012021-2025
Ph.D. in Mechanical Engineering
University of South Carolina
Thesis / dissertationRecognizing the Unexpected: Deep Learning Across Complex Environments
Academic profile · 2026
Ge Song
Postdoctoral Research Associate
Department of Mechanical Engineering
University of Connecticut
I am a mechanical engineering researcher developing artificial intelligence methods for manufacturing, robotics, and complex physical systems. My work connects data-driven learning with real-world sensing, physical behavior, and deployable engineering applications.
Artificial Intelligence
Intelligent Manufacturing
Robotics
Physical Systems
01 / EDUCATION
Academic path
A mechanical engineering foundation shaped by robotics, deep learning, and intelligent systems.
022019-2021
M.S. in Mechanical Engineering
Boston University
Thesis / dissertationComparison of Tracking Algorithm for Differential Drive Robots
032015-2019
B.S. in Mechanical Engineering
Nanjing University of Science and Technology
Thesis / dissertationDesign of Automatic Sorting System for Parts
02 / NEWS & UPDATES
Recent highlights
Selected milestones from my latest research and scholarly activities.
Paper accepted
My first-authored paper, “Physics-Distilled Neural Network enabled by Large Language Models for Manufacturing Process-Property Predictive Modeling,” was accepted for publication in the Journal of Computing and Information Science in Engineering.
Paper accepted
My first-authored paper, “Seeing with Words: Auto-Caption-Guided Graph Transformer for Remote Sensing Segmentation,” was accepted for publication in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing.
Conference presentation
I presented “Physics-Distilled Neural Network With Large Language Models for Process-Property Mapping in Manufacturing Processes” at ASME IDETC/CIE 2026 in Houston, Texas.
03 / RESEARCH
Selected projects
AI methods grounded in real-world sensing, physical constraints, and deployable systems.
01 / ONGOING · GEN-AI
Generative AI for Battery Electrodes
Developing data-efficient, physics-aware methods that reuse local microstructure knowledge to support analysis of larger electrode domains.
- Ongoing Research
- Generative AI
- Physics-Aware ML
02 / EDGE-AI
Manufacturing Health Monitoring
A side-channel energy auditing system for real-time anomaly detection in robotic and 3D-printing processes on edge devices.
- Edge Computing
- Digital Manufacturing
- Anomaly Detection
03 / ONGOING · SAFE-MOBILITY
Geosocial Risk Modeling
An ongoing explainable spatial-AI project connecting physical, community, and regional context to proactive railroad-crossing safety decisions.
- Ongoing Research
- Spatial AI
- Explainable AI
04 / VISION
Abnormal Situation Awareness
A real-time safety platform that learns normal pedestrian motion from skeleton trajectories, then detects and localizes risky behavior at railroad grade crossings.
- Computer Vision
- Deep Learning
- Real-time Systems
04 / PUBLICATIONS
Research record
Seeing with Words: Auto-Caption-Guided Graph Transformer for Remote Sensing Segmentation
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, Early Access
Physics-Distilled Neural Network enabled by Large Language Models for Manufacturing Process-Property Predictive Modeling
arXiv preprint, arXiv:2606.11605
A hybrid transformer deep Q-network for path planning
Neurocomputing, vol. 696, 134135
GPU-enabled decentralized, multi-robot path planning based on global evolutionary dynamic programming and local particle swarm optimization
Expert Systems with Applications, vol. 321, 132321
A memory and retrieval transformer-based unsupervised learning model for anomaly detection and segmentation
Pattern Recognition, 113004
Graphics processing unit-enabled path planning based on global evolutionary dynamic programming and local genetic algorithm optimization
Applied Soft Computing, 113167
An energy auditing system based on learnable positional embedding and edge computing for in-situ monitoring of 3D printing processes
The International Journal of Advanced Manufacturing Technology, 1-17
STGCN-PAD: A spatial-temporal graph convolutional network for detecting abnormal pedestrian motion patterns at grade crossings
Pattern Analysis and Applications, vol. 28, article 2
A dual-contrastive-attention transformer for unsupervised anomaly detection in Lamb waves structural health monitoring
Annual Conference of the PHM Society, vol. 17
U-TFF: A U-Net-based anomaly detection framework for robotic manipulator energy consumption auditing using fast Fourier transform
Applied Sciences, vol. 14, no. 14, 6202
Narrowing signal distribution by adamantane derivatization for amino acid identification using an alpha-hemolysin nanopore
Nano Letters, vol. 24, no. 5, 1494-1501
Hybrid path planning based on adaptive visibility graph initialization and edge computing for mobile robots
Engineering Applications of Artificial Intelligence, vol. 126, 107110
Energy consumption auditing based on a generative adversarial network for anomaly detection of robotic manipulators
Future Generation Computer Systems, vol. 149, 376-389
Analysis of abnormal pedestrian behaviors at grade crossings based on semi-supervised generative adversarial networks
Applied Intelligence, vol. 53, no. 19, 21676-21691
An energy consumption auditing anomaly detection system of robotic manipulators based on a generative adversarial network
Annual Conference of the PHM Society, vol. 15
A deep generative adversarial network (GAN)-enabled abnormal pedestrian behavior detection at grade crossings
IEEE SoutheastCon 2023, 677-684
A deep learning framework for detecting and localizing abnormal pedestrian behaviors at grade crossings
Neural Computing and Applications, vol. 34, no. 24, 22099-22113
05 / CONTACT
Let's connect ideas
to the physical world.
I'm interested in research conversations and collaborations across AI, manufacturing, robotics, and data-driven physical systems.