While Simultaneous Localisation and Mapping is in principle a solved problem, building a reliable SLAM system presents a number of engineering challenges which make it much more than just implementing tracking and mapping algorithms. This article describes some of the more difficult aspects of practical SLAM, and ways in which the various problems can be addressed.
In previous articles we have introduced SLAM (Simultaneous Localisation and Mapping) and discussed some of the issues relating to using SLAM systems in various situations. From reading these, one might be forgiven for thinking that SLAM is basically a solved problem, and that implementing a SLAM system is just a matter of putting together the required components. However, while the core problem of finding the location of a sensor while mapping its environment in real time can be addressed in a variety of ways, there are plenty of difficulties which make building a practical and reliable SLAM system very challenging. This article presents a selection of these problems, from data acquisition through to map refinement, to try to give a sense of the significant engineering challenges and research questions involved when doing SLAM in the real world.
Read the full article here.