Electrical and Computer Engineering

• ECE-01 Medical IoT – AIRwAI

  Rohan Patel, Sima Noorani, Christina Strobel, Reilly Shaffer
  Advisor: Dr. Kapil Dandekar and Malvin Nkomo

  Currently, ventilators are an essential part of treatment for patients who are unable to breathe on their own. Ventilators are often interfaced to the patient via a mask or an intubated tube that is inserted into the trachea. The ventilator's breathing cycle is initiated by detecting breath, known as triggering. When adult patients initiate a breath, modern ventilators detect changes in air flow or air pressure through sensors located in the intubated tube to trigger ventilation. However, in neonatal, or newborn, patients inserting an endotracheal tube is associated with higher risks of developing chronic lung disease. The standard methods of detection such as air flow or pressure sensing are ineffective in non- invasive ventilation such as nasal intermittent positive pressure ventilation (NIPPV). Noninvasive triggering methods do already exist, such as the Graseby Capsule used on the InfantStar 950 ventilator, but they rely on a cumbersome number of tubes and electrical sensors. This paper will discuss a novel, wireless, and non-invasive method of triggering to improve neonatal patient-ventilator synchronicity. An open-source emergency ventilator was constructed, and a custom microprocessor-based controller was developed to drive the ventilator. We will be utilizing the Drexel University developed IoT (Internet of Things) platform, VarIoT, and a fabric-based RFID (Radio Frequency Identification) antenna to track respiratory behavior and transmit information to a predictive algorithm. The final ventilator system can send as well as receive information through the IoT.

• ECE-02 Raspberry Pi Nuclear Reactor Simulator

  Joey Arnold, Tristan Marshall, Matthew Czerwinski, Casey Adams
  Advisor: Dr. Christopher Peters

  Nuclear power generation will play a key role in inhibiting the steady growth of carbon emissions in the upcoming years. With the recent advancements made in the nuclear industry such as the development of Small Modulator Reactors (SMRs) nuclear power is starting to receive more attention world-wide. With this resurgence of interest in nuclear power, it is essential to be able to educate students and non-nuclear reactor operator employees on the functionality of a nuclear reactor. This will ensure that those involved with the topic have a thorough understanding. Currently, educators use software to show the effects of a nuclear reactor, and the primary hardware models are costly, complex, and reserved for training personnel directly involved with nuclear reactor operations. While this software-based approach does manage to help educate, the lack of physical systems can inhibit certain concepts from being understood completely. The proposed method enables a more immersive and accessible learning atmosphere by incorporating a combined hardware and software model to create a nuclear reactor simulation program. The software models a homogenous cylindrical Pressure Water Reactor (PWR). This system is set up to be used remotely or locally with a Raspberry Pi and an Arduino, allowing the user to log-in to the simulation program hosted on a web browser GUI called Node-RED and control the software and hardware functionalities of the reactor.

• ECE-03 Classifying Patents and Automated Routing

  Jonathan Yu, Jordan Franklin, Catherine Nguyen, William Stewart, Connor Strohecker,
  Advisor: Dr. John Walsh

  Every patent application arriving at the United States Patent and Trademark Office (USPTO) must be classified into a subject category within the Cooperative Patent Classification (CPC) system. Patents are assigned to a maximum of 13 classifications selected from more than 250,000 possible classifications. Presently, classification assignment is a multi-step process where USPTO subcontractors make an initial recommendation that gets either changed or confirmed by USPTO examiners. This subject classification assignment ensures that the patent is routed to an examiner with extensive experience in that specific subject. However, this process is labor intensive. It is not only expensive, but also has poor scalability, which degrades results as patent applications increase. The USPTO could improve its patent examination quality and speed if the classification process was improved to become more economical and scalable. Our patent classification software aims to improve classification speed and reduce cost through automation. Our system leverages state-of-the-art Natural Language Processing, text recognition machine learning models, and other techniques to match classifications to patents and route patents to examiners. Our approach addresses issues such as a patent having more than one classification and is flexible enough to update with the changing CPC schema. With the help of an extensive database, these predicted classifications can then be leveraged to recommend the best examiners based on user-configurable criteria. Authenticated users can use both algorithms delivered through the web; each interaction is logged for performance tracking.

• ECE-04 Plagiarism Detection and Deterrence System

  Simon Cohen, Abhishek Kumar, Michael Martin, Chance Shipley
  Advisor: Dr. Andrew Cohen

  Plagiarism in introductory programming courses is one of the biggest problems facing universities. Existing tools can identify cases of high confidence plagiarism through a variety of sophisticated machine learning processes. By identifying such cases, course instructors are alleviated of the pressure to manually seek out students who may be plagiarizing. However, currently available tools are inadequate for addressing the full scope of the issue. The similarity detection between student submissions typically produces arbitrary scores, which make the results difficult to interpret. Additionally, the results fail to consider the social nature of the problem; students tend to plagiarize within the same cohort repeatedly. Finally, existing tools fail to proactively deter plagiarism and instead serve as a reactive solution for punishing students. These shortcomings warrant the development of a preventative model that can detect low-confidence plagiarism and intervene earlier than previously possible. This study demonstrates a plagiarism detection and deterrence system to monitor student-cohort behavior and actively discourage the act of cheating in introductory programming courses. The model uses pairwise, compression-based similarity detection to capture relationships more accurately. An automated deterrence system is used to warn students that their behavior is being closely monitored. High confidence instances are displayed on a secure, online dashboard for instructor review. An unbiased scoring system aids students and the instructor in understanding true independence of effort. Preliminary results indicate that the system can provide meaningful measurements of independence from week one, improving the efficacy of technical education.

