2020

SEN IQ.....,_____. ESIGN

The ECE Senior Design Capstone course serves as an opportunity

for students to execute the education they gained in the classroom

to produce prototypes for real-world clients. Student teams serve

volunteer customers drawn from industry, government, small

businesses, non-profits, schools, artists, faculty, and staff. The

course offers:

Technical, communication,

personal, and team skills

needed for successful design

in electrical and computer

engineering.

Knowledge of specifications

and standards, information

collection, design strategies,

modeling, computer-aided

design, optimization, system

design, failure, reliability, and

human factors.

Proficiency in oral and written

communication of technical

information.

Understanding of team

dynamics and ethical issues in

design.

Experience in completing a

design project for a small­

scale electrical or computer

system.

ECE DAY AWARDS

Best ECE Senior Design Project Award

Design Excellence Award

Michael F. Ruane Award for Excellence in Senior Capstone Design

Entrepreneurial Award

Team 1 - Cube Sat

Members: Dylan Nash, Karle Erf, Thuc Nguyen, Matthew Chamieh (ME), Cameron Winslow (ME)

Client: Professor Brian Walsh

The surface of the Moon and its atmosphere is constantly changing due to meteor strikes and past scientific lunar missions. Researchers at Boston University’s Center for Space Physics have been working on ground-based imaging systems that can view the soil and dust content of the Moon’s atmosphere using spectrometry to see how it has changed from the different interactions with the lunar surface. Our team is attempting to adapt such an imager for a cubesat. Being able to use this imaging system in space on a cubesat orbiting the moon would allow the researchers at CSP to collect more meaningful data about the dust composition of the lunar atmosphere. We hope that successful development of this imager system will aid CSP in producing a fully realized cubesat that can be launched into space to conduct this science mission.

Team 2 - Cyber TapMembers: Felipe Dale Figeman, Alex Fatyga, Evan Lang, Noah Malhi,

Justin Morgan

Client: Cybereason

CyberTap is a high throughput hardware network tap designed for industrial control systems closed off from the internet. Other network monitoring solutions are software based with overutilized CPUs which could theoretically result in packet loss. Packet loss is particularly undesirable in the context of the high data throughput involved in industrial control systems, especially as said systems directly rely on input data for operation. CyberTap will be able to collect Operational Technology (OT) network packets, parse and generate metadata for all relevant network protocols of a system, and store the collected packets and generated metadata in persistent storage. The product will be implemented on a Field Programmable Gate Array (FPGA) to utilize their ability to quickly process large data loads. The final deliverable will contain the following: An emulated high-data-volume OT network using 2 Raspberry Pis and a network switch, an FPGA that sniffs, parses, and outputs the packet data coming from the SPAN port of the switch, a solid state drive that contains the outputted metadata, and a web application for querying the drive and displaying the metadata.

Team 3 - Safe IoT

Members: Yahia Bakour, Chase Clarke, Ryan Ewing, Nicolas Hunt,David Li

Client: Professor David Starobinski

Our main goal is to develop a complete, innovative solution to automate the security testing of wireless IoT devices. This will be the first handheld IoT fuzzing solution on the market. The software and hardware implemented will act as a toolkit that will fuzz several wireless protocols including WiFi, Bluetooth, ZigBee, and Thread. This software will provide a web interface to interact with the system and guide the testing. Additionally, the software will be capable of exporting any fuzzing results to external, cloud-based database and file-systems for further vulnerability analysis and processing.

Team 4 - Life RobotMembers: Lin Fan, Hung-Ching Lin, Kaihui Gou, Apollo Lo,

Benjamin Wong, Tai Kang (ME)

Client: Professor Alan Pisano

In the past, rescue workers during a severe earthquake or natural disaster have to search on-site using hand-held sensors to locate survivors. They would utilize a blanket search technique in order to cover the entire area. This method is highly risky, inefficient, and costs valuable human resources. Currently, there lacks a method that can cover large area space and stream live information from the disaster area. Real time data regarding the effects of the disaster is crucial to determining possible locations of trapped victims. Life Robot aims to reduce rescue time and increase efficiency of search methods by utilizing drones and mini robots (containing IoT sensors). These devices will be used to survey the surface for structural conditions and monitor the activities underground. The solution proposed by the design team is a network of portable sensors deployed by drones that can work alongside rescue workers. When disaster strikes, rescue teams can send out drones to distribute mini robots and scatter them throughout the disaster site. The sensor data transmitted back to the rescue command center through an established network can be analyzed to plan rescue efforts. Each mini robot can identify life signals, detect fire hazard and volatile gas leakage, and monitor structural stability around the region. Drones and mini robots can replace humans to make searching more effective in search and rescue missions. They can work uninterrupted day or night and always maintain an objective observation, which will greatly improve search efficiency and rescue personnel efficiency to rescue as many lives as possible.

