JackHopkins
mapping actions from human data to fle
As a corollary to this issue, we'll also need to map the actions in human trajectories to our FLE primitive actions.
To kick things off, I think the obvious way to mine human data is to predict the next thing to place down: i.e [existing factory] -> [placement to improve factory]
Any improvements over this?
Not sure if that's enough to be really helpful since there's placement but also configuration (rotation, recipe-setting). I think we just need to look at the event trace from @hrshtt 's process, write an algorithm for parsing that into the actions we've defined, and then run the actions we derived and assert that the ending state is identical