• ECE-05 Campus-Scale Internet of Things

  Har Patel, Matt Rantz, Kyle Russo, Saem Jeon
  Advisor: Dr. Kapil Dandekar, Dr. William Mongan (Ursinus College), and Dr. Malvin M. Nkomo

  The Internet of Things (IoT) describes networks of sensors and software designed to share data amongst each other, applicable across several domains including health, environmental studies, and even home life. Within these networks, gateways serve as access points for sensors to connect to a larger network and send information to a centralized location. There are various gateways available in the market today. However, many of these tend to be centered around protocol- or application-specific use cases without flexibility to expand the infrastructure. To use a new gateway for each use-case is costly and introduces technical hurdles many of those interested in IoT may not be well-equipped to overcome. The motivation for Campus-Scale IoT is to develop a multimodal infrastructure that is easily accessible across campus regardless of technical background. Many off-the-shelf commercial sensors will be able to connect to this gateway across a diverse set of protocols. The hardware component of the solution is a printed circuit board gateway, while the software primarily consists of an open-source IoT platform, ThingsBoard, and a Flask website. The Flask website helps users easily connect their devices via a form. Once a device is connected, users will be able to visualize the device data on ThingsBoard. Currently, data from multiple BLE, WiFi, LoRa, and ZigBee devices communicate via a gateway with ThingsBoard where data is successfully visualized in real-time. Data can be retrieved and manipulated via APIs and rule chains, and other groups have begun successfully integrating their projects with this one.

• ECE-06 Autonomous Multi-Seasonal Landscaping Robot

  Brandon Hall, Isaac Hong, Stephen Johnson, Cody Liu, Michael Livornese
  Advisor: Dr. David Han

  By eliminating the need for both a snowblower and lawnmower, customers will save money and storage space. It will also improve the lives of our stakeholders by reducing the amount of time they need to spend caring for their driveway and lawn. The robot will be able to successfully mow a lawn autonomously and have the capability to swap to a snowblowing system that can successfully clear a driveway on its own. There are two main challenges our team will face. The first challenge is developing a swappable system for the robot that allows it to change from one mode of operation to another. The second challenge is getting the robot to autonomously recognize its workspace and stay within defined boundaries. Our solution to the first problem involves developing mower blades that can be swapped out of the machine for snowblower auger blades and turbine setup. For the second challenge, the team intends to modify the code of an already existing system, like Roomba, or other open-source libraries in order to stay within the workspace and avoid obstacles. Our design shows and reflects a missing and potential industry for a low-cost autonomous multi-seasonal landscaping robot.

• ECE-08 Medical Delivery Drone

  Brendan Higgins, Tri Pham, Jordan Lanzoni, Manh Cuong Phi, Binh Tran
  Advisor: Dr. Adam Fontecchio

  Recent advances in robotics have made it possible to deliver packages with small autonomous drones. Over the past 5 years, companies like Amazon have been rushing to develop such drones that can deliver packages ordered online in a timely manner without any human interaction required. Most of these companies are still in the development phase, while a few have an initial product released. Due to advancements in drone technology being so recent, there are still many possible applications in other fields that have been overlooked. One such field with little drone implementation is the domain of medicine. The idea is to overcome the delays and inefficiency of traditional ground delivery by using the sky as a medium for transport. Our team has outfitted a drone with automatic navigation and onboard flight sensors to send it to its destination via GPS signals while avoiding obstacles while in transit. It is also equipped with a carrying compartment to store and deliver life-saving medicines quickly and efficiently. While existing medical programs like Telehealth allow for wireless diagnoses and medical aid through a computer screen, remote patients not having physical access to emergency medical supplies or sample collection for further diagnoses is still a major constraint. With the development of our delivery drone, life-saving medications and treatments can be delivered fast and directly to the sick without the limitations of ground transportation infrastructure.

• ECE-09 ROVRR: Remotely Operated Virtual Reality Robot

  Bernard Morelos, Ryan Bergauer, Akshat Pokharna, Violeta DeSantis
  Advisor: Dr. John Walsh

  Since the beginning of the 20th century, the world has gotten smaller and smaller with the help of technology. However, despite this closeness, the pandemic has made people more isolated than ever before as traveling and attending large gatherings are still high-risk situations. Many people ultimately desire to congregate at social gatherings and events without worrying about the risk of contracting illnesses. ROVRR aims to satisfy this desire by allowing people to remotely attend events and gatherings through the utilization of VR. Users will be fully immersed in a photogrammetry rendering of the event with multiple sources of live data streaming to them from a physical robot attending the event. So far, we’ve successfully created photogrammetry renderings of relevant areas within the school and of the area we expect to present the prototype. We’ve also successfully integrated a virtual rendering of the robot that is controllable by the VR user. We’ve streamed live video into VR and established a Mumble voice communications server. Our next steps are linking the virtual robot to the physical robot for accurate movement representation, and mounting an iPhone to the robot which will stream live LIDAR data. The rapid expansion of the VR market indicates that this realm will continue to grow into other sectors such as business and entertainment. This project will aid in this expansion by collecting live real-world data from gatherings and streaming it back to the safety of a VR environment for users to interact with and enjoy.

