Team Leader
Amanda Lacy
Texas A&M University
Computer Science
aklacy@tamu.edu
Project Type
Research
Who Can Join
Faculty, Staff/Postdoctoral Scholars, Graduate Students, Masters Students, Undergraduate Students
Project Description
Qualifications- Junior or Senior
Experience using machine learning
You must commit to 10 hours per week, please don’t apply if you cannot meet this requirement. If you have time this summer, you can ‘bank’ hours working remotely so that you can have more flexibility during the semester. You will be required to log the hours that you work.Please start your email subject line with “Aggie Research Tactile VR” + your name
We have a system that uses 8 magnetometers to track a stylus in mid-air as well as on a surface (like writing on a page). Our existing algorithm works but has some artifacts that make it difficult to use. The goal of this group is to apply machine learning to create an accurate and efficient code base that outputs ‘explainable’ machine learning. Rather than a machine learning ‘black box’. The goal is to have the system generate mathematical models for what it is doing so that the system can be audited and changed without having to start over. This explainable machine learning paradigm should allow finer control of the system and iterative improvements in small steps.
The initial experimental setup will involve a pendulum with a magnet housed inside it. The pendulum’s swing is quite predictable, giving a wealth of training data with a pendulum’s motion as the ground truth. After training on a single magnet, multi-magnet styluses will be used to train the system for things like pressing a retractable tip and twisting a selector dial. These styluses will be built by the “3D printing and CAD for accessible technology” group.
Use of AI tools is encouraged, but the goal is to keep humans controlling the ‘reigns’ of the resulting system. We want to enable improvements without the help of AI tools, and the ability to employ different AI tools to build upon previous AI efforts.
Here is a video of the magnetometer system in action so you can be familiar with the capabilities: https://www.youtube.com/watch?v=wtNXnpAkpYQ
Team Needs
Machine learning and an ability to apply the math that correlates to the patterns found. Explainable machine learning.
Special Opportunities
While the inspiration for this work is non-visual accessibility, we want you to think beyond this to meet everyone’s needs.
As you work on this project, many applications will become apparent. You can flesh out these opportunities and create new tools.