Can this item defined AI behavior know when an item is inappropriate for a target? If some information, such as a target's heat resistance gets sent to the item, such as a heat grenade, the heat grenade can be somewhat intelligent about whether or not it adds itself. The tile weighting also seems a bit odd in conjunction with the turn system. I don't think the AI understands that time dilation affects how traversible the tile is, but I might be mistaken. Really beautiful anyway. The AI has surprised me a few times. They'll do wild things like teleport to a target when their legs are taken away and they can't walk, or vice versa. I think they can chain together movement, such as using mechanical wings to cross a gap, though they won't leave the zone to go around an obstacle. Some more weapons such as vibratory ones should probably grant wall digging for when hearing and extrasensory enemies notice you through the walls.
Great talk. But I am confused about the actual implementation. What is the difference between the classes Goal and Originalntent? Are MoveEast and KillSomething both subclasses of Goal or does the code distinguish between Goals and SubGoals?
I know this comment is two months old, but I will reply anyway. Basically the way the system works is that every AI has a list of goals that need to be achieved in order. If a goal cannot be achieved, the AI will revert back to a previous goal that might work out. Original intent and goal are both the same class, original intent is just the goal to revert back to if things do not work out. There is no need to distinguish between goals and sub-goals in this implementation. Basically it is a simple way to execute tasks and revert to a previous task if the task fails.
Thanks for this talk! I really like Brian's attitude, such a humble developer.
Can this item defined AI behavior know when an item is inappropriate for a target? If some information, such as a target's heat resistance gets sent to the item, such as a heat grenade, the heat grenade can be somewhat intelligent about whether or not it adds itself.
The tile weighting also seems a bit odd in conjunction with the turn system. I don't think the AI understands that time dilation affects how traversible the tile is, but I might be mistaken.
Really beautiful anyway. The AI has surprised me a few times. They'll do wild things like teleport to a target when their legs are taken away and they can't walk, or vice versa. I think they can chain together movement, such as using mechanical wings to cross a gap, though they won't leave the zone to go around an obstacle. Some more weapons such as vibratory ones should probably grant wall digging for when hearing and extrasensory enemies notice you through the walls.
Really useful talk thank you!!
Great talk. But I am confused about the actual implementation. What is the difference between the classes Goal and Originalntent? Are MoveEast and KillSomething both subclasses of Goal or does the code distinguish between Goals and SubGoals?
I know this comment is two months old, but I will reply anyway. Basically the way the system works is that every AI has a list of goals that need to be achieved in order. If a goal cannot be achieved, the AI will revert back to a previous goal that might work out.
Original intent and goal are both the same class, original intent is just the goal to revert back to if things do not work out. There is no need to distinguish between goals and sub-goals in this implementation.
Basically it is a simple way to execute tasks and revert to a previous task if the task fails.
patrt 1:
ruclips.net/video/U03XXzcThGU/видео.html
Came to the comments to ask if anyone had a link to part 1. thanks!
I found the talk very interesting and constant microphone noise awful beyond words.