Видео

"Awesome video, buddy! I was wondering if you could help me out with creating a combat system similar to World of Warcraft. I’ve been searching for guides but haven’t had much luck finding anything specific. If you could point me in the right direction, or let me know if you’ve already made any videos covering combat mechanics, I’d really appreciate it!"

I have written a GOAP system before in UE. I would get really frustrated with how bad performance could get, and debugging them was double frustrating. You mentioned doing it all in blueprint being a pain point, but doing some of it in C++ introduces the pain of restarting the editor after changes. Of course any workflow has pros and cons, but in the end, the performance cost was too frustrating to deal with to warrant it.

What cannot be stressed enough, that Blueprints or not, you still need to learn how to code.

wholesome, love it!

I like using C++ for things like structs or other "data container" kind of classes/actors, but it's a lot faster to actually implement most of those things in BP. Or at least feels less complicated. I think C++ is also just easier to kinda throw code between projects, instead of having to open the editor to migrate it to another project as is the case with blueprints. Honestly I think they both work pretty well and wouldn't say anyone should pick one or the other. Like you mentioned, Epic has even outlined some great workflows where you use both. As a programmer on a team I'd absolutely want to give game designers tools to be able to implement some things themselves, instead of trying to do everything in C++. Lets me focus on the parts I need to, and I can always jump in and make C++ versions of things to help if there are inefficiencies.

I guess within 5 years AI can efficiently write C++ code but not BPs. I used AI to write short Python code that took 2 minutes. What do you think?

Just a heads up man, there's tonnes of free and cheap C++ utility nodes on the fab marketplace you can just install which cover pretty much every complex math or function you might need. Huge packs of nodes for exactly these things.

you know you can code custom nodes too, you don't need to have a whole pawn for example be coded in C++, I personally like having a "Utilities" C++ class in my projects where i can store global blueprint callable nodes to run complex math or functions. if the blueprint doesnt have a function i need, i pop into my utilities class and whip it up as a node. then continue on with my BP. 90% of fortnite runs on BP, i imagine the other 10% is C++ code tucked in utilities like i do it.

Want to pursue a career in game dev? 🎓Go to my sponsor snhu.edu/Shinobi and learn how you can get started with a Game Development degree!

You could create different types of zombies some that are drawn to different areas or objects. Kinda like in Romero’s Dawn of the Dead.

Awesome explanation! It's really interesting to hear about when to use GOAP and I'm now wondering if it's something useful to me.

Fantastic to hear your story! I had a similar issue with my early games and losing them so I know that pain. Luckily it looks like you're in a much better place now.

I absolutely agree with you! The best way to learn is to do something the "wrong" way and experience the limitations first hand instead of just using the "better" solution right away based on conclusions made by other people. That way you will really understand the differences between these different options and build a strong foundation. I see too many students getting stuck on behavior trees trying to implement a simple follow and attack, if they could have just done it raw in blueprints within minutes. No need to overcomplicate things as long as they work out just fine and we should only move on to more complex solutions as the need arises.

More derail into the implementation would be amazing.

The main downside is performance, using a pathfinder when changing environments mean you have to throw the plan out often does not scale well at all, it's only suitable when you only have one or a handful of agents. Other frameworks like Dave Marks Infinite Axis Utility Systems is a better approach if you want to scale to more than 25 agents and still keep the agent autonomy but too runs into issues over several hundred agents. Finally regardless of approach collaboration isn't as complicated as you'd think, essentially you have a party system where agents can choose to join an empty slot in a group. Once a group is full then you add a bunch of new goals to all of the agents and due to their higher priority the agents will typically do the same things together, that and reacting to each others broadcasts gets you seemingly cohesive collaboration. (Applies to BT, GOAP, IAUS, or any planner really)

Awesome tutorial man, really enjoyed it, I was just looking for something like this, please, make a video with the implementation you made, even if it is simple, it will be extremely helpful..

i cant seem to find any GOAP tutorials for unreal. youtube is only showing videos for unity and gadot

They ain't ai, they're bots. They do what when and how you tell it to do. They won't correct your input if it's wrong. They ain't thinking, they're bots. Chatgpt is virtual intelligence, it's not thinking, it just repeats things what it's been given in a sequence which reasonably makes sense(the sentence as a whole may not, but individual words do. Of what I assume chatgpt o1 does, that could be considered AI as it actually uses reason, it has some thinking capacity. I don't know if it can write something new without reusing stuff from Internet tho

