Quote:
Originally Posted by csThor
Quite honestly I am not sure a static "if this happens than that should happen" is what should be the goal. That is canned AI, utterly predictable and after a while players know what the AI will do. I find it difficult to say what an AI should do in a given situation since I'm not sure of the exact circumstances in the first place and I don't really want a "AI reaction manual" with only pre-scripted reactions.
The whole issue is complex, very complex ...  |
Yes and no... probability can be programmed. Human behaviours can be determined with high degree of probability, not perfect of course.
If a pilot is veteran and he is engaged in a certain fashion by 1 or more fighters there is high probability he will make several responses. One choice might require difficult maneuver, another might be a safer move, another a risky move but could be very effective, etc. Yes there should be some reliance on skill for responses, yes there should be reliance upon the condition of the pilot and/or aircraft (tired, injured, damaged, out of fuel).
Many factors need consideration, but the trial and error portions of creating a competent AI performance can be programmed. The probability aspects of various action and responses can also have a low probability of a "random responses" as well.
Simple example of probabilities, if a enemy is diving on you head on from higher altitude at high rate of speed with a lot of E. 30% chance you will turn up your nose and try to hit him on pass which will also lower your E and airspeed, 60% chance you will nose over negative G's and try to gain E and airspeed in opposite direction, 10% chance you will turn hard and maneuver to get behind him, and then there is the random choice - yaw hard and try to deceive the enemy of your flight path.
Yes there are many scenarios that can play out with 360 degree attacks, speed and E, but with enough probabilities, trial and error a good AI performance can be achieved. The package doesn't have to be done all a once. It will have to be developed over time, with help from the community as it was done with the BOB II WOV community.
There is no AI performance package comparable or as competently programmed as the BOB II WOV AI. The sim is over 10 years old and it cannot be a state of the art application with the old code and graphic capabilities.
I would be very content to have a fairly competent AI package to start, which has options for the community to have access to areas of the source code where AI is programmed.
Anyone that thinks a competent AI cannot be done just needs to go the BOB II WOV site, buy the download of BOB II and apply the update patches. Spend some time with the BOB II in action and you will be astonished what can be done.
The BOB II WOV has no multi-player or programmable mission builder, but the campaign is very good. The action is air combat, but that is what the Battle of Britain was for most part.