Nazrin Gurbanova

About Me

I am a graduate student at James Clark School of Engineering, University of Maryland, specializing in Robotics.

My focus areas include ROS2, computer vision, motion planning, and deep learning. I am passionate about advancing intelligent systems and automation by integrating AI with real-time robotics applications.

Download CV

My Tools

Projects

Pick and Place Using a Mobile Base Manipulator

A UR3e manipulator arm on a mobile base, designed for autonomous pick-and-place tasks. It uses ROS2, forward and inverse kinematics, and sensors (LiDAR, odometry) for navigation and object handling. The system aims to automate tasks in industries like manufacturing, offering increased efficiency and autonomy.

TrackAI Human Detector and Tracker

A human detection and tracking system designed for autonomous robots, using YOLOv8 for real-time human detection and CSRT / SORT for continuous tracking. It transforms camera coordinates into the robot's frame for precise navigation, enabling safe path planning in dynamic environments. The project is implemented with OpenCV and relies on real-time processing for collision avoidance and task execution.

Finder Bots Swarm Intelligence

An autonomous multi-robot system developed with ROS and TurtleBots to optimize warehouse operations. Robots collaborate to autonomously locate misplaced items using SLAM, Aruco markers, and computer vision, ensuring efficient navigation and retrieval. The system improves operational efficiency, reduces retrieval costs, and provides real-time updates to operators.

A* Pathfinding in Obstacle Space

Implementation of A* algorithm on a Differential Drive TurtleBot3 robot to navigate it from a given start point to a given goal point. The implementation involves considering the differential drive constraints while planning the robot's motion.

Dijkstra Algorithm

Navigation of a point robot from a designated start position to an end position within an environment containing obstacles. The goal is to ensure the robot avoids contact with obstacles while using the Dijkstra algorithm to achieve successful navigation.

RRT, RRT, and RRT-Smart Path Planning

A comparative study of RRT, RRT*, and RRT*-Smart algorithms for mobile robot path planning. The project explores their efficiency in finding collision-free paths, analyzing path cost, runtime, and convergence rate in complex environments.

Path Tracking Control of an Autonomous Vehicle using MPC and PID Controllers

The use of Model Predictive Control (MPC) and PID controllers for precise trajectory tracking in autonomous vehicles. It highlights the comparison of control strategies through simulation results, offering insights into advanced vehicle dynamics and steering systems.

Paper Corner Detection and Boundary Overlay

Implemented a computer vision pipeline using OpenCV to detect paper corners and overlay boundaries in a video. Key steps include video frame extraction, blurry frame skipping via Variance of the Laplacian, background segmentation, edge detection with Hough Transform, and corner detection using Harris Corner Detector. The output video highlights boundaries and corners, providing real-time visualization of the detection process.

Image Stitching (Panoramic Mosaic)

Developed a panoramic mosaic from four overlapping images of a far-away building. Utilized feature extraction with SIFT, feature matching using RANSAC to filter outliers, homography computation for image alignment, and image stitching. Applied post-processing to crop black borders, resulting in a seamless panoramic image.

Object Trajectory Tracking using Least Squares

Utilized OpenCV for frame extraction and color-based object detection to compute centroids. Applied the Standard Least Squares method to fit a parabolic trajectory and predicted the object's y-coordinate for a given x-value. Visualized the trajectory on the video frames for clear representation.

Image Collection and Calibration Pipeline

Collected approximately 50 images of a calibration board, detected corners using a chessboard pattern, and optimized intrinsic and extrinsic camera parameters. Analyzed reprojection errors to evaluate accuracy and visualized detected points before and after calibration.