I think the best use I can see of this, is to quickly replace the blockout instances with more finalized assets. _(ie; replacing all cylinder 'lights' with fixtures and bulbs.)_

Hell ueaj

Does the AI only look one step at a time in the network grid? Or does it look to the total cost in the chain of steps to the final goal? There are occasions where it would be more efficient and cost less effort to make the more cost expensive decision on the first step. What emerges as the best chain of actions may change during the progress through the steps. How much sense would it make to let the AI chart the steps in the grid all the way to the end goal before making a decision, and impose a cost on changing the charted plan?

Wow an actual use case for prolog i think

Please do made another video about your experience developing GOAP. I appreciate your breakdown. I just subscribed and will wait patiently 😊

Thank you!, useful and straight to point

I think it should be easy to implement in blueprints since it's it the implementation of it comes down to what the developer is looking for from the AI.

The more i learn about AI stuff the more over my head it goes 😂 sounds really fascinating though. Would definitely be interested in that longer form video if you make it 😁

Thanks for the overview, this was really great! I was going to ask about reaction synthesis because Google only gave me results about chemistry, but I see you included a link in the description. Bless you. 🙏 Would love to see your implementation, even if it is BP spaghetti. 😂 It's been a while, but trying to work through it in the past I always got tripped up by the whole MoveTo idea. Like, it's an "edge" of the graph that's basically always available, but it has no fixed cost and no fixed effect because it's parameterized, so how do you incorporate that into an A* solver, especially one that runs backwards? Say there's an available weapon that's close by but less effective, and another one that's farther away but more effective, how do you decide whether it's worth it to move to the better weapon? Actually, FML, I just realized an action like PickUpWeapon would also be parameterized. Or what if your game has vehicles... Getting into a vehicle is an action that would reduce the cost of future MoveTo actions, and is _also_ parameterized (assuming you have more than one vehicle type available). My brain hurts.

Can GOAP broadcast events to others? Like, a zombie could broadcast "pack leader" or have that tag and nearby zombies might then choose to follow, giving groups more purpose. I didn't even know this existed, I like how UE has so many varied tools. As far as possible zombie goals, I'd say think about them like dumb animals. They're not stagnant, so they would forage any interactable objects to see if food is in there, like drawers, cabinets, breaking plants because you never know if meat is hiding under a plant. haha Hunting might be faster patrol of a zone ignoring most forage options, focusing on travel hot spots like trails and doors. Might they be smart enough to nest, which means they could forage for specific supplies which might lead them into player gathering areas. If a zombie just fed, would others know and want to attack it if they drop below a certain hunger level. Zombies mindlessly doing jobs they mindlessly did while alive? Anyway, I'm done pitching stuff that might not be possible. :p

i should've watch the full video before my last comment XD, but is that what you think the case or that the original goap design that the one action's pre conditions are another action's effects

i have made a goap implementation too and it works pretty well , but my confusion is with pre condition, currently my setup if the action has let's say two pre conditions that are not met by single action everything fall a part , like how do you account for that case ? because if the pre conditions are not met by one action means every pre condition should be considered a goal and we run the planner for it to get a plan for it so we get a sequence of plans and that makes things pretty complicated to me so what i want to ask is do you think i over do it ? like from design perspective i should when i am making the actions objects i should make sure this never the case like if there an action needs condition1 and 2 then there should be an action that has the effect1 and 2 so planning works as dominos and hooking a sequence of actions become much much simpler plus i would really appreciate your thought on my setup if you have time : ruclips.net/video/4B6-KEeXjnE/видео.html thanks much love

Thank you! I've been thinking about the larger game I'd like to make and how I could set NPC jobs. This might be a good way of doing that. I'm glad you're posting again, I enjoyed your early state tree vids

nice clip but the zoom effect on screen makes it hard to follow

🎓Go to my sponsor snhu.edu/Shinobi and learn how you can get started with a Game Development degree!