Team 5 - Huzzah

Members: Elizabeth Slade, Rubeena Zaveri, Laura Joy Erb, Natasha Frumkin, Gennifer Norman

Client: Emily Lam and Professor Thomas Little

What is the nature of the problem? Detecting people in a room non-invasively.

What will be your final deliverable? Indoor positioning system that incorporates Pucks, Beacons, the cloud, and a visualization component.

What is your proposed technical approach? We first focused on full system integration with two pucks, where we sent temperature and acceleration data to a database and displayed it on an application. After we got the full system functioning, our next steps are to get more granular with each component. For example, our next goal for the hardware is to increase the number of pucks and implement trilateration. Similarly, our next goal for the machine learning component is to get a functioning algorithm that can accurately detect a human on a chair.

Team 6 - CUPSMembers: Yifan Wang, Ryan McKeever, Faisal Alhuqil, Faisal Al-Bawardy,

Faris Aljamed

Client: Professor Michael Hirsch

Team 6 is determined to develop a user-friendly, innovative parking system and app that can ease the frustrations associated with obtaining parking in urban garages. The user can utilize our C.U.P.S (Computer-assisted Unmanned Parking System) app to search for available parking garages and open spots as well as sign up for an autonomous payment program. In addition to this, C.U.P.S aims to use multiple ultrasonic sensor arrays to determine what specific parking spots in a specific garage are available and provide this information to the user. Lastly, Team 6 will also utilize an image processing“smart gate” to automatically allow registered users into the garage by reading theirs license plate and connecting it to an account all while charging payment for parking .C.U.P.S aims to rescue customers from the frustration of endlessly circling parking garages for a free spot and save their time.

Team 7 - Blitz (Store Map)Members: Cooper Salmon, John Gillig, Kyle Martin, Lindsey Volk, Mariano

Pache

Client: Self-defined project

The purpose of Blitz is to create a total shopping platform to simplify managing and shopping at a grocery store. Store managers and customers can both benefit from Blitz’s automated processes to aid in everyday tasks. For management, tracking inventory is extremely important and with Blitz, inventory is updated when an item is taken off or put on a shelf. Management can know and respond immediately, before too much time has passed waiting for customers to check out, if an item is low or out of stock. In addition, when items are moved or there are promotions, employees will not have to change the item stickers on the dedicated shelves themselves. To help the customer be more efficient, they can use our mobile application to create a shopping list and receive a populated, shortest ¬path route through their selected grocery store. Instead of wasting time, customers can know exactly where to go in the store to find exactly what they need in the shortest time possible. Overall, we think that shopping can be more efficient with Blitz in terms of time and effort.

Team 8 - Swim Coach Members: David Abadi, Frederick Akiki, Kenwood Harris, Javier Roberts,

Ricardo Villarreal, Juan Zapata Gomez

Client: Self-defined Project

For amateur, intermediate, and advanced swimmers, obtaining personalized training at an affordable price that fits their schedule has become impossible in today’s world. This lack of feedback has created a performance barrier to thousands of swimmers around the world. In addition, because of this lack of professional training, there have been hundreds of cases where long-term shoulder, back, and even leg injuries occur. With this problem in mind, we have created SMARTSWIM : a platform that evaluates swimmer’s technique using computer vision and machine learning algorithms, consequently returning feedback on how swimmers should improve their technique. Our main customers are pool facilities, preferably those who train dedicated intermediate swimmers. Because of this, our deliverable consists of first mounting a GoPro camera above a swimming pool that is connected to a Jetson embedded processor. Then, members of these pool facilities will be able to go into our web application, log-in, look at their swimming video footage, their stats, and receive feedback from machine learning software that compares their swimming with that of professionals. We are proud to say that with these innovative features, we will be first to market in helping swimmers improve their technique.