• ECE-10 Using HTCondor to Scavenge CPU Cycles from Lab Resources

  Sameh Abedin, Fiona Sarno, Joseph Stewart
  Advisor: Mr. Andrew Marx and Dr. Gail Rosen

  Modern research is highly dependent on computing power whether it be to record data, run simulations, or analyze data automatically. Many research groups outsource this computation to third-party resources where the servers and architecture are maintained for them such as with Amazon Web Services (AWS). For underfunded or unfunded research projects, this outsourcing is not a realistic option. A better option for Drexel’s researchers would be to utilize the unused computing power already available on the campus machines. Currently, Drexel’s ECE and MEM departments have over 100 machines available to faculty, staff, and students at any given time. These machines spend the majority of their time powered on without actively being used. Instead, these machines sit idle wasting CPU cycles. Our project seeks to use these cycles for High Throughput Computing (HTC). Our project strives to accomplish two goals. First, to provide a resource for Drexel students and faculty to solve any kind embarrassingly parallel problem they might encounter through HTC. Second, to utilize the wasted CPU cycles of the idle machines in the College of Engineering computer labs at minimal extra cost. To accomplish these goals, we have elected to use HTCondor, a job scheduler that specializes in scavenging CPU cycles from idle machines for HTC. As of now, we have successfully implemented HTCondor on a small pool of computers. Further, we have shown results in applications such as: Basic Local Alignment Search Tool (BLAST), python, and MATLAB. As this project continues beyond our tenure, we expect a focus on the security and usability of this new resource.

• ECE-11 THRIVE

  Katayoun Hazaveh, Neil Patel, Richard Scott, Tobenna Oduah
  Advisor: Dr. Gail Rosen

  Plant-growth chambers are designed to produce environmental conditions that maximize plant growth. Such technology has inspired urban farmers to increase self-reliance to grow accessible and sustainable food. Unfortunately, despite the initial attempts of individual urbanites to develop their horticultural methods, there seems to be a complete lack of community and resources for urban agriculture.

  Previous prototypes of urban growth chambers lag behind the current technology and fail to utilize community-driven analytics. This problem could be improved by developing a better performing controlled environment system that utilizes state-of-the-art plant characterization and introduces data networking.

  THRIVE’s objective is to use unified open-source, IoT, AI, and 3E (Emulated Earth Environment) based solutions to provide a unique framework that allows for crowd-sourced data to help control individual plant growth environments.
  Advisor: THRIVE will bridge the divide between laggard technology and lack of resources in urban farming by presenting an integrated solution and network architecture that allows an ecosystem for multiple controlled-environment instances to collaborate and share phenotypic data.

  The project features a microcontroller and climate sensors that are readily available. The software includes contemporary machine learning approaches as a recommendation system to provide context of the plant’s health, which enables the user to adjust the control parameters of the 3E. A cloud-based multitenant transfer and storage serves as the backend of the network infrastructure which goes along with the state-of-the-art user interface component created to make the interactive process between the user and the assembled 3E easier.

  The system will serve to foster community-driven analytics for urban food production and will serve as inspiration for better plant growth chamber design.

• ECE-12 Optimizing Power Distribution System Reliability with Inverter-Connected Distributed Energy Resources (DER)

  Devin Thurber, Jake Ward, Renat Ramazanov, Forrest Leffer
  Advisor: Dr. Karen Miu

  The power distribution grid traditionally supplies power radially from a centralized power source to customer locations demanding power. Currently, there is an increase in Distributed Energy Resources (DERs) penetrating the power distribution grid in the form of grid tied solar, wind, diesel, and battery sources. One area of research being studied is what occurs during power loss due to a fault in the grid, like downed power lines. Usually there is a loss of power until the connection is restored. DERs presently installed can potentially supply power when connections to the centralized source is unavailable.

  Generation between residential and commercial establishments can network to power more than just themselves. Providing power to these establishments that normally would have been out of power decreases the customer average interruption duration index (CAIDI) - an important reliability metric. This scenario is referred to as an Island - a live circuit that is disconnected from central generation. IEEE Standard 1547-2018 has strict requirements involving Islands and whether they can be formed.

  This project seeks to present possibilities of using DERs to supply multiple loads in an islanded circuit. The project will produce repeatable Drexel power laboratory experiments using controllable sources/loads and software to further this research. Techniques of inverter controls and switch placement are explored to balance the demands with the generation of the island.

  The team has been researching various equipment used in the experiment including Yokogawa Power Analyzer, Chroma AC Loads, and Chroma AC Sources to accurately emulate real power island scenarios.

