External Packages and Nodes

Terminology

• + : interesting topics and hardware abstraction
• ~ : interesting, but quite a lot of work to do for hardware compatibility or mapping
• - : bad solution

Overview

Figure 1: ROS packages overview

3D Mapping

SLAM

Octomap_server : +

3D occupancy grid mapping, independent from sensor, looks like it does not need odometry

• https://youtu.be/yp0f8-AKvDU
• https://wiki.ros.org/octomap_mapping
• https://wiki.ros.org/octomap_server
• http://octomap.github.io/

Plus :

• maintained
• compatible with melodic
• documentation available as well as many
• no odometry
• independent from hardware (only require the right input topics)

Minus :

• ...

Inputs required :

• sensor_msgs/PointCloud2

Hector slam : +

• https://github.com/tu-darmstadt-ros-pkg/hector_slam
• http://wiki.ros.org/hector_slam

Not sure whether we're interested in hector slam itself, or on the

Plus :

• maintained
• not directly compatible with melodic, but easy to build it from source for melodic
• odometry not needed

Minus :

• mostly created for 2D mapping and robot navigation
• not much documentation

Inputs required :

• ...

REMODE : ~

• https://www.ros.org/news/2016/02/open-source-release-remode-probabilistic-monocular-dense-reconstruction-in-real-time.html

modeling of many 3D objects, like rooms, persons, ...

Plus :

• noise reduction
• nice rendering

Minus :

• not much documentation and precisions about hardware/drivers/topics
• maybe "too much" for our needs ?
• looks like it is not maintained anymore : latest commit was 4 years ago

Inputs :

• ...

LOAM

RTABMAP : +

• http://wiki.ros.org/rtabmap_ros

Plus :

• maintained
• compatible with melodic
• real time mapping
• publishes :
• 3D point clouds
• 2D occupancy maps
• tutorials and documentation available

Minus :

• oriented towards robot navigation, although "top-down" modeling seems to be possible

Inputs required :

• odometry (not mandatory in all cases)
• scan 2D or 3D

Spin Hokuyo : +

• https://github.com/RobustFieldAutonomyLab/spin_hokuyo
• http://wiki.ros.org/spin_hokuyo

It creates a point cloud with a 2D LiDaR and a servomotor. The interesting node compiles small point clouds to make one big point cloud. Could be very useful to make our digital model.

Plus :

• has a node that compiles point clouds and publish them on a topic
• great rendering

Minus :

• designed for another sensor, but the node that compiles point clouds does not care about that
• need some odometry work

Inputs required :

• laser scan
• odometry

Lego-LOAM : ~

• https://github.com/RobustFieldAutonomyLab/LeGO-LOAM

Plus :

• good rendering

Minus :

• designed for robot navigation, not for "top-down mapping"
• designed for another sensor (velodyne)

Inputs :

• ...

Velodyne loam : ~

• http://wiki.ros.org/loam_velodyne

Plus :

• good rendering
• builds 3D maps

Minus :

• for velodyne sensor
• robot navigation

Inputs :

• ..

Bad solution -

• https://github.com/koide3/hdl_graph_slam : not what we need. creates maps with corridors and doors, but not "top-down" mapping
• http://wiki.ros.org/robot_pose_ekf : not what we need
• http://wiki.ros.org/ethzasl_icp_mapper : doc not up to date, slowly not maintained anymore, ...
• https://github.com/ethz-asl/libpointmatcher/blob/master/doc/index.md

Modbus

• http://wiki.ros.org/modbus

Object Tracking

Multiple objects lidar tracking : ~

• https://github.com/praveen-palanisamy/multiple-object-tracking-lidar

Plus :

• tracks objects in real time
• hardware independent

Minus :

• 2D maps, most likely used for robot navigation

Inputs :

• ...

Object Detection

• https://www.acin.tuwien.ac.at/vision-for-robotics/software-tools/v4r-library/
• https://rgit.acin.tuwien.ac.at/v4r/v4r_ros_wrappers
• http://wiki.ros.org/object_recognition
• https://www.osrfoundation.org/ros2-object-detection-demo/
• http://wiki.ros.org/find_object_2d

QR code readers

• http://wiki.ros.org/zbar_ros
• https://github.com/mdrwiega/qr_detector
• http://wiki.ros.org/visp_auto_tracker