Open nickumia opened 2 years ago
Starting with the simplest use case:
Forward
Backwards
Left
Right
Previous Pose
(none)Current Pose
(x,y)Goal Pose
(x,y)Move
Stop
A decent amount of progress has been made on the nlp-backend for this. Word associations need to be tighter so that relationships are more meaningful. Preliminary results show that there is too much overlap between words that should be different. Ironically, this isn't wrong, it's just the wrong perspective on the word associations for the problem at hand.
Left
and Right
can be very similar, but for the purposes of choosing a direction, left
and right
are almost complete opposites.
Update: Given the long break in development, I've decided to merge the old code and try to revive/optimize some things while rediscovering pretty much everything I had done in the past haha.. This will take some more time to solidify the codebase.
Some categorizations that I have been trying to save... if this ever gets revived haha
Utilizing the fuzzy meaning algorithm in Lalita, a word connection network can be generated. The idea would be to do the following,
move forward
,go to (x,y)
, ...)move
could related to bothmove forward
andmove backwards
)Future feature,
Translation, Prediction, Meaning Association... all very different, but tied into the problem above.