Plastic pollution in the ocean is a pressing environmental issue our society faces today, and 90 percent of the ocean plastics originate from 10 river systems. We aim to collect floating plastics on rivers by using an autonomous boat named Trident. A single Trident vessel will be capable of navigating and clearing a section of water without human intervention. With multiple Trident vessels communicating with each other, large-scale cleaning of river systems will be possible. Our senior design project will be a proof of concept with one fully functioning Trident vessel collecting floating plastics in a small section of the Charles River. While beyond the scope of this course, the project’s long-term goal is to have a swarm robotics platform allowing many trident vessels to work in a collaborative, optimized fashion for river cleaning.

Team 9 -TridentMembers: Jacob Davis, Alex Hammerman, Ryan Hoffman, (Jack)

Hanchong Jin, Varun Malikayil

Client: Professor Alan Pisano

Team 10 - BooKBoT Members: Isabella Kuhl, Drew Boodhan, Ashley Hamlin, Victoria

Nowrangilall

Client: Self-Defined Project

BookBot is an autonomous library inventory management system designed to track and recover missing books. The combination of an affordable, modular traversal system and a highly accurate image processing algorithm allows BookBot to be installed in any library, regardless of layout or inventory management system. BookBot first scans and compares the books on each shelf to the library’s database. If a book is found to be miss shelved or missing, a notification with the location and call number is immediately sent to a librarian who can then either retrieve and re-shelve the book or report the item as missing. Using a seamless web server to track and flag missing books, BookBot ensures inventory is organized, maintained, and accounted for.

Team 11 - Wearable Members: Neil Patel, David Kirk, Brian Xu, Donovan Jones, Ismail Mohamed

Client: TymeWear

TymeWear is a start-up located in South Boston developing an IoT Athletic Smart Shirt. This shirt is a polyester based tank-top style shirt equipped with a microprocessor and an array of sensors in order to inform the user and keep his/her biometric ratings in a safe space. Unfortunately, TymeWear outsources the respiratory rate module in their product to another company, wasting power consumption and money. In addition, development is at a halt due to the proprietary nature of the outsourced module’s firmware code. That is where us, Team 11 - Wearable, at BU Senior Design come in!Our team will be developing a solution in order to phase out the proprietary outsourced nonsense and adopt all of its function into the native microprocessor and code of the device. We start our journey by doing a top - down analysis of the product’s initial form, quantifying the behavior andcommunication between desired features and “to be phased-out” modules.Afterwards, we start development in an independent microprocessorsystem, eventually leading to firmware integration on the native product’ssystem of implementation.

Team 12 - Tiny Town Members: Jack Belmont, Jinhyun Park, Shaurya Atal, Linda Cordero,

Yaying Zheng

Client: Marissa Petersile

Electrical consumers have started to utilize personal distributed power sources (such as solar panels) to power their homes. This, paired with the lack of an efficient community style energy marketplace, puts the modern electrical grid in an unstable state. Our team is designing a model grid system that will be proof of concept for a novel grid system that solves the major flaws with current electrical power trends. The TinyTown solution will utilize DC power along with a redesigned power line layout to reduce the reliance on phase matching and increase the efficiency of outage management. A centralized processor will maintain a ledger for the power flow around the grid, along with providing self balancing power outage features through the use of relays.

Team 13 - Speech Interactive Therapy Application - SITA

Members: Zhe Deng, Cindhuja Sramasamy, Jennifer Norell, Diana Deleon, Swathi Jaisankar

Client: Andrey Vyshedskiy

Our team is determined to make a speech interactive therapy application to help young, nonverbal children with autism spectrum disorder (ASD). The game begins with videos of a speech pathologist pronouncing phonemes. As the child’s skills progress, the game will play more complex and advanced levels, up to sentence-level exercises. Parents or guardians will also have the choice to record their own video levels. For every video, the child will receive 2D visual rewards based on an accuracy score. A mix of envelope matching and our client’s own child vocal isolation algorithm will give us an accuracy function tailored for young, nonverbal learners. All of SITA’s components will be built in Unity 3D and deployed to iOS and Android. Our goal is to provide a very simple, yet dynamic therapy environment that gradually builds necessary speech skills.

