combined with a feedback controller that reduces the error between current and desired position.  The control strategy is executed in the image plane to reduce computational load through a strategy called Image Based Visual Servoing (IBVS).  This strategy is tolerant to calibration errors and ROCK CANDY includes an easy-to-use calibration routine that allows the system to adapt to any robotic arm, regardless of mechanical or electrical differences.  ROCK CANDY is a significant improvement over standard open loop control in a variety of manipulator tasks such as manual control, trajectory following, and puppeting.

View: Poster/Movie

Convenience Store Ordering System

Authors: Ceren Durmaz, Rebecca Gittler, Yigit Oskay


Advisor: Peter Hahn, Ms. Monique Guignard

Independent convenience stores have difficulty with inventory management and ordering systems, compared to chain stores, which are able to keep up-to-date on technological advances. Chain stores can afford the newest systems that help them reduce costs and renew stocks efficiently. Smaller stores cannot afford commercially available inventory management systems. Furthermore, due to lack of transaction recording technology, these typically “mom and pop” stores are not able to track purchasing trends. This puts them further behind their chain-store competitors. Our team is devising an affordable alternative to commercial inventory management systems for these stores.
Convenience Store Ordering System (CSOS) is a free web-based platform for managing inventory and ordering products. CSOS updates the inventory levels of convenience stores as products enter and leave the store. Each product is assigned a par level, i.e., the point at which that product needs to be ordered. CSOS warns the user when this par level is reached for a specific product. Orders are placed to the supplier automatically or manually, depending on the user’s preference. The feedback feature of CSOS shows the trends of sales of the products. In addition, suppliers are able to track the orders through CSOS.
After testing CSOS with hypothetical and real data, certain conclusions were made. Stores have a better comprehension of their inventory and can place orders accordingly. This reduces the chance of overstocking and spoilage of certain items. Feedback helps the stores to track trends and set par levels, while the suppliers benefit from aggregation of all orders.

View: Poster / Movie

Spam Detection on Wikipedia

Authors:  Avantika Agrawal , Phillip S Baker, Brittney Exline

Advisors:  Prof. Oleg Sokolsky, Andrew West

Collaborative software models (e.g., wikis) represent an increasingly ubiquitous Web technology, due to decreasing costs of access and increasingly usable software. However, the open access that defines such systems can also facilitate malicious objectives. In particular, this paper examines the use of collaborative functionality to add inappropriate hyperlinks to destinations outside the host environment (i.e., link spam). Wikipedia, a free and open encyclopedia, is the basis for our analysis.
Recent research has exposed vulnerabilities in Wikipedia’s link spam mitigation, finding latencies in the human-editor defenses and dwindling quantities of editors. To remedy this, we propose and develop an autonomous classifier for link additions. This context presents unique challenges. For example, low barriers-to-entry invite a diversity of spam types, not just those with economic motivations. Moreover, link presentation, regardless of the destination, creates a wide variety of undesirable links.
In this system, a large set of link additions and associated attributes is gathered and used to create a predictive model. A spam corpus with 76 features is extracted from over 235,000 link additions to English Wikipedia. These features are computed using wiki metadata, landing site analysis, and external data sources. The resulting classifier attains 64% recall at 0.5% false-positives (ROC-AUC= 0.97). Such performance could enable egregious link additions to be blocked automatically with low false-positive rates, while prioritizing the remainder for human inspection.
Finally, a live implementation of the technique is being deployed on Wikipedia.

View: Poster / Movie



Stadium Express


Authors: Lindsey Eatough, Tosin Osibodu, Jordan Zarrilli & Adam Sharaff

Advisor:  John Keenan


The system of purchasing food and beverage at stadiums is slow, inefficient, and technologically outdated.  As a result, patrons are forced to stand in lines that take away from the experience of attending a live event.   Stadium Express is a system that gives users the ability to order food from their mobiles devices, which can be picked up from an expedited line at the most conveniently located vendor.  The system provides an intuitive interface for concession stands to receive orders in real time, manage the status of these orders, and dynamically update their menu.
The approach to creating Stadium Express involved interaction with patrons and stadium facility officials both at the professional and collegiate level to determine the desired capabilities of the product. To maximize the platform’s reach, the Stadium Express team created a mobile website as opposed to an application specific to one type of smart phone. While testing the product, patrons were responsive to the succinct flow of the ordering process and stadium officials were excited about the technological innovation such a system would offer. 

Stadium Express provides an ordering platform that gives event-goers and stadiums a more efficient transaction.  The system can be easily implemented into the existing infrastructure of venues both large and small.  The process will minimize the time a patron spends away from the main attraction of the stadium, while simultaneously increasing the convenience of ordering food and beverage. Concession stands will be able to serve a greater number of patrons while also improving the quality of their customer service. 


