Need help with mav_active_3d_planning?
Click the “chat” button below for chat support from the developer who created it, or find similar developers for support.

About the developer

ethz-asl
139 Stars 39 Forks BSD 3-Clause "New" or "Revised" License 214 Commits 1 Opened issues

Description

Modular framework for online informative path planning.

Services available

!
?

Need anything else?

Contributors list

mav_active_3d_planning

mav_active_3d_planning is a modular framework for online informative path planner (IPP) design. We provide a modular framework for creating, evaluating and employing primarily sampling based, receding horizon algorithms that optimize a gain while minimizing a cost.

Online-IPP for Exploration (left), 3D Reconstruction (right) & more. git_gif

Table of Contents

Credits * Paper and Video

Setup * Packages * Dependencies * Installation * Data Repository

Examples * Configuring a Planner * Run an Experiment

Documentation * Planner Structure * Planner Design Framework * Running and Evaluating a Simulated Experiment * Code Index

For additional information please see the wiki.

Credits

Paper and Video

If you find this package useful for your research, please consider citing our paper:

  • IEEE | ArXiv | Video

  • Lukas Schmid, Michael Pantic, Raghav Khanna, Lionel Ott, Roland Siegwart, and Juan Nieto, "An Efficient Sampling-based Method for Online Informative Path Planning in Unknown Environments", in IEEE Robotics and Automation Letters, vol. 5, no. 2, pp. 1500-1507, April 2020

    bibtex
    @ARTICLE{Schmid20ActivePlanning,
    author={L. {Schmid} and M. {Pantic} and R. {Khanna} and L. {Ott} and R. {Siegwart} and J. {Nieto}},
    journal={IEEE Robotics and Automation Letters},
    title={An Efficient Sampling-Based Method for Online Informative Path Planning in Unknown Environments},
    year={2020},
    volume={5},
    number={2},
    pages={1500-1507},
    keywords={Motion and path planning;aerial systems;perception and autonomy;reactive and sensor-based planning},
    doi={10.1109/LRA.2020.2969191},
    ISSN={2377-3774},
    month={April},
    }
    
  • The planner presented in the paper is given in

    active_3d_planning_app_reconstruction/cfg/planners/reconstruction_planner.yaml
    .

Setup

Packages

The mavactive3d_planning package is divided into separate packages, such that only the dependencies necessary for your application package need to be built.

Although packages are organized for the catkin workflow, the core package can be built as a stand-alone library for non-ROS use. All packages with a short description are listed below.

Dependencies

Packages and their dependencies: * core:

Central logic of active_3d_planners. Dependencies: *

catkin_simple
(https://github.com/catkin/catkin_simple) *
glog_catkin
(https://github.com/ethz-asl/glog_catkin) *
eigen_catkin
(https://github.com/ethz-asl/eigen_catkin)
  • ros:

Interface to ROS for the general active_3d_planner and ROS specific modules.

  • mav:

Modules and interfaces specific to Micro Aerial Vehicles (MAV), using ROS. Dependencies: *

mav_trajectory_generation
(https://github.com/ethz-asl/mavtrajectorygeneration)
  • voxblox:

Using voxblox as map representation and modules specific to voxblox. Dependencies: *

voxblox
(https://github.com/ethz-asl/voxblox)
  • app_reconstruction:

Application package for autonomous 3D reconstruction with MAVs, including automated simulation and evaluation routines. Dependencies: *

unreal_cv_ros
(https://github.com/ethz-asl/unrealcvros) *
rotors_simulator
(https://github.com/ethz-asl/rotors_simulator) *
mav_control_rw
(https://github.com/ethz-asl/mavcontrolrw)

Installation

Installation instructions for Linux.