Team 14 - Situation-Aware Stoplights

Members: Ashley Cui, Sarina Sabharwal Joshua Bassin, Yuhang Wu, Rachid Tak Tak

Client: Asher McGuffin

The average Bostonian driver loses over 160 hours to traffic congestion at intersections every year. The fixed light cycle used by modern stop lights cannot cope with the change in throughput of cars throughout the day, leading to unnecessarily long wait times during off-peak hours. We propose a system to alleviate this issue: a real-time reaction system that allows the light cycle times to be modified on the fly to adapt to current conditions at the intersection. We aim to produce a camera that is capable of detecting the state of the intersection (if there are cars in the intersection and if they intend to turn left) and ultimately use that information make a dynamic decision to change lights.

Members: Nouf Alamri, Levi Emery, Lorenzo Kinnicutt, Erin Thomas, Siyuan Zhang

Client: Professor Malay Mazumder

Routine cleaning and maintenance of large-scale solar plants requires vast amounts of water and labor; this is costly and unsustainable in many regions where solar farms are abundant. Self-cleaning photovoltaic modules are being developed at Boston University to solve this issue through a pioneering technology known as an electrodynamic screen film, which can be retrofitted onto solar cells to demonstrate water-free, autonomous cleaning. Currently, this film has not yet been integrated into a larger control system for deployment and commercial applications. The Clean Solar team will be addressing this by designing and constructing a photovoltaic module with an integrated EDS and control system. The primary approach is to compare hypothetical short circuit current with detected short circuit current of the solar cell and activate the EDS given certain criteria. Upon culmination of this project, the deliverables will be as follows: (1) A completely assembled EDS-retrofitted PV with control and monitoring system and power supply, (2) Successful demonstration of the prototype device, and (3) An instruction manual of the prototype unit for its third-party operation in a solar field. Special innovative features include data storage, a web application, and a heating and cooling system.

Team 15 - Clean Solar

Team 16 - Chameleon Members: Rene Colato, Sean Brandenberg, Lal Senvurt, Skyler David

Client: Professors Trachtenberg and Ishwar

In the current day, many internet users are becoming increasingly concerned with their lack of control over how their data is used. Many internet companies have a business model that relies on users being willing to give up their privacy in exchange for access to convenient tools such as Google Docs. This relationship leaves users vulnerable to have their data sold or used for things like targeted marketing. Because all of this data is stored by these companies, a data breach could result in the leaking of this data to malicious actors.

The goal of this project is to give users of real time collaborative environments control of their data privacy. Currently, users are encouraged to collaborate on servers like Google Docs, but Google has full access to any data produced using the platform. There are currently no means for users to control whether their data is sold to third parties, or scrapped for valuable information (like a list of phone numbers made on Google Sheets). To give users back some of the leverage in this asymmetric relationship, Chameleon aims to build a plugin that encrypts this data while still giving the user full use of these platforms.

Team 17 - Poolwatch

Members: Yuhang He, Kenza Bensouda, Megan Heydari, Jade Stoffey, Emeka Opara

Client: Professor Manuel Egele

Over 300 children in the US drown in private pools every year. Parents must diligently watch young children at all times to make sure they are not in danger of falling or jumping in the pool. PoolWatch will ensure that accidental drowning does not go unnoticed by alerting a guardian. PoolWatch is a wireless, wearable device and alarm system, made for children who are not ready to swim. An accessible, easy to use, and quick to respond solution will save lives and reduce stress around swimming for parents.

One of our innovative features is our mobile app for individuals with smartphones to set up and stay connected. Our system reliably reduces false negatives so that a real emergency will not be overlooked. By relying on low energy consumption, PoolWatch will last the duration of an entire pool party.

Team 18 - CIRapp

Members: Chenjie (Lavi) Zhao, Alexander Salmi, Jiayue (Cindy) Bai, Scott Hom, Chin-Hua (Tiffany) Yen

Client: Professor Janusz Konrad, Professor Jordana Muroff, Mr. Ozan Tezcan

The goal of Clutter Image Rating (CIR) Application is to help improve the level of care provided by support providers to those who have health conditions with negative effects that manifest themselves as an excessively cluttered home. The core providers, as mentioned by our clients, are clinicians, house professional, and first responders. The level of care is improved by eliminating manual CIR Scoring which is labor intensive and has human bias. Both problems are addressed by automating the process with a deep learning algorithm that automatically rates pictures and outputs a CIR Score. The algorithm is a Residual Network (Res Net) 18 Convolutional Neural Network (CNN)developed by the PhD student Ozan Tezcan [4]. The early stage of the CIR App senior design project is to improve the algorithm accuracy by collecting more data to make the training data set more representative of the full range of cluttered rooms on the CIR Scale. The app is being developed for iOS and Android using the Expo and React Native to put the algorithm in the hands of as many people as possible. The initial stages of public testing will be with a few public housing professionals. The final application is intended to be released to the public and the patients’ data will be protected under HIPAA. The encryption features for part of HIPPA compliances which are TLS encryption for data transfer and FIPS compliant encryption for data storage on SQL server [6][7]

