![]() |
|
IL-2 Sturmovik: Cliffs of Dover Latest instalment in the acclaimed IL-2 Sturmovik series from award-winning developer Maddox Games. |
![]() |
|
Thread Tools | Display Modes |
|
#1
|
|||
|
|||
![]() Quote:
Yes, Humans can adapt in ways AI cannot. However, probability properly coded can handle a very high percentage probable actions. If AI have full out 100% usage of resources programmed into the AI aircraft the AI can definitely cope with C.E.M. of any human. Deflection shots can also be coded as well. AI aircraft are just complex math equations that can be balanced against the equations of other aircraft. Deflection range and lead can be calculated as well. It is doubtful any gamer software programming would go to that level of precision. It may be done one day, but I don't see it in anything as long as people aren't that concerned to demand it. If we get a reasonably sensible AI performance that is probably as good as can be expected. The tools for most realistic air combat will probably have to come with Online gaming. I am talking about mission builder tools with adjunct ability for things to change on the fly. When everything spins around the Online game, and enough players can be involved simultaneously then we will really enjoy air combat experience. |
#2
|
|||
|
|||
![]() Quote:
This is the one aspect of AI coding that sets the "predictable" apart from the "believable". My main gripe all these years with IL-2 was the predictability of the AI. If I recall it was one of the patches during the Pacific Battles(?). Flying around in my FW I’d ALWAYS end up with “pilot killed” messages even if the bullets were no where near the pit. Assigning stats and random events to the AI would create a believable playing experience for SP and COOPs imo. I might give BoB II another go, I’ve had it sitting on the shelf for about 3 or more years now and installed it only once ![]() |
#3
|
|||
|
|||
![]()
AI is exactly as it is programmed to be.
If you'll add code for unexpected/random behavior, the AI will behave as so. If you'll program AI to learn from his experience, it will learn. And it is not even so complicated, some simple neural networks and genetic algorithms, as the empty 3D space which is he air is much simpler to process than on a terrain mesh with dynamic collisions.. CEM code is also straight forward, and not at all so resource-eating as some are presenting it in here (I can elaborate on this, if you really want to understand, but main point is that AI is already watching his airspeed, altitude, position vector of his own aircraft and of the aircrafts around it, and also has visual and audio sensors. do you really believe that also watching RPM and oil/water temperatures and taking simple if/else or switch decisions is sooo resource-eating?!!!) Stop filling the forum with statements like AI is a huge resource eater nonsense, because most of the uneducated ears will perceive and promote them as truths. They just haven't got themselves to do it, that's all.. |
![]() |
|
|