Quote:
Originally Posted by IceFire
My hypothesis is that humans are humans regardless of the situation be it real life or simulation. When presented with the same basic parameters they will do the same basic things on aggregate. In the heat of combat, target fixation becomes a concern and two guys shooting at the same burning plane can (and did) happen. There's even a gun camera shot that I saw where one Hellcat's gun camera captures another Hellcat fly in front of him while what appears to be a Zero is getting shot at by both. So, while it's less likely that point whoring would go on in real life (at least not like we see it)... target fixation looks fairly similar. At least some of the notorious kill stealing and shoulder shooting is caused by total target fixation. You can watch it happen. They are oblivious to all around them.
|
If both pilots were in or not much short of shooting distance, and one lights up the plane, the other then shoots it, too, that would be okay, and that is explainable as target fixation.
But AI gets target fixated from miles out, one shoots up a bomber, lights it up - and the other AI is still x km out, passes a few other bombers on its way -only to shoot on the already burning bomber.
Quote:
Originally Posted by IceFire
The problem with the AI right now is that they have a switch. If the target has bailed out or lost a wing then the aircraft is considered dead. If the plane is burning but still flying then it's considered a viable target. What needs to be programmed is a a greater range of recognition.
|
That. And no more 1/0 switch. Humans don't always react the same way in similar situations -and AI should not, too. Have the 1/0 switch be the trigger, the result not granted, but randomised.
Say "enemy plane burning" triggers: check distance to border AND check other enemys near AND check own ammo:
-distance to border greater than 30 km AND enemys near AND own ammo <50%: stop shooting 90%
...10 other cases...
-over enemy country AND no enemy near AND own ammo plenty: continue shooting 80%
Net result: AI behaves less predictable - same AI in same situation may even do things different