Team 19 - Mars Rover ArmMembers: Yu Lee, Kevin Arseneault, Kathryn Lewis, Hairuo Sun, Junwei

Zhou

Client: Professor Osama AlShaykh

As technology advances, many scientists have encountered problems in collecting research-based samples in hostile or dangerous environments. Thus, they are in need of modern robotics that are capable of performing hand-like operations in order to remove the need for human intervention. On conditions like that of Mars, a robotic arm attached to a rover is needed to help solve this problem. Our final deliverable is a robotic arm capable of lifting objects, grasping handles, tightening screws and opening or closing drawers. Our proposed technical approach is to use an Arduino as our central microcontroller connected to PS4 controller, which can operate linear actuators and servo motors. A scientist would be able to wirelessly control the functions of our arm and collect the data they require in harsh environments.

Team 20 - Low-Power Embedded Fisheye Camera for Deep Computer

Vision - SNELLensMembers: George Sorescu, Vanessa Schuweh, Vindhya Kuchibhotla, Josh Rubin

Client: Professors Janusz Konrad, Prakash Ishwar and Thomas Little

Understanding where and how many people are present in various parts of a building can be applied in minimizing energy waste (through optimized HVAC and lighting control), for emergency response, and for space management. Our team is working to achieve the goal of determining room occupancy by building a compact, self-contained system which will provide the locations and counts of persons within a room based on a mounted, overhead view. Furthermore, the device will be able to acquire images in low-light conditions for optimal viewing and processing, enter/exit sleep-mode upon programmed conditions to conserve energy, and communicate all data securely.

We plan to design, build, and test a low-power, integrated system that will consist of an embedded processor to run an adapted deep-learning algorithm that processes images from a fisheye camera with 360°x220° field of view and of high resolution. Our client’s proprietary YOLO-fisheye algorithm is designed to be executed on a desktop GPU, so our team will work to port and execute this algorithm on an embedded architecture. We will also develop a user interface so that the end-user may not only set up system parameters like detection/standby durations, removal of the IR-cut filter, IR illumination, and the human detection thresholds, but also in order to view the current and prior counts of people in the room. The final deliverable will consist of a robust, aesthetic system that will house our hardware components and processors, and a graphical user interface that can communicate with the system.

Team 21 - IoT Kitchen assistant for meal & recipe preparation

Members: Yuran Shi, Erin Dorsey, Kenny Zheng, Addison Dolido, Saransh Kothari

Client: Annie Lane

Many amateur chefs often face difficulty when trying to follow along with recipes and updating recipes to their preference in real time. In addition, there are many existing technologies to aid in cooking but none that concentrate them all into an easy to use hub. IoT Kitchen seeks to solve these issues as the ultimate smart cooking assistant for amateur home chefs. IoT Kitchen provides hands free assistance so that chefs can follow recipes as well as update recipes to their own preferences in real time. This is achieved with the help of a user-friendly app interface, Natural Language Understanding (NLU) assistant, as well as several IoT sensors including a scale, thermometer, and measuring cups. The Android app interface will allow users to keep a record of their recipes made on their profile page, and find new recipes by viewing other users’ profiles. If a user fails to find inspiration from their friends, they can search our curated recipe database with strategic recipe keywords such as ingredient or dietary restriction. When a user finds a recipe, they can initialize auditory recipe instructions through the NLU assistant. This assistant will help a user step through a recipe or jump to a specific ingredient or instruction. If the user comes across an element of the recipe that they’d like to change, they can vocalize this change to the NLU assistant, which will send that information back to the recipe database and app. Throughout cooking, the IoT sensors will track cooking time, temperature, and ingredients added, providing feedback to the app to update the recipe and to correct a user if they are following the recipe wrong. Combined, IoT Kitchen will make following along to recipe and cultivating new ones easier than ever before.

