iPlanner ROS 2

This is a ROS 2 port of the iPlanner package, integrated into the Autonomous Exploration Development Environment.

Overview

iPlanner is a deep learning-based path planner that generates local paths from depth images. This package migrates the original ROS 1 functionality to ROS 2 Humble.

Dependencies

• ROS 2 Humble
• Python 3.10+
• PyTorch (tested with 2.5.1+cu124)
• Torchvision
• NumPy (<2.0)
• OpenCV
• cv_bridge

Installation

1. Clone specific branch (if applicable) or copy to your src folder.

   Note: This repository uses Git LFS for large model files. Ensure you have git-lfs installed:

   sudo apt-get install git-lfs
   git lfs install

2. Install dependencies:

   rosdep install --from-paths src --ignore-src -r -y

3. Build the package:

   colcon build --symlink-install --packages-select iplanner_ros2

Usage

0. Download .pt file

Download the pre-trained network weights plannernet.pt here and put it into the models folder. Noted this pre-trained network has not been adapted to real-world data.

1. Start Simulation

First, launch the vehicle_simulator with the desired environment:

ros2 launch vehicle_simulator system_garage.launch

2. Launch iPlanner

In a separate terminal, source the workspace and launch the planner:

source install/setup.bash
ros2 launch iplanner_ros2 iplanner.launch.py

3. Send Goals

You can check iplanner_node status via:

ros2 topic echo /ip_planner_status

Send a goal using Rviz "2D Nav Goal" or publish to /way_point (PointStamped).

Topics

• Subscribed:

  • /camera/depth/image_raw (sensor_msgs/Image): Depth image input
  • /way_point (geometry_msgs/PointStamped): Goal point
  • /joy (sensor_msgs/Joy): Joystick input (optional)

• Published:

  • /path (nav_msgs/Path): Generated path
  • /path_fear (nav_msgs/Path): Alternative path when "fear" (obstacle) is detected
  • /ip_planner_status (std_msgs/Int16): Planning status (0: Planning, 1: Success, -1: Fail)

Known Issues / Notes

• Visualization: The original iplanner_viz node has not been ported. Use Rviz to visualize the /path and /camera/depth/image_raw topics.
• Model File: The pre-trained model plannernet.pt is loaded from src/iplanner_ros2/iplanner_ros2/models/.

Original Citation

If you use this work, please cite the original paper:

@article{yang2023iplanner,
  title={iPlanner: perception-aware path planning for autonomous navigation in complex environments},
  author={Yang, Fan and Zhang, Tingrui and Hutter, Marco},
  journal={arXiv preprint arXiv:2309.02700},
  year={2023}
}