M61A1/M61A2 20mm Gatling Gun Velocity
-
okay maybe you are not seeing like i do the level of the problem
an error of like 0.1 deg in the calculation is a problem for AI.
Let’s take an example :
You know what was the BEST method to avoid beeing gun killed by AI before 4.35 => just fly straight (we never told you )
look at this :
It took me literally years and many attempts to be able to manage better the aiming accuracy of the AI against a steady target. Of course i could have artificially increased the dispersion for AI, and problem solved, but this was not very satisfying
another method is to more or less gun, then run a probabilistic code to know if you hit or not….not very satisfying either…
having to fly each bullet is a freaking PITA
Here is 4.35 now:
-
another interesting problem that has been fixed in 4.35 is the wind effect
Watch this :
in 4.34
in 4.35
-
Silly question here. Is it not possible to have a table of known ranges?
This gives me an idea , see you in 3 to 4 weeks
-
This gives me an idea , see you in 3 to 4 weeks
Nice if you can make table but isn’t it going to be huge one?
-
@M79:
Nice if you can make table but isn’t it going to be huge one?
i said it gave me an idea …not that i will make tables
-
i said it gave me an idea …not that i will make tables
OK. You didn’t said you got better idea…
-
I think that BMS should have proper guns modelling.
But it’s not urgent, it’d be fine if it takes 6-8 weeks.
BTW, what’s so hard about teaching AI to use guns? Shouldn’t it be possible to recycle a large portion of the EEGS code? Like giving AI EEGS and then teaching it to put the thing on the thing.
-
I think that BMS should have proper guns modelling.
But it’s not urgent, it’d be fine if it takes 6-8 weeks.
BTW, what’s so hard about teaching AI to use guns? Shouldn’t it be possible to recycle a large portion of the EEGS code? Like giving AI EEGS and then teaching it to put the thing on the thing.
You can imagine that that’s the sort of thing I ´ve tried
When I say it took me years of attempt I mean it and I am not particularly known to be bad in physics modeling
Teaching anything to AI is vey complex
When we want some precision it is necessary to use PID implementation which is not easy because the AI does not use a physics model anyway but a cinematic model instead
Moreover for a human brain putting a dot on a moving target seems natural , but for AI this is far from beeing obvious with so many dofs you have plenty of option to put the dot in the target ….
-
You can imagine that that’s the sort of thing I ´ve tried
When I say it took me years of attempt I mean it and I am not particularly known to be bad in physics modeling
Teaching anything to AI is vey complex
When we want some precision it is necessary to use PID implementation which is not easy because the AI does not use a physics model anyway but a cinematic model instead
Moreover for a human brain putting a dot on a moving target seems natural , but for AI this is far from beeing obvious with so many dofs you have plenty of option to put the dot in the target ….
What is PID?
-
Proportional Integral Derivative. Basically, a control system that uses feedback for continual control.