DiffBot Slam Package
This package contains launch files and configurations for different simultaneous localization and mapping (SLAM) algorithms to map the environment of the robot in 2D, although some of these algorithms can be used to map in 3D.
1 2 3 4 5
Additional runtime dependencies are:
are added to this workspace using
As you can see this package has lots of dependencies to test different slam implementations and frontier exploration approaches.
To run this package these dependencies need to be installed and are set as
exec_depend in the
package.xml. Currently only
gmapping provides a ROS Noetic Ubuntu package that can be installed directly with:
In case you want to try more advanced SLAM algorithms, such as
cartographer_ros you need the following Ubuntu package dependencies.
Alternatively you can install from source by building the cloned git repository in your catkin workspace.
Take the required installation size into account. For example
karto_slam needs approximately 125MB because it will also install
SLAM stands for Simultaneous Localization and Mapping sometimes refered to as Concurrent Localization and Mappping (CLAM). The SLAM algorithm combines localization and mapping, where a robot has access only to its own movement and sensory data. The robot must build a map while simultaneously localizing itself relative to the map. See also this blog post on FastSLAM.
To use the following slam algorithms, we need a mobile robot that provides odometry data and is equipped with a horizontally-mounted,
fixed, laser range-finder. Viewed on a higher level, every specific slam node of these algorithms will attempt to transform each incoming scan into the odom (odometry) tf frame. Therefore the node will subscribe to the laser
/scan and the
Transforms are necessary to relate frames for laser, base, and odometry. The only exception is
hector_slam which doesn't require odometry for mapping.
The following SLAM implementations are offered using the launch files explained in the next section. It is suggested to start with
gmapping which is used by default.
gmapping: This package contains a ROS wrapper for OpenSlam's Gmapping. The gmapping package provides laser-based SLAM (Simultaneous Localization and Mapping), as a ROS node called
slam_gmapping, you can create a 2-D occupancy grid map (like a building floorplan) from laser and pose data collected by a mobile robot.
cartographer: Cartographer is a system that provides real-time simultaneous localization and mapping (SLAM) in 2D and 3D across multiple platforms and sensor configurations. See the documentation for an algorithm walkthrough.
karto: This package pulls in the Karto mapping library, and provides a ROS wrapper for using it. Karto is considered more accurate than, for example
gmapping(note: for ROS noetic, see
slam_karto) and became open source in 2010.
hector_slam: metapackage that installs
hector_mappingand related packages. The
hector_mappingis a SLAM approach that can be used without odometry as well as on platforms that exhibit roll/pitch motion (of the sensor, the platform or both), such as drones. It leverages the high update rate of modern LIDAR systems like the Hokuyo UTM-30LX and provides 2D pose estimates at scan rate of the sensors (40Hz for the UTM-30LX). While the system does not provide explicit loop closing ability, it is sufficiently accurate for many real world scenarios. The system has successfully been used on Unmanned Ground Robots, Unmanned Surface Vehicles, Handheld Mapping Devices and logged data from quadrotor UAVs.
gmapping which uses a particle filter,
hector_slam are all graph-based SLAM algorithms.
The least accurate SLAM algorithm is
gmapping but it works fine for smaller maps. Use other algorithms, such as
karto if you operate your robot in
larger environments or you want more accuracy.
Another interesing package is
slam_toolbox which provides ROS1 and ROS2
support and is based on the easy to use
karto is the basis for many companies because it provides an excellent scan matcher
and can operate in large environments. Additionally,
slam_toolbox provides tools to edit a generated map and even create a high quality
map using stored data (offline).
cartographer package is currently supported by OpenRobotics and not by Google where it was originally developed.
It is currently also not setup correctly for DiffBot. Using it will result in errors.
This package provides a main launch file named
diffbot_slam.launch which accepts an argument
Depending on its value, different launch files will be included that execute the specified SLAM algorithm using its configuration in the
As mentioned above, every ROS slam package requries messages from the laser-range finder topic. Usually this topic is named
To distinguish possible multiple lidars, the topic of DiffBot resides in its namespace
Therefore, its necessary to remap the
/scan topic to
/diffbot/scan. The following shows how this was done for the
gmapping launch file.
Inside this package in the
it is important to map the
scan topic to laser scanner topic published by Diffbot.
Remappings are done in the node tag. Here, for the
gmapping.launch in the
1 2 3 4 5 6 7 8 9 10 11
Most of the configrations are the same as
For detailed description of what each parameter does, please check the individual package documentation of the different SLAM methods.
Gazebo Simulation Tests
To test SLAM in the Gazebo simulator run the following two launch files in separate terminals.
First run the simulation with:
and in a second terminal execute the SLAM algorithm:
Here you can choose between different algorithms by changing the value of the
Possible values are
gmapping (the default),
The ROS node graph will look like the following:
In the figure we can see that
gmapping subscribes and publishes to
It requires the transformation from
<the frame attached to incoming scans> to the
base_link, which is usually a fixed value,
broadcast periodically by the
Aditionally, it requires the transform from
odom. This is provided by the odometry system (e.g., the driver for the mobile base).
In the case of DiffBot the odometry system consists of EKF fusion data from the motor encoders and the IMU.
The provided tf transforms are
odom that describes the current estimate of the robot's pose within the map frame.
You can read more about the required and provided transforms in the documentation.
In case you get inaccurate maps follow the official ROS troubleshooting guide for navigation.
The so far described mapping approaches require manually steering the robot in the unknown environment. Frontier exploration is an approach to move a mobile robot on its own to new frontiers to extend its map into new territory until the entire environment has been explored.
Other SLAM Packages (for 3D Mapping)
hdl_graph_slam: Open source ROS package for real-time 6DOF SLAM using a 3D LIDAR. It is based on 3D Graph SLAM with NDT scan matching-based odometry estimation and loop detection. This method is useful for outdoor.
- RTAB-Map: stands for Real-Time Appearance-Based Mapping and is a RGB-D SLAM approach based on a global loop closure detector with real-time constraints. This package can be used to generate a 3D point clouds of the environment and/or to create a 2D occupancy grid map for navigation. To do this it requires only a stereo or RGB-D camera for visual odometry. Additional wheel odometry is not required but can improve the result.
- Loam Velodyne: Laser Odometry and Mapping (Loam) is a realtime method for state estimation and mapping using a 3D lidar, see also the forked Github repository for
loam_velodyne. Note that this is not supported officially anymore because it became closed source.
- ORB-SLAM2: Real-Time SLAM for Monocular, Stereo and RGB-D Cameras,
with Loop Detection and Relocalization Capabilities. See
orb_slam2_rosfor the ROS wrapper.
- slam_toolbox: This package provides a sped up improved slam karto with updated SDK and visualization and modification toolsets. It is a ROS drop in replacement to gmapping, cartographer, karto, hector, etc. This package supports ROS1 and ROS2 and is suitable for use in commercial products because it can map large environments. And it provides tools to edit the generated maps. See also the related ROSCon 2019 video.
- A Tutorial on Graph-Based SLAM
cartographerReal-Time Loop Closure in 2D LIDAR SLAM
hector_slamA flexible and scalable SLAM system with full 3D motion estimation.
- A practical introduction to to pose graph slam
- The Normal Distributions Transform: A New Approach to Laser Scan Matching
- RTAB-Map as an Open-Source Lidar and Visual SLAM Library for Large-Scale and Long-Term Online Operation
- LOAM: Lidar Odometry and Mapping in Real-time
- ORB-SLAM2: an Open-Source SLAM System for Monocular, Stereo and RGB-D Cameras