Team 22 - SmartOilMembers: Zachary Bachrach, Brian Butler, Gerard Devlin, Christopher

Jung, Zachary Wollman

Client: Professor Alan Pisano

While there has been a surge in development for smart meters that monitor electric energy consumption in residential applications, the movement has not yet begun to address needs for customers who heat with fuel oil. Their systems are often older and difficult to keep tabs on; oil consumption, an important parameter, is either tracked using sight-based mechanical gauges on oil tanks or estimated with private algorithms that fuel companies use to schedule deliveries. There are some automated tank gauges on the market that allow users to see their current fuel usage rates and levels on their smartphone, but such devices ignore the rest of the system beyond the oil tank. Our project aims to develop a full-home smart system that can offer users the information needed to monitor the performance of their heating equipment without complicated installation. It will encompass a temperature sensor network throughout the home, an oil level sensor on the fuel tank, and a central hub that receives and uploads all recorded data to one location. Customers will be able to interact with the many devices and to learn about their system – in terms of remaining fuel, home efficiencies, costs, and more – on readable graphs shown in our mobile application. The interface will report the results obtained from our developed software, taking the raw sensor measurements and converting them to meaningful metrics relevant to oil heating. Our solution will do more than just provide a history for their system; a specialized machine learning program will analyze the data in real time to alert users to future fuel costs as well as to changes that could reduce such costs if made in their home.

Team 23 - BirdWatchMembers: Zach Carlson, Yi Wei Chen, Joe Paller, Gordon Wallace, Rachel

Yang

Client: Ned Utzig

Bird populations have been rapidly declining since 1970, and conservation societies such as Massachusetts Audubon Society urgently needs to track bird numbers more than ever. In order to help tackle this issue, our team has designed the BirdWatch, an automated bird observatory. The main technical components of BirdWatch are a camera that will automatically take pictures of birds that land at a bird feeder, an image processing algorithm that will identify the bird species, and a user interface that will provide meaningful data to researchers. Using our devices, researchers will be able to quickly collect large, accurate datasets as well as saving time and manual labor.

Team 24 - Community Managed Security System

Members: Samuel Evans, Esteban Hernandez, Daniel-John Morel, Steve Numata, Cameron MacDonald Surman

Client: Dan Ryan and VergeSense

With recent advances in machine learning technologies, there is a growing concern over video data tampering. Currently, most privatized security systems are owned and operated by third-party companies with little accountability. Our project aims to get rid of third party involvement and provide accountable security through a community audible security. The final deliverable will be a video surveillance system that can be accessed through a web-based interface. The website will allow the user to view the live stream of the cameras and the recorded video data uploaded to the cloud storage. All actions taken by the users account will be recorded in a log that can be viewed by all users. This will ensure the privacy and security of the community.

Team 25 - AR Tour

Members: Jessica Barry, Adiran Martinez, Quianna Mortimer, Sebastien Nevarez, Edwin Sun

Client: Professor Thomas Little

This exciting and innovative project is an Augmented Reality Tour of BU College of Engineering campus made available at anyone’s fingertips. With a touch of your smartphone and a little hardware help, guests can come and experience BU in a whole new way. Our product uses innovative methods and technologies to add more to the budding world of AR to produce a new, exciting, and customizable experience. The AR App which can be downloaded by students on their phones will guide them around the engineering campus. Location technologies such as GPS, BLE beacons and NFC will communicate with the App and guide the user. It is a chance to see BU through new, innovative eyes.

Team 26: Audio Watermark Authentication

Members: Andrea Farina Fernandez Portillo, Dennis Harrison, Harshavardhan Ogoti, Aixa Davila, Aminata Seck

Client: Cambridge Consultants

Our group is tasked with making 2FA more user friendly through audio watermarking. There will be no need to memorize passcodes, wait for authentication codes to be sent to you via email or text message, or type digits to authenticate; rather, this form of 2FA will be something you have in addition to something you know (i.e. password) and conveniently applied in the form of audio. Our final deliverable will include a cross platform phone application, a backend server that records the response from that application, and a database to store logistical login information. For our code generation we will be implementing TOTP, a Time-based One Time Password, which has specifically been requested by our client. For our audio watermarking technique, we plan to embed the code using a combination of signal transformation techniques that include amplitude shifts, phase shifts, frequency shifts, or fast Fourier transform. Our innovative feature is inserting a dynamically generated code with an expiration time into a given audio file as an audio watermark. Our competitors SoundLogin and Sound-Proof have similar approaches but between using dial tone frequency or checking of ambient sound on two devices, our approach is far more robust and secure.

