Index   ¦   Tags   ¦   Archives

Tutorial on robotics for developers

I've written a project-based tutorial to introduce other developers to the exciting world of robotics software. By the end of it, you'll have created a fairly complete SLAM system that uses cameras and IMUs for localizing a robot.

The project is built using some of the most common technologies, including ROS and OpenCV and requires some proficiency in C++ programming and Linux. I've assumed no previous knowledge in computer vision or robotics, though.

If you want to have a look before starting to read, the source code is available on GitHub.

The getting started post talks about the different software components in a robotic system and introduces the objective of the project.

The architecture post discusses the different parts of a SLAM (localization) system and how to connect them together with ROS.

The camera post lays down some basic concepts of Computer Vision and non-linear optimization needed to find the camera position when observing a fiducial marker like this one:

The next post talks about some ideas in filtering and estimation theory as a prelude to tracking the camera position as it evolves over time.

We proceed by adding a map to operate on larger environments and fully realize the promise of SLAM.

The final article adds accelerometer and gyroscope data to make the localization more robust when there's little or no information from cameras.

© Nicolò Valigi. Built using Pelican. Theme originally by Giulio Fidente on github.