Edge Perception & Temporal Consistency

Jul 1, 2026·
Ömer Kurkutlu
Ömer Kurkutlu
· 1 min read
projects

Overview

This project develops lightweight perception for resource-constrained edge platforms, from Raspberry Pi to microcontrollers. Images are converted into a compact grid-based obstacle representation, and a lightweight temporal smoothing mechanism improves detection consistency under motion blur, occlusion, and lighting changes with virtually no computational overhead.

Key Features

  • Temporally consistent, grid-based obstacle detection
  • Embedded inference with YOLOv8n, SSD-MobileNet v2, and FOMO
  • BLE communication and energy-aware sensing
  • Deployment across desktop GPUs, Raspberry Pi, and MCUs
  • Lightweight Temporal Consistency for Grid-Based Obstacle Detection in Edge Devices. O. Kurkutlu and A. Roohi. IEEE ACSOS-C, 2025.
TinyML Edge AI Embedded AI Object Detection Computer Vision
Ömer Kurkutlu
Authors
Ömer Kurkutlu
PhD Candidate in Electrical and Computer Engineering
I am a Ph.D. candidate in Electrical and Computer Engineering at the University of Illinois Chicago, advised by Dr. Arman Roohi. I specialize in robot learning, Physical AI, and autonomous systems. My research focuses on deep reinforcement learning, vision-language-action models, vision-based navigation, sim-to-real, and resource-constrained robotics, with hands-on experience in NVIDIA Isaac Sim/Isaac Lab, ROS/ROS2, and embedded AI.