IEMS 2021

Permanent URI for this collection


Recent Submissions

Now showing 1 - 5 of 13
  • Item
    Real-time kinematics directed swarm robotics for construction 3D printing
    (Industry, Engineering & Conference Management Systems Conference, 2021-03) Wang, Darren; Zhu, Robert
    Architectural advancements in housing are limited by traditional construction techniques. Construction 3D printing introduces freedom in design that can lead to improvements in building design, resource efficiency, and cost. Designs for current construction 3D printers have limited build volume and, at the scale needed for printing houses, transportation and setup could become issues. This work is an exploration of and proposal for a swarm robotics-based construction 3D printing system that could bypass these issues. In this system, a central computer will coordinate the movement and actions of a swarm of robots which are each capable of extruding concrete in a programmable path and navigating on both the ground and the structure. The central computer will create paths for each robot to follow by processing the G-code obtained from slicing a CAD model of the intended structure. The robots will use readings from real-time kinematics (RTK) modules to keep themselves on their designated paths. The current progress in hardware and software is available on GitHub for collaborative development of this proposed approach towards construction.
  • Item
    LungStat: improving lung cancer diagnostic accuracy through computer vision
    (Industry, Engineering & Conference Management Systems Conference, 2021-03) Sonawane, Kaustubh; Rai, Aditya
    Non-small cell lung cancer (NSCLC) results in over 1.8 million deaths worldwide every year; however, most of these deaths are preventable via early diagnosis, which reduces the mortality rate by more than 50%. Currently, physicians use a CT (Computed Tomography) scan as a preliminary method of identifying cancerous tumors. Unfortunately, this visual process of identifying NSCLC scans becomes time-consuming and inaccurate, leading to high misdiagnosis rates. The goal of this project is to create a cloud-based web application that can take an inputted CT scan and identify regions of potentially cancerous tumors at an accuracy >90%. This is accomplished by first standardizing all inputted scans to a standard size and range of pixel values. A 3D CNN is then trained to classify an inputted scan as either "positive" or "negative" for cancer. Class Activation Mapping (CAM) is then used on scans classified as "positive" in order to identify the location(s) of cancerous tumors. This algorithm core is accessed through a cloudbased user interface on AWS allowing physicians to upload and organize their patient's NSCLC scans as well as receive dynamic results based on patient biometric and cancer history background. Using the LungStat platform, an oncologist can simplify the procedure for lung cancer diagnosis by cutting down the average tumor identification time from about 1 hour to a few minutes. Overall, this project establishes a critical tool needed for the accurate diagnosis and treatment of NSCLC, leading to a severe reduction in the death rates caused by lung cancer.
  • Item
    Theoretical study on instability of the base solutions of Lorenz system via Ordinary Differential Equations
    (Industry, Engineering & Conference Management Systems Conference, 2021-03) Davani, Sina; Asmatulu, Ramazan
    The Lorenz system is well-known for producing chaotic solutions for a particular range of technical systems and process characteristics. To examine the instability, fluctuation parameters in the form of exponential functions are introduced to the base solutions of the ODEs, and numerical and computational methodologies are used to determine the range of values that cause instability under different conditions. This paper will focus on the Lorenz system's instability of Ordinary Differential Equations (ODE) that regulates the thermal-hydrodynamic behavior of a two-dimensional fluid layer which is equally warmed (distributed) from bottom to top. The boundary layer of instability is also examined and compared with numerical and computational methods, and the accuracy of the solutions is thoroughly investigated. The study will give considerable insight into the requirements for stability of numerous technical systems such as DNA analysis, chemical reactors, thermosyphons, Malkus waterwheel, electric circuits, DC motors, pipelines, and power generation using micro as well as nanoscale techniques.
  • Item
    A multipurpose robotic glove designed to teach sign language through guided manual motions
    (Industry, Engineering & Conference Management Systems Conference, 2021-03) Joshi, Soham; Malik, Raaghav
    Projections show the numbers of deaf and deafblind individuals are quickly increasing, as both the number of deaf and blind individuals are expected to double within the next 50 years. Additionally, there is limited support for deafblind individuals, as they have little sensory access to the world. Most existing solutions are inadequate because they are either not portable, readily accessible or are very expensive. The goal of our project is to create a low cost, accessible solution for those with these disabilities: a portable device that utilizes the sense of touch to teach sign language, instead of conventional sign language learning methods that rely on sight. By doing this, a more immersive and personalized experience is created. The implementation of this solution is two-pronged: first, the physical portion of the solution is equipped with servo motors that control the pulling and release of a cord threaded through rings on the fingers that moves the user's hands into various sign language positions, which is closely modeled after human hand anatomy. Secondly, feasibility is determined for an AI algorithm to take in data from an external camera to efficiently add new signs to the glove, and also to track the user's sign language patterns and rate the user's accuracy with various signs. Through these mechanisms the user is engaged in learning new signs and expanding their sign language vocabulary, all without their sense of sight. We were successful in creating a working prototype analyzed through hand-tracking mechanisms, and achieving a 91-92% accuracy for the American Sign Language alphabet.
  • Item
    A brain computer interface system for the improvement of cognitive and communication abilities for patients with neuromuscular disorders
    (Industry, Engineering & Conference Management Systems Conference, 2021-03) Ramakrishnan, Navya
    More than 16 million people in the United States are living with cognitive impairment. Reports suggest that 12,000-15,000 people have amyotrophic lateral sclerosis and approximately 17 million people have cerebral palsy, globally. The engineering goals are (i) to design an experimental study to analyze and improve cognitive performance and (ii) to develop a communication aid to support people with neuromuscular disorders using non-invasive brain-computer interface. A low-cost EEG device, Emotiv EPOC+ is used to record EEG data and a Python interface is used to stream the data for analysis. The features extracted are used to train the classifier, Linear Discriminant Analysis. The participants' cognitive performances were measured initially and after giving 20 days of feedback sessions with alpha-numeric speller. The performances were in the range of 76%- 81% initially and accuracy improved for all the participants and are in the range of 84%-89.3% after feedback sessions. The proposed feedback training design is an excellent way to improve cognitive abilities and can be used for healthy individuals as well as people with attention deficiency to improve their attention. A software application is developed to use the system as a communication aid for neuromuscular disorder patients who are unable to communicate. The accuracy in identifying the words of participants' choice are measured only using their brain activity. The communication aid was able to predict more than 91% of the words correctly. The system is low-cost and easy-to-use with a short setup time regardless of users' expertise.