Last week I gave an update regarding the progress that I had made with the literature review that I had been conducting into range only localization techniques. The issue with this line of investigation was that it was slightly off topic since we are trying to solve the issue of not having rotation information where as the papers being reviewed were describing various other techniques such as time of flight calculations with beacons.
This week I shifted focus more onto searching for lidar specific localization techniques. The search has been tough for the most part since there are very few papers which contain the key words lidar and localization earlier that 2000. This being said it is important to note that this issue can be extended to the sonar application case as well since the two modalities are fundamentally the same.
Several papers dating back to the 70s and 80s address the localization problem using a sonar ring such as Drumheller, M. (1987). Mobile Robot Localization Using Sonar. In this paper the author addresses the localization issue in four steps:
- Extract straight line segments from the sonar data
- Generate interpretations - as to weather a segment is part of a wall
- Global Model Test - test interpretations against the wall equations
- Sonar Barrier test - refine the selection according to the properties of sonar
It is pertinent to note that in this case a model of the room was known and utilized. This paper therefore addressed the localization issue exclusively as opposed to the full SLAM problem. Another downfall of the method is that it extracts stright line segements for localization which means the system is not robust to curved walls and other discrepancies.
Investigting more papers on this topic will hopefully shed light on the situation and make a solution more clear