its not a wast of time it a wast of money it keep you paying out to get new system and update's to run the UE 4.01 to UE5.5 software smooth over the years engine update alot you needed better pc or laptop. i just got use to it but getting a job in the UK is a nightmare even back in 2008 it really bad for money and try to keep what you saved up over this years is hard with the UK laws tell you cant fight back and you have let them robber's get away with are stuff or we get lock up if we take matters in are own hands. hell they mite be looking at my comment right now haha

Niceness! I'm making a Turn-based RPG, so I'm using blueprints. It's really about how much the AI needs. >3 actions or simple actions, use blueprints, 3-7 is easy in states, everything else should probably be a tree. Like my base enemy class doesn't use a tree, but my boss child class will probably need a tree, because it needs to act with more strategy, and sometimes control the other enemy's. Might just add the tree to my base tho, who knows. Not hard to dump half a dozen events into a tree.

Dude ive been binging your videos. Appreciate the authenticity and insights i dont really hear elsewhere. That video explanation you gave on world partitions was the best ive come across and i watched at least 12-13 today trying to wrap my head around the principles

I dont think enough people realize the weight of what you're saying when you say "do it however you can" and "there's going to be times where youre just going to have to live with imbalances". It's so true dude. Its true for doing anything big with your life

Hands down the best explanation and breakdown of the World Partition system in UE5 as of 12/16/24. Thank you <3

I had to start my project over 2 years ago because of the same reason, and dealing with it now, I been telling myself to switch to SSD but I'm a procrastinator

version control 😳

Heh your life sounds like mine. Family with 2 kids, 2 jobs and end day time for dev. First game is up on Steam dough! It sucks! BUT ITS MINE :D

I just got started with UE5 a while ago with no coding or game dev background at all and I have to say it's more than just amazing. And while I do have a day job, I always look forward where I have full days off, weekends and holidays alike. Sure - and that's my opinion as a beginner - quitting the day job and giving up your whole life so to say isn't a good idea or even a good suggestion. However, when new to game dev, such as in my case, I appreciate it to have as much time at hand as possible. For a simple example: I want to do some level design and add a new mesh to the scene that makes me think "Hmm, wouldn't it look awesome when it would look like <that>?" and an hour later I am still fiddling around with materials or textures. In any way, UE5 feels like a drug to me. On days off I get up, do the usual things I do in the morning (like feeding the cats or brushing teeth) and as soon all that is done -> start my project, continue working on it. And what goes for tutorials in my opinion: Yes, if you follow them without the plan to understand them sooner or later, those will be just tutorials. However, and I am speaking from my own experience, over time and with a bit of a dedication and interest in all that stuff, you'll find out what the heck you were doing and what was once "a tutorial to follow" becomes a "wow, NOW I get what that means/that is for/what I was doing/etc". Having understood BPi's and event dispatchers so far, the thing I currently struggle the most with understanding is line/sphere/whatever trace to implement things such as foot step sounds on/in different surfaces. Besides all that: Nice video, appreciate it. Very grounded, motivating and inspiring. Thumb up and sub left.

good video Lesson 11 take breaks no one works everyday

oh you are not an hokage. you are still a man, enough here.

Lesson #6: Acquire Skills - definitely a key to success.

Love how real you are bro. Don't over sensationalize your stuff or make it seem dramatic or crazy. it just is what it is. I really appreciate that type of content creator.

Great lessons! I also absolutely agree with not quitting your job just to make games. The grass is always greener on the other side and you might imagine that you're gonna get more done with the pesky day job out of the way. But not having the safety net of a steady income cause a lot of anxiety and having too much pressure to make money can lead you down the wrong path. Not having a job anymore also isn't gonna magically make you more disciplined or motivated either and you'll have to be really good at managing yourself.

Not trying to brag but I'm pretty sure I am the world record holder for the game at 6:04

You make a great point on the take a course route. There are a lot of great youtubers that put out great content but nobody can learn a system in unreal in 20 minutes. You'll need 1hour minimum to scratch and get a general idea of what a system is. Courses allow for that polish and time to be taken to properly learn things. Coba Code makes a ton of great 2d/2.5d courses that are on Udemy.

Honestly, I think doing game development while also having a job is actually the preferable option. Because, yes if you're full time game dev, you'll get more done. But unless you have plenty saved up on the side to sustain your daily life, eventually you're gonna run out of savings to continue that full time dev life. At least that's been my experience so far.