Page 1 of 1
Drone stuck in loop.
Posted: Tue Feb 25, 2020 6:24 pm
by Nemo4809
It flies towards the ocean, run out of power, returns to the nearest roboport, then zips off towards the ocean again. Rinse and repeat.
Any idea how to stop it?
Edit: Never mind, I managed to right click it out of the air. Devs should consider fixing up bot AI so that doesn't happen.
Re: Drone stuck in loop.
Posted: Tue Feb 25, 2020 6:30 pm
by Deadlock989
Nemo4809 wrote: ↑Tue Feb 25, 2020 6:24 pm
It flies towards the ocean, run out of power, returns to the nearest roboport, then zips off towards the ocean again. Rinse and repeat.
Any idea how to stop it?
Don't build logistics networks where that is possibly a thing.
This is like the trains thing, except exactly the same.
Re: Drone stuck in loop.
Posted: Tue Feb 25, 2020 7:33 pm
by coppercoil
This can be fixed. Need an extra condition while choosing nearest roboport: calculate an angle beetwen the robot course and roboport, and accept only those where angle < 90 (no return allowed).
Calculations can be made in chunk precision, to not repeat it for thousands robots individually.
Re: Drone stuck in loop.
Posted: Tue Feb 25, 2020 7:33 pm
by Deadlock989
Or you could not build terrible logistics networks.
Just saying.
Re: Drone stuck in loop.
Posted: Tue Feb 25, 2020 7:42 pm
by ptx0
coppercoil wrote: ↑Tue Feb 25, 2020 7:33 pm
This can be fixed. Need an extra condition while choosing nearest roboport: calculate an angle beetwen the robot course and roboport, and accept only those where angle < 90 (no return allowed).
Calculations can be made in chunk precision, to not repeat it for thousands robots individually.
any kind of change like this that makes bots better while nerfing their UPS advantage is fine by me, bots are the worst. but you won't be able to use millions of them with this kind of change, which makes them better.
Re: Drone stuck in loop.
Posted: Tue Feb 25, 2020 8:18 pm
by coppercoil
ptx0 wrote: ↑Tue Feb 25, 2020 7:42 pm
you won't be able to use millions of them with this kind of change, which makes them better.
Why not? Just precalculate chunk route tables in, say, 30 degree precision, that would be enough.
I believe actual implementation wouldn't be so easy, but you got the point: calculate once.