AMAI
AMAI copied to clipboard
Can you evaluate this operation? Ranged avoiding melee attackers
【【魔兽微操AI测试】大猎头 vs 骑士-哔哩哔哩】 https://b23.tv/bT1ELsn
This is the work of a Chinese author. He also has several videos,Mainly judgment The unit is attacked
However,There is no decoration on the ground,Slope,Bridge,none, and the Arms single,last,He did not have formation,Unable to concentrate attacks
But I think This is true How much can we learn from it
So there is some behaviour in AMAI to avoid melee units, but its weakened as it caused more issues in tight spaces (mainly the AMAI units would never attack as they constantly dodged attacks instead)
And its not as smooth as shown here, the natural AMAI timer makes it not as efficient.
Need to find a way to do this that doesn't require an insane amount of processing power to be optimal.
So there is some behaviour in AMAI to avoid melee units, but its weakened as it caused more issues in tight spaces (mainly the AMAI units would never attack as they constantly dodged attacks instead)
And its not as smooth as shown here, the natural AMAI timer makes it not as efficient.
Need to find a way to do this that doesn't require an insane amount of processing power to be optimal.
Is there any way to send information about when the game is running to some process? And the reverse also, for example, send data to position units? Well, if there is, maybe we could use Machine Learning or Deep Learning to help with the choice process.
Possibly but ultimately the issue is you have to disable all hardcoded AI on the unit so you have full control of the unit via script or external and build perfect unit control in its place.
We currently still piggy back on the hardcoded AI rather than disable it. When we tried disabling the hardcoded AI in the past we have to run AMAI scripts on a timer which can mean by the time it reaches the unit again it has been too much time passed to give the next order or restore control so the unit just stands around doing nothing/being killed.
Possibly but ultimately the issue is you have to disable all hardcoded AI on the unit so you have full control of the unit via script or external and build perfect unit control in its place.
We currently still piggy back on the hardcoded AI rather than disable it. When we tried disabling the hardcoded AI in the past we have to run AMAI scripts on a timer which can mean by the time it reaches the unit again it has been too much time passed to give the next order or restore control so the unit just stands around doing nothing/being killed.
I think we would have to build something that was based on feedback. But for that, what Warcraft sends us must include things positioned on the map, etc. Among many other statuses. But what I intended was to use a deep learning that would play against itself in various scenarios. Then she would learn as she played.