Implement default strategy and then optimize it

[manpages]

This is all about getting our feet wet with the actual solvers. We want to have a good baseline solver that performs marginally better than default strategy. Hopefully, learning tricks like when to Flip, at the very least.

[kevroletin]

I think there are several optimizations to the default approach:

• improve bounding box approach by computing bounds for each layer or better for each line
• switch between harmonics (in trivial case we just run solver in low harmonics mode and leverage our solutions DB to pick up cases where it worked)
• draw "thick line of 3 points wide" - here we will save some movements
• trivial sharding: split the map into 4 or 16 equal zones and fill each zone by a separate drone

[kevroletin]

Current progress: there is a (seemingly) working default solver. It was checked on several tasks using the online visualization tool.

[kevroletin]

We've adapted production/pyjs_emulator/run.py to test the solution. We've discovered several failures on first 20 test and investigating where is the problem.

[ixahedron]

So the default solver now finally behaves like a proper default solver and is marginally optimised on top of that, by just returning to 0 faster (in 15-decrement steps) after filling in all the needed voxels. @kevroletin and I continue working on it.

[manpages]

@kevroletin what is the status of this issue?

[kevroletin]

• using a simple bounding box
• draw "thick line of 3 points wide" - here we will save some movements
• return back using long SMoves where done.

AFAIU ixahedron wants to improve it so default strategy can be used as a library function in other solvers.

[kevroletin]

I did "improve bounding box approach by computing bounds for each layer or better for each line" (but, unfortunately as a separate implementation default_solver2.py). It takes a working rectangle as a parameter, so it can be used as a building block for other solvers.

[kevroletin]

Also parallelized to 4 robots. Going to add deconstruction phase and turns the solver into a full solver.

[ixahedron]

Default solver (v1.4) works for all problem types, but no further improvements were introduced. Will focus on integraring @kevroletin 's work