I'm currently learning how to use utility ai, you should check it out. Its basically ai that considers a task based on the information they have, which would be a float value. So like distance to ammo, health, etc. Supposedly can create great emergent gameplay which I'm excited about.
Every time is heard about something new I go from happy to disappointed to frustrated to sad to just ok it is what is it. Because I’ll never find a good source of education that teach proper usage of the tool. In the end it’s like just trying to learn Chinese with an Indian accent teacher. I don’t get it.
I was trying to learn unreal for like 5 months, but I got annoyed when I couldn’t get visual code to work correctly with unreal, mainly intelligence, so I’m using godot now lol
So you mentioned the performance issue with So you mentioned the performance issue with tick. However, you can also trigger states using events so you can send an event to your state tree and have it changed the state that is super powerful, especially with the gameplay messaging system.
Maybe it's just the way I've been using Behavior Trees, but every time I see people getting excited by State Tree and going 'oh, now with State Tree I can finally do X,' I'm like 'You could do that with Behavior Tree in Unreal. It's not that hard.' And when I looked through the documentation it seemed like State Tree was just Behavior Tree, but turned sideways. The only thing Behavior Tree can't really do from an AI design standpoint is GOAP, and it doesn't seem like State Tree can really do that either.
one positive you are missing...unlike behavior tree that requires a aicontroller state tree doesnt require a controller...meaning you can use it on a regular actor in the level to give it gameplay logic consistent with behavior trees capability
I had some experimentation with state trees in the past on its experimental state And the ammount of transitions needed for Complex AI behaviours is quite massive That made me stick to behaviour trees for AI However state trees can be used for other systems to create gameppay mechanics and systems and unlike Behaviour tree they arent limited to AI controller only
The true purpose of State trees is Mass AI, which stands for large-scale clustered artificial intelligence. You can refer to assets like the Matrix and the City Sample for examples.
It's pronounced "data" hehe Anyway, I'm working on a lesson plan for one of my students and AI is the focus of it. Since she's new to the engine and game design in general, I'm going to start by creating the entire system in blueprints, creating an FSM within there. Once we've made a feature complete AI with that, I was going to teach her how to make the same thing with behavior trees, but after this video, I will be considering using state trees. I am worried about the crashing though, so maybe not. It will be at least a month before we're there though. I'll also be uploading videos and lesson plans on it.
I'm currently learning how to use utility ai, you should check it out. Its basically ai that considers a task based on the information they have, which would be a float value. So like distance to ammo, health, etc. Supposedly can create great emergent gameplay which I'm excited about.
You just helped me make a bunch of decisions. Thank you!
What about AI percision?
May I request Detailed State Tree Tutorial for your watching community (tick functions and all)...please 🙌💯🙏
I'll see if I can put a good one together soon!
@@blackshinobi956 U da Man! 💯
Very informative video! Thank you :)
Glad you enjoyed it!
Every time is heard about something new I go from happy to disappointed to frustrated to sad to just ok it is what is it. Because I’ll never find a good source of education that teach proper usage of the tool. In the end it’s like just trying to learn Chinese with an Indian accent teacher. I don’t get it.
That's what documentations are for.
Yeah, I've felt the same way too. It's what made me start relying on documentation more than tutorials
New sub :) I LOVE your explanations and research into your content!
I was trying to learn unreal for like 5 months, but I got annoyed when I couldn’t get visual code to work correctly with unreal, mainly intelligence, so I’m using godot now lol
So you mentioned the performance issue with So you mentioned the performance issue with tick. However, you can also trigger states using events so you can send an event to your state tree and have it changed the state that is super powerful, especially with the gameplay messaging system.
You probably deserve a Nobel prize for that thumbnail picture.
Unreal 5.3 got the StateTree Debugger and it works great! Still has way to in term of readibility and bug fix but so far so good!
Maybe it's just the way I've been using Behavior Trees, but every time I see people getting excited by State Tree and going 'oh, now with State Tree I can finally do X,' I'm like 'You could do that with Behavior Tree in Unreal. It's not that hard.'
And when I looked through the documentation it seemed like State Tree was just Behavior Tree, but turned sideways.
The only thing Behavior Tree can't really do from an AI design standpoint is GOAP, and it doesn't seem like State Tree can really do that either.
Definitely agree, seems like a lot of things are a matter of preference
You are not kidding on how fragile these are. Previously I would have said Unreal 5 was rock solid... now? Save every 2 minutes, It's gonna crash!
Sheesh! It's tough! I did start working on another project using them in 5.2 and I've had less issues with crashes. I'm not sure what causes it
Great few videos, thanks for sharing.
one positive you are missing...unlike behavior tree that requires a aicontroller state tree doesnt require a controller...meaning you can use it on a regular actor in the level to give it gameplay logic consistent with behavior trees capability
Good point! Thanks for pointing this out!
C++ is still superiour then.
Very cool video. And spot on 😊
I had some experimentation with state trees in the past on its experimental state
And the ammount of transitions needed for Complex AI behaviours is quite massive
That made me stick to behaviour trees for AI
However state trees can be used for other systems to create gameppay mechanics and systems and unlike Behaviour tree they arent limited to AI controller only
I can imagine how big the transition list would be fore each state! Do you think that's unavoidable for certain systems?
The true purpose of State trees is Mass AI, which stands for large-scale clustered artificial intelligence. You can refer to assets like the Matrix and the City Sample for examples.
day-tuh for me haha good video!
Thanks for watching, glad you enjoyed it!
Great vid bro thanks!
Thanks for watching!
Yay nee video!!
Thanks for watching!
It's pronounced "data"
hehe
Anyway, I'm working on a lesson plan for one of my students and AI is the focus of it. Since she's new to the engine and game design in general, I'm going to start by creating the entire system in blueprints, creating an FSM within there. Once we've made a feature complete AI with that, I was going to teach her how to make the same thing with behavior trees, but after this video, I will be considering using state trees. I am worried about the crashing though, so maybe not. It will be at least a month before we're there though. I'll also be uploading videos and lesson plans on it.
That sounds like a great way to do it! I think there may have been some State Tree improvements in 5.3, so maybe crashing is less of an issue now
Do you have discord by chance?
I have a discord, but it has been neglected. I'll be working on revamping it soon!