I just started studying localization for robotics. I started with Markov localization. There was a cool lecture on Udacity. I also came across another localization called hidden Markov localization, but couldn't find many blogs on it. Can anyone explain to me briefly what's the difference between both?
$\begingroup$
$\endgroup$
Add a comment
|
$\begingroup$
$\endgroup$
From some basic googling.
https://dtransposed.github.io/blog/Robot-Localization.html
https://www.cs.ubc.ca/~carenini/TEACHING/CPSC322-10/SLIDES/lecture32-2010.pdf
http://web.mit.edu/16.412j/www/html/lectures/L4_Introduction%20to%20SLAM%20II.pdf
http://www.columbia.edu/~mh2078/MachineLearningORFE/HMMs_MasterSlides.pdf
Also the wikipedia article on Hidden Markov models is pretty clear.
A hidden markov model is just a case where the state/variables you are interested are not directly observable. This is the more common situation as generally you don't have a nice sensor that directly measures the variable you want. Instead you have to infer the hidden state through measuring other visible variables.
For a localization example, you are interested in the robots global position and heading. These are your hidden variables ,because in many cases you don't have a sensor that can directly measure this. Instead you might have something like range measurements to 2D landmarks, and a motion model which predicts how your robot moves. Now neither of them directly measure the hidden state(position,heading), but you can use a series of them to infer the hidden states.