• ECE-13 LBPAlert - Wearable for Low Back Pain Risk

  Bunmi Oluwatudimu, David Ajibade, Ekenedilichukwu Illoh, Kathrina Waugh, Kweku Aboagye
  Advisor: Dr. John Walsh

  Low Back Pain (LBP) is one of the most prevalent musculoskeletal disorders. Four out of five people will experience some variation of LBP in their lives which can have a profound impact on their quality of life due to its associated physical and economic burdens. Although there has been a plethora of research efforts towards pain and its association with fatigue as well as health-care resources devoted towards diagnosing LBP, these efforts are hindered by their lack of accuracy and convenience due to their invasiveness. LBPAlert combines novel non-intrusive sensors, an advanced cloud-based machine learning framework, and a cross-platform software application to sense, predict and alert an individual of their risk to low back pain in a highly accurate and convenient manner. LBPAlert presents itself as an early predictive solution in the form of a wearable with the potential to reduce the steady growing population of LBP patients.

  Non-invasive sensors such as Surface Electromyography (sEMG) sensors record muscle activity based on biopotentials, and Inertial (IMU) sensors - Accelerometer, Gyroscope and Magnetometer - record the body’s relative position and track motion. The sensor data is streamlined to a cloud-based machine learning model which makes a real-time numeric prediction within the 11 point Numeric Pain Rating Scale where 0 represents no pain and 10 represents extreme pain. The numeric prediction along with consolidated low back muscle data is communicated to an iOS or Android based application - giving users their health data in the palm of their hands.

• ECE-14 Frame-Mounted Bike Safety System for Vehicle and Road Hazard Detection

  Pankil Chauhan, Isaac Deerwester, Cory Spiner
  Advisor: Dr. David Han

  Cycling is a popular form of transportation and leisure, but the inherent dangers of sharing the road with much larger and faster vehicles pose a threat to riders’ safety. Even with significantly decreased vehicle traffic due to the COVID-19 pandemic, 675 cyclists were killed in accidents in 2020 compared to 712 fatal vehicle crashes in 2019. A cyclist’s rear blind spot is a gap that raises concern, especially for hearing-impaired cyclists and the growing popularity of quieter electric vehicles. Devices on the market today that seek to address this issue often struggle to perform well in busier cities due to difficulty distinguishing real incoming hazards from false positives such as parked cars. They also require the rider to divert their attention from the road. We propose a rear-view, frame-mounted device that will communicate with a helmet-mounted heads-up display to relay hazards detected by a camera module. Together, the two devices will augment and enhance riders’ awareness by indicating when there is a vehicle approaching from their blind spot on either side without the need to take their eyes off the road. The camera-based approach, utilizing machine learning, will create a more reliable system capable of far greater awareness than the radar-based systems in use today. To test and demonstrate the system, physical prototypes of the system components, along with functional software, are created, mounted, and tested. The performance is measured by quantifying and evaluating the number, reliability, and quality of hazard detections along with the warnings subsequently triggered by them.

• ECE-15 Off-Grid Electrically Heated Suit

  Madeline Harmes, Ben Siti, Michael Bonacquisti
  Advisor: Dr. Kevin Scoles

  Enjoying outdoor winter activities in much of the United States is difficult or unrealistic due to the freezing temperatures. Winter sports such as skiing or snowboarding thrive in the cold weather. However, other activities like hunting and hiking become uncomfortable on the body and potentially dangerous if exposed to freezing weather for too long. A variety of battery-heated clothes have been created to combat freezing weather, such as: vests, gloves, and boots.  All these products fall short for two reasons: the garments do not heat the entire body and there is no option to recharge the battery from a remote location. Staying warm and safe is especially important in a freezing, remote location. There is one battery-heated product that encompasses the entire body; however, it is more similar to a sleeping bag and does not allow for mobility. To solve this issue, a full-body electrically heated suit has been designed. The heat suit can be charged by a wall outlet or by an included solar panel. A full-body electrically heated suit with a remote charging option solves all the problems listed above. This product will allow people to safely enjoy their favorite outdoor activities even in freezing conditions. Suit temperature can be adjusted with the control belt on the suit. The battery can be recharged whenever the user is stationary and wherever there is sun. 

• ECE-16 Reducing the Latency of Millimeter Wave in 5G Cellular Communication

  Davis Ranney, Emil Snyman, Jamie Wu, Kevin Yu, Md Abu Saleh Tajin, Kyei Anim
  Advisor: Dr. Kapil Dandekar

  Millimeter Wave (mmWave) is becoming the new standard for high-speed 5G cellular communications. It can deliver gigabit speeds and low latency streaming and handle massive deployments of receivers. One of the inherent drawbacks for mmWave is that antennas need to radiate the wireless signal directly at the receiver, and these beams are very narrow to overcome path loss and signal attenuation. Antennas must search through many beam-states to identify the receiver's direction which drastically affects the latency of mmWave communications. Several solutions have been proposed, such as applying deep learning, to reduce the search time or space. For increased reliability, sub-6 GHz antennas are often deployed in conjunction with mmWave antennas. With a sub-6 GHz antenna capable of beamforming, it can inform a mmWave antenna of an approximate direction of a receiver. The approximation can be estimated more quickly because sub-6 GHz beams are wider than mmWave beams and thus have fewer states to search through. Together, sub-6 GHz and mmWave antennas can swiftly identify optimal beams and reduce the latency of 5G communication. Previous research has validated that this concept works by manually steering the mmWave antenna. We have designed a solution to automate steering of the mmWave antenna and implemented an intelligent algorithm to control the antennas by predicting the Angle of Arrival (AoA) based on Received Signal Strength Indicators (RSSI). We can compare this system with other search algorithms and quantify the reduction in latency that commercial systems should experience if they deploy a similar strategy.

