projects

Public work first — SIU field UAV localization, open-benchmark air and ground localization, then perception demos. Older wrappers under fun.

work
project thumbnail

GNSS-Denied Visual Localization for UAVs using Satellite Imagery

SIU 2026. SuperPoint–LightGlue matching of UAV camera frames to satellite tiles; 10 m MAE / 14 m RMSE (X–Y) at ~500 m vs RTK, ~5 Hz on Jetson Orin Nano.
project thumbnail

UAV Visual Geo-Localization on Public Benchmarks (OrthoLoC + AerialVL)

Absolute UAV pose from camera-to-orthophoto registration on OrthoLoC and AerialVL. Prior-guided map warping, SuperPoint–LightGlue, DSM-lifted EPnP + RANSAC, and a gated track filter. The open-data counterpart of the SIU 2026 field system.
project thumbnail

Ground Vehicle Localization Under GNSS Dropout (LiDAR–IMU + GNSS)

LiDAR–inertial odometry with gated GNSS fusion on MulRan and UrbanNav. Frame alignment, innovation gating, continuous odometry through outages, and drift / reacquisition metrics on open data.
project thumbnail

Follow-Me Robot Using YOLOv11, SAM2, and Visual Servo Control

A sophisticated visual servoing system combining YOLOv11 object detection, SAM2 segmentation, and image-based visual servo control for autonomous robot following behavior.
project thumbnail

Crazyflie Flight Tests and Experiments

Autonomous flight tests and experiments with Crazyflie 2.1 nano quadcopter, including control system development and real-world flight demonstrations

fun
project thumbnail

LIO-SAM Gazebo ROS2

This repository contains the implementation of LIO-SAM (Lidar Inertial Odometry via Smoothing and Mapping) integrated with ROS 2 (Robot Operating System 2) for robust and accurate SLAM (Simultaneous Localization and Mapping) applications.
project thumbnail

Deep Reinforcement Learning for Mobile Robot Navigation

This project implements **Deep Reinforcement Learning (DRL)** for mobile robot navigation using the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm.
project thumbnail

Obstacle Avoidance with Imitation Learning

This project explores the implementation of obstacle avoidance in autonomous systems using imitation learning. The aim is to train agents to navigate environments while effectively avoiding obstacles.
project thumbnail

YOLOv8 with ROS2

This repository contains the implementation of **YOLOv8** integrated with **ROS 2 (Robot Operating System 2)** for real-time object detection.
project thumbnail

Use of Kalman Filter in Autonomous Vehicles

Kalman filtering for state estimation and prediction of the ego vehicle and surrounding objects from noisy LiDAR, radar, camera and GPS measurements.