Robots search for a goal from indirect clues, using Bayesian inference, and share tasks through decentralized allocation when links are weak or lost. We now run these protocols on the real robots.
Student: James Lott
Real-time person and face detection on Jetson hardware, running on a field robot that localizes itself with lidar, GPS and an IMU. Our earlier face recognition work won best paper at IEEE UEMCON 2025.
Students: Sofia Sacchetto, Alex Beheshti
A car with five ultrasonic sensors, controlled over Bluetooth, drives a maze. We compare machine learning methods across Raspberry Pi 3+ and 4, Jetson Nano and Orin.
Team: Dr. Alireza Pakravan, Richie Andrews, Zoe Hassett
Building a three-dimensional view of the environment from a two-dimensional lidar.
Student: Abigail Trevillian
Computer vision for composite materials. With the Department of Mechanical Engineering at USD, we apply deep learning (U-Net) to detect and measure defects in fiber-reinforced composites.
Team: Dr. Paulina Díaz-Montiel (Mechanical Engineering), Julia M. Greely
Automated defect segmentation in carbon fiber composites. With Mechanical Engineering and Computer Science at USD, we extend this work to segment several defect types (voids, resin-rich regions and cracks) in microscopy images of carbon fiber-reinforced polymers, to replace slow, subjective manual inspection.
Team: Dr. Paulina Díaz-Montiel (Mechanical Engineering), Dr. Md Nafee Al Islam (Computer Science), Arnav Chaturvedi