• ECE-17 Hyperspectral Imaging Drone System

  Francis Franco, Mansoor Khan, Fredy Sanchez, Jeffrey Tharakan, Alan Zhou
  Advisor: Dr. Adam Fontecchio

  Crop production experiences numerous drawbacks, each requiring improvements in technology to increase efficiency. Hyperspectral imaging, a newer technology involving the measurement of light from pictures taken at narrow spectral bands in a contiguous spectral range, is used to characterize large spectrums of biological, chemical crop traits, and soil by analyzing the reflective properties over its spectral band's range. Through this technology, several parameters such as soil classification and weather can be done quite easily, making it functionally better than spectral imaging. Hyperspectral imaging has been successfully employed in a variety of remote sensing applications that requires the calculation of physical parameters of many complex surfaces. It also identifies visually similar materials with fine spectral fingerprints due to these characteristics. Plant diseases, soil erosion, weeds, crop identification, are factors that have derailed crop production. These processes can help in improving the efficiency of production to meet ever-rising agricultural demands. Drones in use can be improved, incorporating hyperspectral imaging technology to enhance crop production efficiency. Multispectral and thermal sensors can be used to sample and calculate the chlorophyll, pesticides absorption, water deficiency, and diseases based on reflectance. The design & construction of the drones will require precision and will be costly, but the results will prove worth the cost. This paper focuses on explaining the current crop production problem in this field and the creation of a low-cost optic lens-based hyperspectral imaging system for around $800.

• ECE-18 Astrocelium

  Rhea Dutta, Celine Khoo, Bethany Pittman, Daniel Rodriguez
  Advisor: Daniel Belquer and Andrew Wiggins

  For decades, technology has been used to enhance and optimize the music and entertainment industry. Today, this is more prevalent than ever with the new capabilities of wireless technology and real time data transmission. After achievability, the next concern of technology’s integration with music is accessibility. How do we leverage this technology to allow everyone to enjoy the art that it elevates? The simplest examples of this are seen in services like streaming platforms for music and TV, or live streams of performances and concerts to make entertainment more accessible to all. However, technology can take the accessibility initiative to heights never previously imagined in the music industry. For example, Daniel Belquer has dedicated much of his professional career developing technologies that provide unconventional musical experiences for impaired individuals. Most notably, he developed a vibrotactile to create a better live music experience for the deaf. The goal of Astrocelium is to continue and build on this idea of accessible and inclusive entertainment by implementing available technologies into an integrated musical experience through senses other than hearing. The project aims to integrate multiple available musical applications, electronic devices, sensors, and microcontrollers into a large plant-like structure that will be controlled by live musical pieces to create visual, auditory, and sensory displays.

• ECE-19 Hindsight’s Personal Safety Backpack

  Spencer Fan, Steven Lee, Kevin Ly, Christina Phan, Jacob Wong
  Advisor: Dr. Steven Webber

  According to a RealTime research survey, “50% (of people) feel unsafe walking alone at night”. This concern not only affects students but parents and loved ones who want to keep them safe. Current solutions consist of defense weapons like pepper spray and tasers, or alert devices like whistles or phone apps. These are all reactive solutions that attempt to treat the issue after it’s already happened. We are the first to explore a preemptive personal safety device. This device will allow people to avoid confrontation entirely without staking their lives on how calm and rational they are in a high-stress situation, how well they can use a defense device, how fast others would come to their aid from an alert device, or how dangerous the attacker is. No matter how aware a person believes they are, they can not reliably shift their focus to cover 360 degrees. We are proposing a device that will mitigate these physical and mental vulnerabilities. This will take the form of a backpack with lidar sensing to provide real-time haptic alerts with information on a pursuer’s distance and direction. The backpack will contain vibrational motors on the straps and a gimbal stabilizing the lidar sensor. This device will also interact with a web app that can read the alerts and also notify emergency contacts associated with the user’s account. After validation testing, the system was able to accurately detect pursuers and increase the user’s spatial awareness making their nightly walks safer and less stressful.