Team 27 - AI Admissions

Members: Ahmed Albahar, Aman Gupta, Paschal Igusti, Varun Lalwani

Client: Professor Ari Trachtenberg

Boston University’s graduate admission committee requires a recommendation system that will catalyze the extensive admissions process. The proposed Artificial Intelligence (AI) enabled system will autonomously analyze scanned transcript PDFs and provide a recommendation score for the applicants based solely on their transcript’s academic quotient, thereby allowing the graduate team to examine prospective students faster. The data to train an AI algorithm to accomplish this task consists of highly sensitive information and is highly volatile in terms of input quality itself. Hence, our system utilizes computer vision to extract text from the input transcripts; followed by, natural language processing models on the extracted text to ensure privacy by using a data anonymization model innovated by our team. Ultimately, we fit a support vector regression model to predict the recommendation score.

Team 28 - Mars Rover: Autonomous Navigation

Members: Daniel Crawley, Brian He, Tommy Lam, SeungYeun Lee, Linden Vo

Client: Professor Osama AlShaykh

Our goal for this project is to design, build, integrate, and test a rover capable of autonomous movement to a specified AprilTag with the use of GPS, visual and inertial navigation. We have been following guidelines outlined by the University Mars Rover Challenge. The system is a prototype meant for use on Earth, but can be designed to be used on Mars. While adhering to size, weight, power, budget, and environmental constraints, the design of the rover used an Intel RealSense T265 tracking camera and D415 depth camera to generate a map of its surroundings, detect obstacles and simultaneously navigate around them to acquire the AprilTag target. The entire software stack used ROS nodes with different topics subscribing and publishing to each other. While much of the final prototype testing was in progress, the team accomplished many of its goals in the final semester and left the project in a state where another group would be able to come in, replicate the results, and complete the task of autonomous navigation.

ADVISOR: PROFESSOR AYSE COSKUN

Human-in-the-Loop Experimentation Platform for Cloud-native Applications

In modern cloud-native application development, the progressive movement to-wards microservice architectural designs and continuous experimentation has brought new challenges. The development process usually follows a continuous integra-tion and continuous delivery (CICD) practice, where new changes to the software are deployed several times a day. These changes are then represented by a new version of a microservice, where it would be deployed alongside the base version for online experimentation. The experimentation process allows a subset of users to interact with the new version of the microservice, which can lead to the early detection of potential problems. Given the competing versions of a microservice, however, there is difficulty in balancing the efficient traffic routing towards the optimal version while continuously collecting experiment data points. This thesis project consists of the research and development of a novel human-in-the-loop experimental plugin for cloud-native applications using Kubemetes. The first objective is the simplification of the experimentation process by provid-ing a dynamic experiment creation interface. Another objective is to allow the developers to involve themselves in the traffic shift decisions during the on-going experiments, assuming that the developer operating the experiment is the most knowledgeable. After every iteration of the conducted experiment, this developed plugin will provide developers with a summary of performance attributes and traffic split suggestions that minimize operational risks. The decision to execute a different traffic split or to apply the algorithmic suggestion would be completely up to the developer. Through the inclusion of these new features, the experimental platform will serve as a point of reference for developers intending to optimize their software rollouts.

AVERY HUANGTHESIS STUDENT

THESIS STUDENTZhiyuan (Ryan) RuanADVISOR: PROFESSOR RICHARD WEST

FlyOS: Mix-Criticality UAV with Quest-V

Drones are becoming increasingly ubiquitous due to the lowering costs, sizes, and weights of computational hardware. They are becoming a viable option for applications such as package delivery, search and rescue, surveillance, building inspection, etc. The current paradigm for the control of these drones is to use a STM32-based board with a more powerful board, where STM32 runs the flight controller while the other board carries out additional computational tasks. Those tasks can potentially be autonomous, re-configurable missions. While using two boards ensures that the mission computation will not interfere with the timing and predictability of the flight controller, the setup is not ideal. As most STM32 based boards are not designed with high speed buses in mind, the communication speed between two boards is greatly limited. FlyOS explores how to use a single computational platform to host both the flight controller and the mission computation. This work utilizes a lightweight partitioning hypervisor, Quest-V, to ensure the isolation of the flight controller tasks from the mission tasks. FlyOS has been implemented on the Intel Aero board, without requiring additional computational platforms. We perform evaluation of our platform by comparing the latency of our solution to an STM32-based platform. And we have implemented an object tracking mission to showcase the flexibility of our platform.

