This repository extends the HKUST-Aerial-Robotics FUEL planner by adding a diffusion-based sampling stage for frontier expansion. Instead of only following detected frontiers, our method samples extra points in free space around each frontier to reduce uncertainty and improve goal selection.
- Diffusion Sampling: Randomly sample points within a radius around each frontier to gather more information.
- Frontier Expansion: Merge diffusion samples with original FUEL frontiers and re-cluster to select higher-value goals.
- Real-Time Map Updates: Subscribe to live point clouds and update the occupancy map continuously.
- ROS Nodes: Separate nodes for diffusion, frontier expansion, and FUEL integration. Parameters can be tuned via launch files.
- Optional GPR Integration: Use Gaussian Process Regression to predict unknown regions if needed.
# 1) Create or go to your Catkin workspace
mkdir -p ~/catkin_ws/src
cd ~/catkin_ws/src
# 2) Clone this repo
git clone git@github.com:romaster93/DUF-Planner.git
# 4) Install dependencies
cd ~/catkin_ws
rosdep update
rosdep install --from-paths src --ignore-src -r -y
# 5) Build
catkin build
Update will continue...