View: Poster / Movie


Improving Cross-Docking Efficiency through the Use of ITS Technology

Authors: Lisa C Zheng, Jacci K Jeffries, Ciara E Kennedy

Advisor:  Eric Bruun, Ms. Monique Guignard, Peter Hahn              


Cross-docking is a consolidation practice in logistics that facilitates the transfer and sorting of products from suppliers to distribution centers, eliminating warehouse holding, minimizing costs, and allowing for the realization of more efficient deliveries. There are several logistic and integration problems inherent in the cross-docking process. Accordingly, key issues identified at the National Retail Systems (NRS) cross-docking facility in Bergen County, NJ include scheduling uncertainty, inefficient staffing and record keeping, and limited integration of statistical data.

The project is derived from a need to improve the efficiency and transparency of the cross-docking process through the design of an Intelligent Transportation System (ITS) tailored to meet the needs of the NRS facility. The approach of the project, in accordance with systems methodology principles, consists of project definition, research of the existing system and ITS applications, the collection of statistical data, the design and comparative analysis of alternative systems, and the testing of the optimal system design.

The new system design, derived from an analysis of the paper-based and computerized communication links that comprise the daily processes at NRS, is built around a centralized database, The Efficient Dynamic Display of Incoming Exchanges (E.D.D.I.E.). The result is a modular interface that, when integrated with NRS’ existing container management system and data terminals at the freight checking stations, establishes data connections that streamline daily operations, increase transparency, and perform more accurate labor calculations. In addition, by increasing the firm’s access to archived data, this system lays the necessary groundwork for future optimization endeavors.

View: Poster/Movie

Energy Performance Rating of Penn Buildings

Authors: Saksham Karwal, Sugyan Loyiaa

Advisor:  Andrew Huemmler              



In 2007 Dr. Amy Gutmann, President of the University of Pennsylvania, signed the American College and University Presidents’ Climate Commitment (ACUPCC) and laid out the Penn Climate Action Plan, according to which Penn would develop ways to reduce its carbon footprint. Under this plan, one of the measures applied is to improve the energy performance rating of the 150 buildings that are owned and operated by Penn.

Calculating the energy performance rating of a building uses extensive data on many different aspects such as the building’s power consumption, size, location, hours of operation, number of occupants and number of electrical appliances in use. Till date, no international standard has been developed to assess the energy performance of university buildings. University buildings are unique because they have multiples building types under a single roof – labs, offices, classrooms and cafes. This project takes into account all these factors and in turn will help the Facilities and Real Estate Services (FRES) calculate the energy efficiency of Penn buildings in a uniform manner.

A detailed regression analysis has been used in this project to develop a calculation methodology specific to university buildings. The user will be able to take advantage of the same via a simple graphic user interface to input data. The program will analyze the data and provide an output with an energy performance rating for the building. Once the program has rated sufficient number of Penn buildings, there will be a ranking system to clearly identify the most and least energy efficient buildings.

View: Poster/Movie

High Speed Water Purification System for Developing Countries

Authors: Ritwik Lodhiya, Tal Bar-Or, Gaurav Gupta, Juan Jacobus, Ishan Sehgal

Advisor:  Andrew Jackson             



Access to clean water is one of the major societal and human rights issues that exists in the world today. 884 million people lack access to safe drinking water and 1.8 million die every year due to waterborne illnesses. The High Speed Water Sterilization System for Developing Countries is a low-priced stand-alone device that will be able to purify dirty water at relatively high speed, by utilizing a new technology which incorporates silver nanowires and carbon nanotubes.

Our main objective is to implement the purification system in rural areas of developing countries, where waterborne illnesses are most commonly found. The design of the filter is based on a two tier system. Firstly, the water will initially pass through the pre-treatment stage, where two water mesh filters will lower the water’s turbidity (impurities such as dirt, sediments and other unwanted organic compounds). After, the water will continue into the treatment stage, where the Silver Nanowire / Carbon Nanotube filters will inactivate (kill) the bacteria. The sterilizing filter utilizes electroporation to destroy cell walls of bacteria in order to clean the drinking water.

The main advantages the filter will possess are the low electric power consumption required for its operation and its very high filtration speed. Most self-sustained water purification systems currently employed in rural areas have an average filtration speed of between 60 and 2,000 liters of water per day. Our water purification system could potentially filter up to 7,400 liters in one day. Most other filters use materials such as clay, ceramic or stone with very small pores that physically separate the microorganisms, which makes the process very slow. This system does not separate microorganisms from the water, rather it electroporates and kills them while the water is passing through, making the water safe to drink without actually removing them from the water.

View: Poster/Movie




ESE Home Page > ESE Undergraduate Labs > Senior Design Home Page > 2010-2011 Abstracts
©University of Pennsylvania, Department of Electrical and Systems Engineering.
This site maintained and administered by Siddharth Deliwala, ( ), Last updated Apr 14, 2011