reported AI (still) detect contacts at extreme/unrealistic distances.

[Justin1Ntime]

Question, when I am playing single player I have a wing man and when we engage bandits he can seem to lock them up and shoot them way before I can with our AIM120s. By the time I lock them up with my radar his missiles are hitting or close to hitting the target. I see the bad guys on my radar via the datalink but it seems like as soon as the wing man is about 40 miles out he can fire and will get a hit. it seems like my radar cant lock them up  until I am about 20-30 miles away and my wing man seem like he can shoot them at twice that distance some times. Am I doing something wrong here? I am running 3 bars and even gone down to one bar with a focus on a single target pointed my plane right at the bad guy. 

[Pizzicato]

I've spent the afternoon running tests on how various factors affect the AI's behaviour (skill level, altitude, target type, target size, etc.). It turns out that the Hornet still detects and begins to engage large targets such as the IL-76 from 370 nm away. Given that the AI should be subject to the same constraints (i.e. a radar screen that only displays out to 160 nm), that's pretty problematic and immersion-breaking level of advantage over the player. It's also problematic from an AI fuel conservation standpoint, as the Hornet immediately goes to full afterburner in pursuit of its very distant prey.

Mission and screenshot attached. (The screen shot shows the moment at which the Hornet detects and starts to maneuver towards the IL-76.

Test CAP 01.miz

Edited November 6, 2022 by Pizzicato

[Exorcet]

While AI sensors are a significant issue that needs to be addressed, the engagement range and fuel conservation problems are more because you're using a general CAP task I assume rather than Search Then Engage, which places range limits on the AI. Any remotely realistic mission would need to use Search Then Engage type tasks. CAP/Fighter Sweep/etc are for simple missions where you just want the AI to shoot stuff ASAP.

[Pizzicato]

Yep, I know. Like most sane people, I typically delete the CAP task as the first step of any A2A group's setup.

This was more just about a raw exploration of what;s going on under the hood and how the AI behaves in its default state.

[Pizzicato]

On a related note, an Su-27 will happily spot and engage an F-15E at 166nm.

Edited November 7, 2022 by Pizzicato

[norman99]

I would have thought including a somewhat realistic, individual unit based "max detection range", especially for AI, would be a core fundamental concept in any modern warfare simulator?

Even if these aren't dynamic and/or situation dependant, something as simple as "F/A-18 Max Detection Range=80nm" etc, would prevent the ridiculous examples above.

Edited December 6, 2022 by norman99

[NineLine]

I'll have a look, thanks. 

[NineLine]

On 11/6/2022 at 4:20 PM, Pizzicato said:

On a related note, an Su-27 will happily spot and engage an F-15E at 166nm.

 

Can you send me a track or mission for this, I set up a F-15 and Su-27 orbiting at 140nm and they never engaged each other. So maybe I am missing something. 

[NineLine]

Actually no worries, I was able to reproduce with other aircraft and at greater distances. This is reported. Thanks.

[NineLine]

Hey guys, those of you that reported this, please check again in the latest version and see if you still see any issues. Thanks!

[Dragon1-1]

Latest version (literally flown a few days ago), the AI F-5E detects and reports targets at 45nm. Not only is it far beyond the capability of the F-5's radar to see a MiG-21 sided target, the range scale doesn't even go past 40nm.

[NineLine]

2 hours ago, Dragon1-1 said:

Latest version (literally flown a few days ago), the AI F-5E detects and reports targets at 45nm. Not only is it far beyond the capability of the F-5's radar to see a MiG-21 sided target, the range scale doesn't even go past 40nm.

SO, only noticed on the F-5E AI? Any others? Thanks!

 

[NineLine]

I have updated the report with the F-5E AI, I was able to reproduce, thanks!

[Pizzicato]

@NineLine I've done some (by no means exhaustive) testing, and the behaviours are generally much improved from when I first reported this. There are still some weird anomalies in there, though. In terms of testing, I kept things very simple:

• Two aircraft flying head-to-head towards each other on the Caucasus map from 250 nm apart.
• The fighters were just given the default CAP task. No tweaks.
• The map was empty, so there were no distractions and no AWACS or EWR support.
• I gauged the detection ranges as the points at which the fighters lit their afterburners and started manoeuvring for position.

My initial findings were:

• Western Fighter versus Transport
  • F-14B v IL-76            - 176nm
  • F-15C v IL-76             - 95nm
  • F-16CM v IL-76         - 73nm
  • F/A-18C v IL-76        - 82nm
• Western Fighter versus Small Fighter
  • F-14B v Mig-29        - 91nm
  • F-15C v Mig-29        - 70nm
  • F-16CM v Mig-29     - 73nm
  • F/A-18C v Mig-29     - 73nm
• Russian Fighter versus Transport
  • Mig-29A v C-130    - 75nm
    Su-27 v C-130         - 75nm
    Mig-31 v C-130       - 110nm
• Russian Fighters versus Transports
  • Mig-29A v F-16     - 75nm
    Su-27 v F-16          - 70nm
    Mig-31 v F-16         - 72nm

I'm not an expert by any means, but these numbers seem much more sane and plausible versus the omniscient AI radars and suicidal charge-across-the-entire-map-on-full-afterburners-and-run-out-of-fuel-before-reaching-their-target behaviours of a couple of years back.

As I mentioned, though, there are still some comparatively minor weirdnesses remaining. As an example, Mig-31 appears to have the radar cross-section of a small skyscraper as both the F-16 and F/A-18 begin engaging it at over 150nm. That seems a little odd to me given that they only "see" an Il-76 at 70-80nm. There are probably other such issues hidden away waiting to be found, but all in all it seems generally much better. Thanks for the heads up.

 

[Ironhand]

On 5/1/2025 at 12:01 AM, Pizzicato said:

• Russian Fighter versus Transport
  • Mig-29A v C-130    - 75nm
    Su-27 v C-130         - 75nm
    …
• Russian Fighters versus [Fighter]
  • Mig-29A v F-16     - 75nm
    Su-27 v F-16          - 70nm
    …

The MiG-29A’s radar is weaker than the Su-27’s. The ranges shouldn’t match vs transports and the MiG-29 shouldn’t “see” a fighter before an Su-27 does in identical situations. So there’s still a problem in that regard.

[Exorcet]

On 5/1/2025 at 12:01 AM, Pizzicato said:

As an example, Mig-31 appears to have the radar cross-section of a small skyscraper as both the F-16 and F/A-18 begin engaging it at over 150nm. That seems a little odd to me given that they only "see" an Il-76 at 70-80nm.

 

Does the MiG have radar on? Maybe that is playing a part.

I have a script that breaks down AI detection by sensor, it might help in figuring out exactly what is going on. It's embedded in a test mission here: