For 5B, we started thinking about an approach for model estimation mode based on conversation with Jeff Newman, the developer of larch. I'll soon put my notes into a wiki page to get this task started.
We plan to review pros and cons of the estimation packages and select one for prototyping mode choice and destination choice integration
larch, pylogit, pybiogeme, choicemodels, etc.
choicemodels doesn't do nested logit
pybiogeme now has a pandas version
review ease of use, coding, performance, support community, maturity, etc.