• ECE-20 CHEF: Robotic Kitchen Assistant

  Julie Le, Gavin Fox, Phil Jones, Abhigyan Khullar and Samantha Efman
  Advisor: Dr. Christopher Peters and Dr. Tom Chmielewski

  61 million Americans live with a disability in the United States. These disabilities can include fine motor, mobility, vision, and cognition impairments, among other variations. These, unfortunately, have far reaching implications that disrupt the lives of millions of people on a daily basis. Amongst this population, fine motor disabilities end up leaving 7% of disabled American adults struggling with self care, and independent living, consequently. Specifically, it can be difficult to perform the tasks that are necessary in order to prepare a home cooked meal. This can result in over-dependence on assistance from others, unhealthier diets, and a myriad of non-ideal circumstances for those affected. CHEF will alleviate this pain point for those who struggle with this particular challenge, allowing them to achieve more independently in their own home. It will result in more independence for the disabled individual, improve the quality of their diet and provide an extra level of safety when performing tasks that may have otherwise put them at risk. The CHEF robot provides an alternative to existing expensive, “nice-to-have” kitchen gadgets that are not targeted for this demographic. CHEF is a tabletop gantry system that will perform all appropriate food-safe actions using various stations to prepare, cook and plate a full, nutritious meal for the user. CHEF will provide end-to-end assistance for someone who would have otherwise been unable to perform those actions safely. Ultimately, the user will provide very little input to the system, which will autonomously recognize and prepare the meal.

• ECE-21 Mini-Golf Modernization

  Declan Beaudoin, Joshua Boniuk, Finn Clements,
  Advisor: Dr. Christopher Peters

  Mini-golf is an industry which currently uses old-fashioned methods of score keeping. The current method uses a significant amount of paper and mini-pencils and requires a lot of user intervention. An example of an industry that struggled with outdated score keeping methods, and saw a surge in popularity when modernizing its methods is bowling, which is now almost exclusively automated. Evidence suggests that, as seen with bowling, automating the score keeping process enhances gameplay by removing the tedious tasks required by the current system, allowing players to focus on the game experience itself. Additionally, the added flares of having a computerized scoring system appeals to a wide variety of mini-golf players. The promising nature of the popularity increase that bowling saw leads us to believe that this too can apply to mini-golf. The purpose of this project is to extrapolate from the success seen with bowling and apply that knowledge towards improving the mini-golf experience. The effectiveness of the project is being confirmed through market research surveying of mini-golf beta test players. Initial survey results have shown that most players enjoy our alterations to mini-golf and the majority would choose automated mini-golf over standard mini-golf.

• ECE-22 U-Grow: Autonomous Herb Garden

  Bryce Trimble, Frank Tuminello, Tobias Green, Zoe Sucato
  Advisor: Dr. Bruce Eisenstein

  Growing plants can be a daunting task. Factors for creating an optimal environment include temperature, humidity, sunlight, water, pH balance, and nutrients. If one of these factors is incorrect or neglected, a plant’s growth success rate can significantly decrease. The goal of U-grow is to automate the climate conditions of a miniature greenhouse for optimal growth of specific plant selection. Current solutions on the market are lacking in multiple places. Most merely have lights and a water supply on a timer. For people that don’t want to give raising plants a second thought, this won’t do. This project’s objective is to bridge the gap between overly-professional and overly-simplistic indoor greenhouses. U-Grow was designed to keep environmental conditions stable while being sleek and suitable for an amateur gardener. U-Grow autonomously monitors and controls conditions, while posting real-time data to an app. A prototype is under development that will demonstrate cultivation of a variety of herbs and maintenance of environmental control. So far the electrical subsystems have been successfully tested and controlled, the communications scheme has been developed and is in testing, and the mechanical body of the system has been designed and assembled. After elaborating on the need for this product, this poster will outline the subsystems of the design, including mechanical construction, hardware requirements, control scheme, and application communication design, and detail U-Grow’s economic impact and marketing plans.

• ECE-23 Control System Design for a Climate Controlled Enclosure

  Amanda E Hipple-Bornhorn, George Guo, Michael Yu
  Advisor: Dr. Bruce Eisenstein

  Reptiles are getting more popular as pets during the past several years in the United States. Social study shows that the number of households owning reptiles had doubled from 2007 to 2020. Demographic change had an influence on the rising popularity of reptiles. The college tuition has raised far more dramatically than minimum wage and average salaries. Which made more young people live with parents and stuck in smaller apartments. So they won’t have room to have a dog. The changing perception of reptiles also let more people consider reptiles as a potential pet. Media and popular culture created more positive images instead of evil and scary reptiles.

  Unfortunately, most reptile and amphibian enclosures provide improper husbandry causing them to experience illnesses and sometimes death. These enclosures could be improved by maintaining proper temperature, humidity gradients, and light cycling.

  This paper presents and evaluates a new solution which properly automates the environmental conditions of an enclosure for a herptile reduces the burden of husbandry on a keeper. Instead of constantly checking and altering the conditions in the enclosure manually with devices by the keeper, a remotely control system can be implemented to check and maintain conditions. The widespread adoption of such a remotely control system can lead to significantly fewer premature deaths of reptiles through the reduced change of improper environmental conditions, particularly among inexperienced keepers.

• ECE-24 Smart Refrigerator

  Yelnur Abilakim, Oluwasijibomi Akinkugbe, Wyatt Brisbane, Keidi Xhagolli
  Advisor: Dr. Jaudelice de Oliveira

  Smart Devices have been making huge strides in the world today. From Amazon’s Alexa devices to Apple’s Home Dot and even Google’s Lifx Mini Smart Bulb, smart devices are making the lives of individuals and businesses easier and more convenient. Smart Refrigerators are the next generation of smart devices.

