Haha just yesterday finished setuping the goap ai. Does this continue or replace it? A good time to choose before I add all the content of the game to the systems.
@@yonatanabergel My personal opinion is to use Utility AI to supplement your GOAP decision making so you get the full power of both. Ideally you will end up with less true/false beliefs and more beliefs that are based on Utility which has the potential to make smarter choices.
My bachelor thesis was about creating an Asset that combines Utility AI and Behaviour Trees. I didn't find many Assets of videos about Utility AI 2 years ago, but it seems to have gained a lot of popularity since then.
I was just looking into Utility AI today and you dropped this video, absolutely insane timing. Thanks for all your hard work, it's seriously changed they way I've tackled programming!
It can be useful to add a small constant to the utility value of the agent's current action to avoid the agent flip-flopping between different actions when their utility values are very close to each other.
I'd love a comparison between utility AI and GOAP - specifically the upsides and downsides. I'd assume that GOAP is more expensive at least - but it honestly seems like it's easier to set up GOAP actions and prerequisites than manage all of the considerations you were writing once you get into composite considerations.
Literally could not be more perfect timing to release a video on this exact subject. I am experimenting with A Rimworld like game and was already considering using Utility AI after reading up on it a bit.
Thanks for uploading this, this was a great video! I am currently messing around with state machines and I'm using Playmaker for enemy AI, but I have been wanting something more like this. This video gave me some insight to how to handle decision making with math. Very cool, thanks a lot!
Very nice tool! And it's nice that this kind of knowledge representations are gaining popularity. I've been working on a similar project to represent different kinds of planning problems for my PhD in unity and is always nice to see both the AI and game dev community pushing these kinds of projects!
I love your videos, your explanation of design patterns makes game dev seem simple. Designing AI behaviors spunds like a colossal beast of a task to figure out how to design. But knowing there is a design pattern that can solve a particular problem makes me feel more confident that I can make complex games.
Your videos are always awesome. Is there a roadmap of your future videos? I'm struggling for modding and would love to know your knowledge on how to code thinking about modding.
You think you might be interested in talking about some of the Experimental features/packages Unity is working on? For example, Behavior Package and/or App UI? I think these current versions are for Unity 6 Preview and beyond.
Another great great video! I have a dumb question to ask. Since every AI has a brain to it and this script has things to calculate in update. In short we are having n number of brain updates methods working. That may affect the performance of the game right. Instead can I have a one global update method and using static delegate can subscribe that method to calculate. Can I take this approach?
You can do that, but it's going to be roughly the same amount of processing. My suggestion is to not calculate every frame, but instead put it on a timer so that calculations only happen once every 1-2 seconds, and then also do a calculation whenever the sensor detects something has changed.
Utility AI is such a powerful AI tool. The only issue is how much overhead ends up being produced when debugging anything. I would not recommend making a utility AI for real productions without a fully fledged editor window to help debug what's happening in real time. The inspector stuff going on in the video is nice, but try to imagine a complex AI with dozens of actions. It would be unreadable and hard to reason through.
I mean, the editor here is a pretty good start isn't it? wouldn't Editor.RequiresConstantRepaint solve the issue of having to mouse over the editor so you can see the changes in realtime?
Hi, first of all nice video and plz continue making more . Sorry for the noob question but whats the diference between using utilityAI for beahaviours instead of a State Machine , like we can switch states based on considerations or am i wrong . or we can improve utilityAI with a state machine , im a litl confuse.
While this is a great and well-made system, I wouldn't say it extends GOAP AI. It could replace the algorithm that calculates the best action path to a selected goal, but that is the core of GOAP. Personally, aside from using Scriptable Objects or upgrading a regular state machine system, I don't quite understand the advantage of this approach. It seems more complicated and a bit of a cudder for every action you want to add. Why not instead add more beliefs and actions that lead to the same goals? For example, with a bow and arrow vs. sword, Utility AI would pick depending on distance. In GOAP, the AI would see that using a sword means 1) getting to the player and 2) attacking, versus using a bow which is just 1) attacking from a distance. It's less simplistic but still more organized.