seq2seq for pathfinding

Can we use these seq2seq matchers for robot planning and pathfinding? I.e. given points A and B, find point C, laying between them. Because dijkstra's is fucking slow and A-star is useless, when space in non-linear (i.e. teleporters are present).

>>1
Pro game programmer here, recently worked on an RTS for a big publisher. One of the things I did was optimize the path finding. We used Jump Point Search Plus (JPS+), which gave us an improvement of 2 magnitudes in performance over A* for our needs.

However, as of a three months ago, a Nintendo programmer iterated on JPS+ and came up with Jump Point Search Plus with Goal Bounding, which can be a few or even several magnitudes faster than A*, depending on how much data you have--it scales way better.

https://github.com/SteveRabin/JPSPlusWithGoalBounding

>>2
Back to Reddit, please!

Just kidding, nice work.

>>2

The compromise with using Goal Bounding is that the Goal Bounding data must be preprocessed offline taking O(n^2) time, thus making it not feasible to support dynamic runtime modifications of the search space (adding or removing edges/walls)
JPS+ is an optimized preprocessed version of Jump Point Search.
JPS+ only works on grid search spaces with uniform cost.

Interesting.
Not bad at all. I've been using A* all the time because A* is easy.

>>1
Whom are you quoting?