4. Conclusion

As we have shown in this article, SLAM techniques are extremely varied, in terms of the representations produced, the algorithmic techniques employed and the sensors used to implement them. It should be borne in mind that the selection of methods presented here is extremely partial and is essentially intended to give a broad overview of the fundamental techniques, overlooking many related problems such as the association of points of interest in vision or loop closure detection, to name but two.

Most of the methods mentioned have Open Source implementations available on the OpenSLAM website, which also contains many other representatives of the three categories of methods presented. The ROS middleware can also be used to test some of these approaches in simulation or on real robots.

To give an idea of the level of maturity of the different...