My case is different, I am not using the sensors as for localization problem, I am using the 1D Kalman and Bayesian filter to combine/fuse multiple distributions together.
The data I am getting is based on the article PDDR: a statistical time-domain damping parameter for structural damage identification. So, I get the probability distributions of several cases and I tend to fuse each distribution together of similar case. I managed to achieve that with the 1D Kalman filter, but I am now trying to achieve the same with Bayesian filter. My aim is concentrating on the sigma parameter since this is important parameter.