THESIS STUDENTHIROKI KAWAI

ADVISOR: PROFESSOR RICHARD BROWER

Lattice Gauge Theory on a Noisy Quantum ComputerThe computational resources and time complexity for the exact simulations of quantum systems grow exponentially as the size of the system increases due to the complicated interactions. Hence, interesting problems in lattice gauge theories and many-body dynamics require a huge amount of computational resources or are even incapable of being simulated in a realistic time. Computers whose information units process in a quantum mechanical manner, so-called quantum computers can resolve this issue. Since no fault-tolerant universal digital quantum computer is realized yet, and we are currently in the very early stage of the noisy intermediate-scale quantum (NISQ) device era, the complexity of quantum algorithms that may be performed is quite limited. Optimizing the circuit and mitigating the noises for each specific application are critical to maximizing the capability of the noisy devices. In this thesis, I construct a quantum circuit for the quantum link model for lattice gauge theories which is an alternative of the Wilson's lattice QCD and originally formulated for U(N) and SU(2) gauge theories and later extended to SU(N) and more generalized as so-called D-theory. In particular, I am interested in the simplest U(1) QED in 2+1 dimensions on a single triangle plaquette lattice which is formulated originally with fermionic creation and annihilation operators and is converted into the spin basis, and simulate its real-time evolution via the Suzuki-Trotter expansion on a quantum circuit. The constructed circuit is tested on digital quantum hardware called IBM Q with specific optimization for the target hardware. I also add supplemental gate operations to mitigate noises which are inevitable during the gate operations on the noisy quantum device by empirically modeling the noises specifically for the circuit. The experimental results demonstrate that the mitigation technique dramatically reduces the error in the measurement distribution of the output state so that the time-evolution simulation gives useful insights into the transition of the state.

converted into the spin basis, and simulate its real-time evolution via the Suzuki-Trotter expansion on a quantum circuit. The constructed circuit is tested on digital quantum hardware called IBM Q with specific optimization for the target hardware. I also add supplemental gate operations to mitigate noises which are inevitable during the gate operations on the noisy quantum device by empirically modeling the noises specifically for the circuit. The experimental results demonstrate that the mitigation technique dramatically reduces the error in the measurement distribution of the output state so that the time-evolution simulation gives useful insights into the transition of the state.

ALAN PISANO Associate Professor of the Practice

Dr. Alan Pisano received a Ph.D. in electrical engineering

from Northeastern University in 1974. He retired from

General Electric in January 2010 after a 39 year career

there in both Power Systems and most recently Aircraft

Engines. There, he was responsible for numerous

advanced controls technology programs and held a variety

of managerial positions including Manager of Turboshaft/

Turboprop Controls and Manager of Advanced Controls

Technology and Planning. After retiring from GE as a

Department Staff Engineer, he was appointed to the full­

time faculty in the ECE Department at Boston University

as Associate Professor of the Practice. He is currently the

lead professor and course coordinator of the capstone

Senior Design course in ECE and also regularly teaches

courses in control systems and electric energy.

OSAMA ALSHAYKH Lecturer & Asst. Research Professor

Dr. Alshaykh is CEO of Nx T ec. He was CTO of Packetvideo

corporation, Scientist at Rockwell and Visiting Researcher

at UC, Berkeley. Osama received a Ph.D in Electrical

and Computer Engineering from Georgia Institute of

Technology in 1996. Osama received Fulbright Scholarship

and served as associate editor for IEEE Transactions on

Circuits and Systems, Video Technology. He served

as a consultant, board member and advisor for several

companies and groups.

MICHAEL HIRSCH Research Scientist

Dr. Hirsch is President of SciVision, Inc .. He consults and

advises companies and institutions in remote sensing

systems, rapidly going from ideation to Series A, market or

deployment. He fuses real-time heterogeneous networks

of GPS, radar and optical sensors, breaking through legacy

spatiotemporal sensing limits.

Iris Celeb

Shuaibin Chang

Dorde Gluhovic

Meghan LeMay

Jianing Liu

-0

-0

--()

0 0

0 "

0

ft d