Hi, I've been reading some papers, mainly to do with mapping and navigation and there is a lot of mention of Kalman filters. From the context of the papers I assume these filters are able to help a robot in mapping its environment because, when using sonar - possibly other sensors I'm not sure - if a location is visited a signature of that location will be accurate and more likely to be unique? I could be very wrong here. Like I say I've only just started looking into these.
This has raised a couple of questions:
1) With my current robot, could it benefit from using a KF on the sonar readings. 2) Do the reading have to accumulate over time to be of any use, and if so how long. I assume it can't be the case that the robot have to keep visiting the same place time and again to ensure useful values. 3) Can a KF be used in read time to simply help filter out sonar inaccuracies? I.E. if it just wandering about which out any specific purpose. 4) I plane to use a neural network as a navigation tool, would KF data be on any benefit here? Or do NN's and KF's amount to the same thing in this respect? - I have some experience of NN's and in some ways the KF seems as thought it might have similar properties.I would be most grateful for any advice.
Regards
Mark