Using 2d histogram filter for localization
-
Sence - When a robot senses, a measurement update happens; this is a simple multiplication that is based off of Bayes' rule. This step is followed by normalization to ensure that the resultant distribution was still vald (and added up to 1 probability).
-
Move - When it moves, a motion update or prediction step occurs; this step is a convolution that shifts the distribution in the direction of motion.
After this cycle, we are left with an altered posterior distribution!
- convolution means adding of beliefs(initial belief + posterior = prior).
- And after convolution we multiply beliefs (prior x posterior) { using bayes rule for conditional probability }
- measurement implies that robot reaches the green door.