A couple of months ago I announced that x-io Technologies would sponsor FAUST Informatik by providing them with two x-BIMUs with Serial Breakouts. Since then, Hauke and team have made allot of progress and the latest post on the project website includes an impressive demonstration of a RatSLAM algorithm running on the Carolo Cup v.2 platform.
SLAM (in “RatSLAM”) is an acronym for Simultaneous Localisation and Mapping and it encompasses all the processes required for autonomous navigation of a robot in a new environment. If your or I were to walk around a city for the first time, get lost and then recognise a land mark seen earlier in the day, we would instantly be able to understand our location and the build some sort of mental map of the journey we just made. This is effortless for us and other animals but designing a robot that is able to do this effectively is very difficult. RatSLAM is a SLAM algorithm that attempts to recreate the neurological mechanisms observed in rats navigating new environments. It is a great example of biologically inspired AI.
The RatSLAM demonstration uses only two sensory inputs: a camera mounted on the platform and an x-BIMU. The location of the platform is tracked through the optical flow of the video image provided by the camera. The demo video first shows the tracked location of the platform without the RatSLAM algorithm and the estimated location can be seen to drift over time. The second half of the video shows tracking using the RatSLAM algorithm and you can see the error in the estimate location be correct once the platform returns to its starting position.
If you have any questions about this project, please contact Hake or other members of the team through the project website.