Trajectory Planning Library for Autonomous Robots

This repo implements multiple trajectory optimization methods, such as min-snap trajectory planner, ViGO (our local trajectory planner), based on the occupancy voxel map and the Octomap for autonomous robots.

Author: Zhefan Xu, Computational Engineering & Robotics Lab (CERLAB) at Carnegie Mellon University (CMU).

If you find this work helpful, kindly show your support by giving us a free ⭐️. Your recognition is truly valued.

This repo can be used as a standalone package and also comes as a module of our autonomy framework.

I. Installation Guide

This repo has been tested on ROS Melodic with Ubuntu 18.04 and ROS Noetic with Ubuntu 20.04 and it depends on map_manager which provides the occupancy voxel map implementation and octomap for octree-based map. It also depends on global_planner for global waypoint generation.

# install dependency
sudo apt install ros-[melodic/noetic]-octomap* # octomap

cd ~/catkin_ws/src
git clone https://github.com/Zhefan-Xu/trajectory_planner.git

cd ~/catkin_ws
catkin_make

II. Run Planner DEMO

a. The example of the local trajectory planning using the B-spline-based trajectory optimization (ViGO):

Please modify the correct path for the prebuilt map in cfg/bspline_interactive/occupancy_map.yaml. All the prebuilt maps are under trajectory_planner/map.

roslaunch trajectory_planner bspline_interactive.launch

Use 2D Nav Goal in Rviz to select start and goal position in the map as shown below:

bspline_demo.mp4

The related paper can be found on:

Zhefan Xu, Yumeng Xiu, Xiaoyang Zhan, Baihan Chen, and Kenji Shimada, “Vision-aided UAV Navigation and Dynamic Obstacle Avoidance using Gradient-based B-spline Trajectory Optimization”, IEEE International Conference on Robotics and Automation (ICRA), 2023. [paper] [video]

b. The example of the global trajectory planning using the min-snap trajectory optimization with the global planner:

roslaunch trajectory_planner poly_RRT_goal_interactive.launch

Use 2D Nav Goal in Rviz to select start and goal position in the map as shown below:

min_snap_demo.mp4

III. Code Exmaple & API

a. For the B-spline-based trajectory optimization (ViGO), the example code can be found in trajectory_planner/src/bspline_node.cpp

b. For the min snap trajectory planner, the following code explains how to use the trajPlanner::polyTrajOctomap to generate collision-free trajectory from waypoints:

#include <trajectory_planner/polyTrajOctomap.h>

int main(){
    ...
    ros::NodeHandle nh;

    // Initialize Planner
    trajPlanner::polyTrajOctomap polyPlanner (nh); // trajectory planner

    // Load Waypoint Path and generate trajectory
    polyPlanner.updatePath(path); // path: nav_msgs::Path
    polyPlanner.makePlan();

    // Get pose from trajectory
    polyPlanner.getPose(t); // get pose at time t
    ...
}

You can check src/poly_RRT_node.cpp for more details.

IV. Parameters:

All planners parameters can be edited and modified in trajectory_planner/cfg/***.yaml.

V. Issues

If you cannot visulize the Octomap with ROS noetic, that is because the octomap_ros package has some tf name incompatibility issue. To solve that, please try building the octomap_ros from source with the following steps:

Step1: download the octomap mapping source files

cd ~/catkin_ws/src
git clone https://github.com/OctoMap/octomap_mapping.git

Step2: modify source files and catkin make Please change all the frame id /map to map in octomap_mapping/octomap_server/OctomapServer.cpp and octomap_mapping/octomap_server/octomap_server_static.cpp. There should be 2 places.

cd ~/catkin_ws
catkin_make

VI. Citation and Reference

If you find this work useful, please cite the paper:

@inproceedings{xu2023vision,
  title={Vision-aided UAV navigation and dynamic obstacle avoidance using gradient-based B-spline trajectory optimization},
  author={Xu, Zhefan and Xiu, Yumeng and Zhan, Xiaoyang and Chen, Baihan and Shimada, Kenji},
  booktitle={2023 IEEE International Conference on Robotics and Automation (ICRA)},
  pages={1214--1220},
  year={2023},
  organization={IEEE}
}