Infrastructure Systems: Smart, Resilient, and Green Energy Systems

Led by Yufei Tang, Ph.D.

REU Scholar: Briana Deloatch

REU Scholar Home Institution: Lehman College

REU Mentor: Yufei Tang, Ph.D.

Advance AI-Driven Urban Safety Systems for Smart Cities using YOLO9

The Center for Smart Streetscapes (CS3) is dedicated to revitalizing smart cities worldwide using Artificial intelligence for the safety of pedestrians and drivers. This project aims to fulfill that mission using a powerful 2D machine learning algorithm and 3D LiDAR data to estimate the social distance between two people and their surrounding urban environments and ultimately, enhance urban designs and safety. Building on previous work, this project introduced the needed significant advancement in urban safety, including image augmentation, and upgrading from yolov5 to yolov9 to enhance the efficiency of the machine vision algorithm. Utilizing these augmentations makes the model resilient to diverse real-world situations ensuring reliable performance. Including new datasets of cars, trains, and people to broaden the scope of what a real-world scenario looks like for a machine. The system is a combination of 3D Ouster Lidar Data and Velodyne LiDAR Sensors with a 2D machine algorithm that’s trained on a Convolutional Neural Network (CNN) specifically You Only Look Once (YOLO) framework, which is a popular CNN that is known for their ease in training and application process. Integrating different systems and processes replicates urban cities accurately, by identifying human behaviors and vehicles.

REU Scholar: Siyuan Du

REU Scholar: Rithika Mathew

REU Scholar Home Institution: Florida Atlantic University

REU Mentor: Yufei Tang, Ph.D.

PowerGPT: A ChatGPT-like Solution for Smart Grid Data Management

PowerGPT is an innovative AI personal assistant framework designed for secure and efficient big data analysis and management. Focused on enhancing Smart Grid systems, the project addresses the critical need for advanced methods to manage and visualize smart grid data, enabling intelligent prediction and prevention of system faults. The technical objective of PowerGPT is to integrate various advanced features supporting Smart Grid systems into a robust chat engine assistant. Comprehensive case studies ensure optimal performance of each feature, both individually and collectively. By integrating multiple grid data processing features with Natural Language Processing (NLP), PowerGPT bridges impactful fields previously uncombined. A user-friendly NLP interface allows clients to submit prompts and receive relevant feedback. When users’ inputs require more than simple conversation, the system’s framework automatically redirects users to advanced features, such as power transformer fault detection, smart grid geolocation, and wind farm production prediction. This seamless integration of deep learning models within the chat functionality ensures users can access powerful features with limited prior knowledge, and the cloud-based engine allows this software to be accessed virtually anywhere. PowerGPT packages these models and features into an accessible, user-friendly format, with the chat model serving as the central launch point.