Better support for mixed-observability models

[zsunberg]

Mixed-observability models present a problem for this framework because the observation probability is only positive if the observation matches the new state exactly. We should have a better support mechanism for this.

[lassepe]

Is there a specific strategy that you had in mind here?

[zsunberg]

Not other than what I talked to you about. I think it will involve integrating with the DDN Structure support for POMDPs.jl. We need to think carefully about the DBN for a mixed observability Markov model and what distributions should be used for a particle filter.