  There are smart refrigerators on the market, however their smart features usually provide help with cooking or maintenance. Almost none of them target the issue of food waste directly. Implementing features that can notify the owner of expired or close to expiring food can help fill a niche that no fridge manufacturer has yet filled.

  We have developed a smart refrigerator device that can help combat food waste by allowing the user to always have a correct idea of the contents and its respective state in their refrigerator and make choices that will help reduce the amount of wasted food. Our solution includes an (a) application, (b) touchscreen, and (c) barcode reader. The touchscreen is mounted on the refrigerator to allow the user to manually input and view refrigerator contents without opening it. The barcode reader will be used to automatically add food items to the smart refrigerator database to accurately predict the expiration date. The application will be used to send notifications about the approaching end date of and will additionally display the refrigerator’s contents to the user.

• ECE-25 Retrofit Vehicle Camera System

  Jack McVeigh, Clayton Sheffield, Shawn McHugh
  Advisor: Dr. Prawat Nagvajara

  In 2018, a law went into effect that requires all new vehicles to have backup cameras and video displays. Less than half of cars out on the road today are equipped with these capabilities, and it is not projected to be a majority for years to come. Many aftermarket rearview camera modifications can cost hundreds of dollars, and require the consumer to pay for a monitor as well as professional installation. Apart from do-it-yourself projects, there are no cheap products that solve this problem. This product will not only act as a backup camera when reversing but also as a dashcam. The Retrofit Vehicle Camera System (RVCS) is also capable of recording and storing video feeds using Bluetooth or Wi-Fi capabilities. RVCS will set the standard by combining an affordable camera system with a smartphone, eliminating the need for a separate monitor.

• ECE-26 Improved Projectile Accuracy with Drone Assistance

  Kanstantsin Paulau, Philip Vee, Kevin Vo
  Advisor: Dr. Prawat Nagvajara

  The United States has one of, if not, the largest military budgets in the world. Criticisms of the high spending has led to the Department of Defense (DoD) cutting costs. Some of the strategies in reducing costs are downsizing personnel and modernizing older equipment with newer technology. This modernization agenda includes unmanned aerial systems. Drones would fall under this category and have seen increasing usage with both the military and civilian population in the past decade. What is a way in which the DoD can modernize equipment and use fewer personnel to be more efficient? Our team has identified the mortar system as equipment that the DoD has been trying to improve on for a significant period. In particular, the accuracy of mortars has been seen as unsatisfactory as GPS-guided rounds were one of the proposed solutions by the military. Our team has been making a system that integrates a projectile launcher and a commercial drone to make the firing process more accurate. The use of a drone is to reduce costs over the long term as guided rounds are expensive and have limited use. Microcontrollers have been attached to the projectile launcher so the drone can go to where the predicted landing will be. The drone provides visual confirmation and maps out a line of its flight path. After the first shot, we want to calculate the error and figure out the best adjustments to make for the subsequent projectiles.

  ECE-27 Smart, Efficient and Light Solar Microgrid Inverter (S.E.L.S.M.I)

  Omar Gomaa, Cody Tran, Modibo Fomba, Henry Nguyen, Fatimata Dia 
  Advisor: Dr. Fei Lu

  In the past 40 years, solar energy has grown and expanded from just powering satellites in space to a technology that powers homes and businesses. Inverters are an essential part of the installation with their main function being converting the DC electricity from the solar panel to AC electricity that can both feed the load, and/or be integrated again in the grid and supplied to the utility. Before the 1960s, solar systems were only 6% efficient and cost $300/Watt. With the advancement in both PV technology and more economical solar inverters, the average solar system performs at an efficiency that ranges from 14 to 18 percent efficiency and could cost as low as $3/Watt. As more PV solar systems get integrated into the electric grid, energy efficiency and installation costs became a concern to residents and businesses. The objective of Smart, Efficient, and Light Solar Microgrid Inverter (SELSMI) is to provide grid support functions, such as frequency support, voltage regulation, and ride-through capabilities. The design aims to reduce heat sink losses, weight, and system’s size, hence reducing the overall installation cost for the customer. This is done by optimizing the capacitor and inductor values for the LCL filter, the most convenient Full-bridge for the DC-DC converter, an STM32 based controller, and isolators.  

