Implement a simple Observation Likelihood Model

[elvout]

Use a pure Gaussian distribution as a model for Observation Likelihood.

Compute a probabilistic weight for each particle using the predicted point cloud for that particle.

Implementation Details:

• [x] Use doubles instead of floats for better precision with really small numbers.
• [x] Use log-likelihoods.
• [ ] Need to tune the Lambda hyperparameter.
  • Note that Lambda is infeasible to calculate or measure. We can assume it's constant for a given sensors and a given "type" of environment.

Slides here.

Note that the final proportionality on slide 28 is incorrect. There should be a 1/2 inside the exp().

[elvout]

We actually didn't need lambda, since we're not concerned about the absolute value of the log-likelihood.