Benchmark for UAV Visual Localization under Low-altitude Multi-view Observation Condition

An offical repo for ECCV 2024 Towards Natural Language-Guided Drones: GeoText-1652 Benchmark with Spatial Relation Matching

Codebase of ACL 2023 Findings "Aerial Vision-and-Dialog Navigation"

Autonomous aerial vehicle localization, path planning and navigation towards full and optimal 3D coverage of a known environment

OctaIndex3D is a high-performance 3D spatial indexing and routing library based on a Body-Centered Cubic (BCC) lattice with truncated octahedral cells.

A reinforcement learning-based drone navigation system using Unity ML-Agents and Stable Baselines3. The drone learns to autonomously pass through gates while avoiding obstacles using PPO.

Python TERCOM (Terrain Contour Matching) Demo for GPS-Denied Navigation

PyTorch3D-based differentiable drone simulator — train vision-based flight policies end-to-end via gradient descent. Extends DiffPhysDrone with CMA-ES, multi-sensor fusion, and 10 policy architectures.

Python DSMAC (Digital Scene Matching Area Correlator) Demo for GPS-Denied Navigation

Experimental simulation environment for drone navigation.

BE Project: High-Speed Drone Navigation for Emergency Response - Oisín Moran, University College Dublin

3D Autonomous Drone Navigation using A*, RRT*, Dynamic Replanning, and PID control implemented from scratch in Python.

Simulate real-time drone interception with advanced A* pathfinding! Solve massive-scale mazes in milliseconds. Optimized for precision, speed, and scalability—powered by theoretical satellite tracking and enemy mapping. Built in Java for mission-grade performance.

A reinforcement learning-based 2D drone gate navigation system built from scratch using PyTorch.

gDroneMapper — 🇺🇸 Maps DNG, JPG, and MP4 files from DJI drones into a single portable HTML map, extracting telemetry for precise pin location with heading line and cones of view (POV) — 🇦🇷 Mapea DNG, JPG y MP4 generados por drones DJI en único HTML portable extrayendo telemetría para pines precisos con línea de dirección y conos de visión (POV)

This project provides a proof-of-concept Raspberry Pi Visual Positioning System (VPS) that generates simple non-GNSS geolocation through LoFTR matching a query image against offline Google Maps tiles.

DQN-based path-planning component for search-and-rescue drone navigation, with a custom 3D environment, geodetic mission generation, and ArduPilot/QGroundControl waypoint output.

Main MS thesis hub for Deep Reinforcement Learning-Based Autonomous Drone Navigation, tracking research pipeline, setup notes, progress, and repository ecosystem.

This project geolocates a nadir drone query image by matching it against Google Maps tiles and extracting an estimated location. It is designed to run on edge compute.