How to map large indoor spaces with Google Cartographer

When you need to navigate in an indoor environment, you won’t have access to GPS satellites to get your location, and you’ll need to rely on other solutions to know your location. Above all, you will need a software framework that captures the environment and builds a maps out of all possible sensor data. This article focuses on the most popular mapping software at this time of writing: Cartographer.

Cartographer is a real-time simultaneous localization and mapping (SLAM) software designed to create high-resolution, 2D and 3D maps of indoor and outdoor environments.

It uses a combination of sensors including lasers, cameras, wheel odometry and Inertial Measurement Units (IMU), as well as advanced algorithms such as scan matching and graph optimization, to build an accurate map of the environment in real time. While it is primarily aimed at mapping applications, Google's SLAM technology can be used in many areas such as robot navigation and mobile point cloud creation. In addition to accumulating real-world data points in real-time, the system is also able to take maps generated elsewhere—such as CAD models—and use them to construct environments more quickly than using real-world measurements alone. Cartographer has become an essential part of exploring real-world environments with robots or other machines.

Creating an accurate map of a large indoor space such as a warehouse or factory using Cartographer is an achievable task, but one that requires careful planning and design. To create a high-quality map, three major points of attention must be checked:

For this to work properly, calibration of the LIDAR sensor with respect to the wheels and IMUs will ensure correct matching of points between scans.

With these factors in mind, it is possible to create a detailed 3D map of any large indoor space quickly using Cartographer.

For those looking to optimize and tune their maps, there are a few tips that can help ensure success. Firstly, be aware of thin walls that are squeezed to one 2D line or 3D surface, as these can lead to bends in the map. Furthermore, tune the parameters of loop closure searching in submaps so that it is not too rigorous – otherwise you won't get a complete map – nor too relaxed – otherwise the poses won't match up sufficiently. If done properly, then using Google Cartographer can quickly produce dependable maps.

Cartographer is a robust and powerful standalone C++ library, designed to make mapping and navigation easy and efficient. But if you don't want the hassle of setting up Cartographer yourself, there are packages which integrate Cartographer with Robot Operating System (ROS) available. For ROS 1 users, Noetic is the preferred release; if you're running ROS 2, there's also a package available for the Humble release.

If you need additional support, our team at Intermodalics would be happy to provide assistance. Contact us today to get started!

Order Cartographer Support