Prerequisites

  1. If not already done so, install ROS (Desktop-Full is recommended).

  2. If not already done so, create a catkin workspace with catkin tools:

shell script
sudo apt-get install python-catkin-tools
mkdir -p ~/catkin_ws/src
cd ~/catkin_ws
catkin init
catkin config --extend /opt/ros/melodic  # exchange melodic for your ros distro if necessary
catkin config --cmake-args -DCMAKE_BUILD_TYPE=Release
catkin config --merge-devel

Installation

  1. Move to your catkin workspace:
    shell script
    cd ~/catkin_ws/src
    
  1. Install system dependencies:
    shell script
    sudo apt-get install python-wstool python-catkin-tools
    
  1. Download repo using a SSH key or via HTTPS:
    shell script
    git clone [email protected]:ethz-asl/mav_active_3d_planning.git # SSH
    git clone https://github.com/ethz-asl/mav_active_3d_planning.git # HTTPS
    
  1. Download and install the dependencies of the packages you intend to use.
  • Full Install: dependencies of all packages can be installed using rosinstall: ```shell script # system dependencies, replace melodic with your ros distro if necessary: sudo apt-get install ros-melodic-cmake-modules ros-melodic-control-toolbox ros-melodic-joy ros-melodic-octomap-ros ros-melodic-mavlink ros-melodic-geographic-msgs autoconf libyaml-cpp-dev protobuf-compiler libgoogle-glog-dev liblapacke-dev libgeographic-dev pip install future unrealcv

# If you already intialized ws tool use 'wstool merge -t' wstool init . ./mavactive3dplanning/mavactive3dplanningssh.rosinstall # SSH wstool init . ./mavactive3dplanning/mavactive3dplanninghttps.rosinstall # HTTPS wstool update ``

   * **Partial Install:** Install dependencies of the packages you intend to use ([listed above](#Dependencies)) and remove unwanted packages from
mavactive3d_planning/package.xml` as well as their source folders.
  1. Source and compile:
    shell script
    source ../devel/setup.bash
    catkin build mav_active_3d_planning # Builds this package only
    catkin build # Builds entire workspace, recommended for full install.
    

Data Repository

Related ressources, such as experiment scenarios and ground truth point clouds, can be downloaded from here.

Examples

Configuring a Planner

The

active_3d_planning_app_reconstruction
is an application package, that launches an active_3d_planner. A verbose example of how planner configurations are specified is given in
cfg/planners/example_config.yaml
. The example planner uses local motion primitives to expand new segments and the number of unknown voxels as gain formulation. To see the planner in action, start an unreal_cv_ros game, e.g. CityBuilding, make sure to tab out of game control (Alt+Tab for Binary, Ctrl+Shift+F1 for Editor) and then run
roslaunch active_3d_planning_app_reconstruction example.launch
The planner will be built from the config file and visualized in RVIZ. A useful parameter to set is
verbose_modules: true
, as all available params of all built modules will be printed to console.

mav_3d_ex_config

A local motion primitive based planner starting exploration.

Note: The reconstruction planner used in the paper is not the demo one and is given in

active_3d_planning_app_reconstruction/cfg/planners/reconstruction_planner.yaml
. You can run it using
roslaunch active_3d_planning_app_reconstruction example.launch planner_config:=planners/reconstruction_planner.yaml

Run an Experiment

In order to record data of the example planner, run

roslaunch active_3d_planning_app_reconstruction run_experiment.launch data_directory:=/path/to/my_data_dir
This will collect and store raw data in a new folder in
my_data_dir
. When the experiment has finished by time limit (30 minutes) or by pressing Ctrl+C, run
roslaunch active_3d_planning_app_reconstruction evaluate_experiment.launch target_directory:=
/path/to/my_data_dir gt_file_path:=/path/to/CityBuilding/gt_surface_pcl.ply
to evaluate the raw data. When the process is finished, the created data directory contains a folder 'Graphs', containing the evaluation results as well as a folder 'Meshes', which can be visualized using e.g. CloudCompare.

SimulationOverview Performance overview of the planner over the course of the simulated experiment.

PerformanceOverview Distribution of computation time for the different modules. The majority is expended for gain computation (red). Notice that the voxblox map serialization increases as the map grows (dark grey).

mav_active Final reconstruction mesh and error coloring visualized in CloudCompare.

We use cookies. If you continue to browse the site, you agree to the use of cookies. For more information on our use of cookies please see our Privacy Policy.