What I want to do is to create an algorithm that gives the most probable position at each moment. I have a (discretized) map, and for every input I know the probability of each node. I also know the probabilities of going from each node to another, so for the next input I have two sources of information to compute the probability of each position with: The last input and the new one.
The more inputs I get, the more precise I can get with computing the most probable trajectory and, therefore, the most probable position at that time. I think I might have to use some sort of modified Viterbi Algorithm, but I can't seem to find nothing remotely similar to what I'd need.
The problem is the following: we have an individual moving on a 2-D space, and every few seconds it emits a signal. When we recive the signal, we know some information about the position of the individual (in relation to us), so we'll have some areas with more probability than some others (and not all probable areas are necesseraly connected). With only one signal, the information we get is basically useless, but since we also know how the individual typically behaves, I'd like to compute the most probable trajectories and, with enough imputs, have a better aproximation.
The nodes should be a discretization of the 2-D space, but I haven't figured out how many I should use yet.
Any ideas on what to start looking?