[REPORTED]AGL oscillations that leave AI stuck in high AoA condition

[ColonelPanic42]

I've attached a mission and track that demos an AI aircraft taking off, ascending toward a waypoint specified in AGL, and then turning toward another waypoint. It seems that in attempt to maintain exact radar altitude the AI commands huge pitch oscillations that bleed speed until they reach 18.5 AoA, at which point they seem to be unable to recover. The exact same flight plan with the waypoint specified as MSL (the altitude MSL is the same in both cases; it's the radar altitude specification that seems to affect the AI). I suspect this could be because the buildings the AI is flying over is causing many sudden changes in their radar altitude.

 

If allowed to use afterburner the AI is able to recover. Somewhat related to https://forums.eagle.ru/forum/english/digital-combat-simulator/dcs-world-2-5/bugs-and-problems-ai/ai-ad/7121294-ai-stuck-at-high-aoa-after-making-sharp-turn-if-afterburner-is-restricted, but the circumstances that get the AI into this bad state are different.

agl oscillation.miz

agl oscillation.trk

[Pikey]

We've seen this for the last 9 months by reproducing with MOOSE GCICAP type scripts which make heavy usage of "RADIO" altitude. It's often easy to get this to happen via scripting. I've no idea why RADIO is even useful at high altitude, it should probably be deprecated as it causes problems that will cause AI to stop responding to instruction, get pegged at 30 degrees AoA and climb slightly until runnning out of fuel. Really the AI should be refactored to use "RADIO" (AGL) when flying very low and "BARO" when their line of sight is unobscured. AI still avoid terrain, its really only useful for getting helicopters to go down into valleys and fly noe, that's it.

[BIGNEWY]

Hi

 

I have reported this to the team to double check what is going on.

 

thanks

[funkyfranky]

Thanks BN. Very much appreciated!