Histogram filter

A histogram filter is a type of Bayes filter that represents the belief as a histogram; a discretization of the state space with one probability value per discrete states. One of the key assumptions in Bayes filters is the Markov property of states, from which follows that the current belief Bel(st) summarizes all information of the entire history of observations and actions that is relevant for predicting the future.

Other key assumptions determine how the belief is recursively updated using two alternating steps: the prediction step based on the last action At−1 and the measurement update step based on the current measurement O-t. Note that these two sources of information are separated, which results from the assumption of conditional independence of observation and action given the state.

More information:

Robot Localization II: The Histogram Filter

Papers:

End-to-End Learnable Histogram Filters

Finding Location Using a Particle Filter and Histogram Matching

 

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Iterative Closest Point Algorithm

Kalman filter