• ECE-28 Smart Wheelchair: Brain-Actuated and Semi-Autonomous Assistive Device Interface

  Mario De Lorenzo, Kaustav Bora, Mohor Bhowmick, Ashley Siddiqui
  Advisor: Dr. Hasan Ayaz

  According to the WHO, 65 million people depend on a wheelchair for mobility to complete day to day tasks. An estimated 40% of the disabled community finds power wheelchairs difficult, and in some cases, impossible to operate. This includes patients suffering from Tetraplegia, Spinal Cord Injuries, Parkinson’s, or cognitive deficits. Conventional motorized wheelchairs’ interfaces use joysticks, head/chin control, tongue control, and face/gaze control, all of which still requires continuous movement or posture control that patients or users have difficulties keeping up. The recent emergence of Brain-Computer Interface (BCI) technology presents an opportunity to capture input user intention for relaying them to computer and machines in diverse communication and control scenarios. In this project, we propose a new motor-imagery (MI) based mobile EEG-BCI to control navigation of a motorized wheelchair. This is also enhanced by collision avoidance IR proximity sensors as an additional layer of safety and semi-autonomous control. All together with these capabilities, we aim to build a Smart Wheelchair. To capture brain dynamics continuously during wheelchair driving, an 8-channel Unicorn mobile EEG system with OpenVibe software stack will be used. The Smart Wheelchair will be enabled to have a recording session to train inter-subject models for 4 classes (opening and closing left, right, both hands and both feet). As an embedded computational platform, Raspberry Pi 4 will be used to perform real-time BCI operations that include visual stimuli display, signal pre-processing, and classifiers. This Smart Wheelchair aims to provide greater accessibility and ease of use for severely disabled patients.

• ECE-29 Autonomous Shopping Cart

  Andrew Judge, Tony Mao, Francis Silvestri, Sudarshan Krishna Bhatia
  Advisor: Dr. James Shackleford and Dr. Tom Chimelewski

  Transporting goods through a grocery store is necessary for customer satisfaction and store success. Pushable shopping carts offer a solution to this problem, but prove difficult to use for customers with disabilities. Electric riding carts have allowed disabled customers to get around this hurdle by offering the choice to steer through the store and reach for groceries. These solutions however, strip customers of a sense of autonomy, sometimes carry stigma that customers would rather avoid, and can make it hard for customers who have trouble, or would not like to push a cart, but can otherwise walk, to reach certain items.

  We have designed an automated cart that has the capability to follow a user around a store. The base has been constructed out of HDPE plastic Omnidirectional wheels are used for improved turning. The navigation of the cart will be accomplished using IR transceivers with overlapping conical detection regions. Object avoidance will be accomplished using ultrasonic sensors. Progress has been made to construct the base of the cart, with wheels and motors, to build and test a motor control circuit consisting of an Arduino, a battery and four motor drivers and to build an IR sensor circuit consisting of an Arduino and four IR sensors. The Motors have been tested to turn when the IR receiver falls in the transmitter’s detection region.

• ECE-30 Self-Balancing Bicycle

  Jason Ellstrom, Phil Huddy, Zach Miller, Peter Nguyen, Theo Strahan, Kenny Tea
  Advisor: Dr. Thomas Chmielewski and Dr. Paul Kalata

  The self-balancing bicycle is an autonomous transport system that can transport a load from one destination to another; the system’s control is based on the inverted pendulum model. The inverted pendulum model works by applying counter forces to keep a system stable enough to remain upright in otherwise impossible conditions. This project is a continuation of a previous senior design group that researched and began the implementation of the pendulum model using state-space control systems. In this field, there have been very few previous attempts at a self- balancing bicycle. Unlike Google Netherland’s CGI version, this project is aimed at creating a real self-balancing vehicle in the future. Although the groundwork has been laid, there is still much more work to be done. For example, a few existing parts of the bike need to be fixed, and the construction of the bike must be completed. Additionally, all the software for the bike needs to be written. Finally, once the bike is working, the model will need to be iteratively tested to maximize stability. The creation of this self-balancing bicycle will be groundbreaking because it takes advantage of the inverted pendulum problem to create a linear model of a complex issue. Doing this would be a huge accomplishment for control theory. Due to unanticipated complications in getting components and subsystems working the scope of this senior design project has been modified to reach a functional state for all subsystems as the final deliverable. These completed subsystems will then be available for use as a steppingstone towards a fully autonomous self-driving bicycle.

• ECE-31 CAMEL LiDAR Drone

  Evan Byrnes, Christopher Lijoi, Andrew Ragni, Laxmi Sanigepalli, Maggie Wang
  Advisor: Dr. Adam Fontecchio, Nicholl Glasco (Collins Aerospace)

  Emergency scenarios in urban environments risk the safety of both civilians and first responders, so gathering information on the situation is critical for efficient response time. Current research on three-dimensional live mapping primarily focuses on the locations of first responders within the building itself. However, having knowledge of the surrounding site can be just as critical in an emergency scenario. By examining the building as well as the area surrounding it, we can direct bystanders away from dangerous areas and identify key areas for the deployment of rescue vehicles. This team is implementing the use of a drone with attached LiDAR sensor to be deployed in fleets in the event of emergencies so as to generate a three-dimensional map of an emergency site and potentially decrease response time. The drone in use is an EXO X7 Ranger paired with a SICK TIM 571 LiDAR sensor. This sensor has a horizontal aperture angle of 270º with a scanning frequency of 15 Hz, allowing for a scan every 20-30 ms. This team has been testing the gathering of these scans to generate a point cloud for 3-d imaging of an area with implementation of a ROS to the raspberry pi. When enough images are spliced for a mapping, a filter is applied to smooth out the data and eliminate any possible outliers. Mounting brackets have also been designed for the hardware attachment of the drone and testing for weight distribution will begin shortly when the elements are added to the drone.