Select Page

SLAM, Lidar and localization.

Simultaneous localization and mapping (SLAM) is the computational problem of constructing a map while simultaneously keeping track of the LiDAR location within in the map. In simple words you can solve the following tasks: 

  • Creating a point cloud from your Lidar indoors and outdoors without any complex Inertial Navigation Systems (INS) or similar,
  • Real time mapping,
  • Keeping track of your position in real time relative to a map, typically a robot inside a large facility or outside,
  • Tracking changes in the environment,
  • Supporting other navigation means like INS where GNSS conditions are poor. 

If you want a Lidar to work with SLAM Visimind has several solutions to help you:

  • Commercial SLAM,
  • Pre-installed kit with Lidar, computer and Inertial Measurement Unit (IMU),
  • An installation guide to Open-source SLAM.

Commercial SLAM

Visimind can guide you with commercially available SLAM algorithms that we know work well with our Lidars. Some algorithms also operate without IMU and instead relies on a fast updating of the Lidar frames. How SLAM works in practice varies with the use case, and we believe we can help you there. Contact us to discuss your Lidar need and use case. Commercial SLAM comes with easier to use packages, support and also good performance.

Pre-installed kit with Lidar, IMU and computer

This is a ready kit with computer, Lidar of your choice, IMU and Open-source SLAM installed. It’s easy to use but assumes you know Linux. As output you will get a point cloud. The Open-source alternative requires a bit more skills than the commercial packages and you are more on your own when compared to a commerical package. But this kit comes ready to run, connect a screen, power it up, press store and you are ready to see your point cloud!

Installation guide for Open-source SLAM

Visimind has prepared an installation guide with dockers and scripts to install SLAM on a Linux computer with Lidar and IMU. You still need to be familiar with Linux but you should be able to quickly have SLAM up an running to generate your own point clouds. Open-source SLAM is powerful and constantly being developed.

About SLAM

 When using SLAM you will get a point cloud which has a metric coordinate system but it’s not georeferenced, ie. you have no latitude and longitude and no idea of where is north.  The quality of the point cloud will be dependent on factors like:

  • Movement of your Lidar! Move your Lidar smoothly, without sudden jumps or fast rotations.
  • The terrain needs to have some features that the Lidar can track, in an empty field that looks the same everywhere SLAM will not function well.
  • Errors will accumulate while you move to new terrain where you have not been before.
  • If you come back to an already mapped area and the loop closure is working properly the SLAM algorithm will correct backwards and correct partly for the errors that have accumulated. 

Loop closure simply means that the SLAM algotrithm can identify if the location has previously been visitisited and use that that fact to correct the accumulated error, not just at the site but also backwards in time.

Some SLAM algotihms can use GNSS trajectory to add a proper map localisation to your point cloud and trajectory.

Visimind’s PLCT, a backpack system for real-time control and measurements of vegetation close to power lines, using Velodyne Lidar’s Puck sensor.

Typically the desired output from SLAM is a point cloud (LAS, ply…) along with the trajectory. This can be used on drones, in backpack mapping systems, on surveying cars or as a mean to create local maps for different industrial machines like,

  • In forestry make models of the trees before harvesting
  • As support for automated vehicles
  • To make the Lidar data much more dense than single frames

Lidar Odometry and localisation

Lidar odometry is the method of tracking movment from Lidar. So like SLAM but all you are after is the relative movement of the LIdar from time to time as your Lidar moves. During a short period of time this can be very accurate. For example OxTS uses that inside their Inertial Navigation Systems to limit drift and errors when the INS is used in locations with very poor GNSS data.