WIP Brawler: Page 4 - Godot Community Forums

This is a READ-ONLY snapshot of the Godot Community Forums when TwistedTwigleg was the owner, which is hosted for archival reasons only. The current Godot Community Forums can be found here.

Some links may point to external websites, including the current Godot Community Forums. These websites are not associated with TwistedTwigleg.

WIP Brawler

casual-garage-coder

Made an update : https://casual-garage-coder.itch.io/unexpected-browler/devlog/340524/week-end-update

A small step toward complete bug fixing. I'm still working on screen resolutions. I've activated some built-in options concerning expanding/streching/ratio conservation.

cybereality

Okay, it plays a lot better now. This physics are working and it's really fun. Thanks for listening to my feedback, it feels like a real game now. One thing, though, that I'm sure you don't want to hear, but I think you need to code AI. People may not have 4 friends locally they can play with, or maybe there are 2 people that want to play but it would be more fun with 2 AI characters as well. I think it really needs it.

casual-garage-coder

@cybereality said: One thing, though, that I'm sure you don't want to hear, but I think you need to code AI. People may not have 4 friends locally they can play with, or maybe there are 2 people that want to play but it would be more fun with 2 AI characters as well. I think it really needs it.

And you're absolutely right ! I've though about it for months. I've been into AI in my early days (pathfinding, gauge driven emergent behavior, and so on). I must admit it could be a lot of fun to implement. I've got some ideas, some tricks you might say, but, well, AI in games is all about trickery :) So, right, let's do this ! I've some content ahead. A major "AI update" seems to be the right thing to do.

cybereality

Well AI is actually very simple. I don't mean real AI, like to play chess. But for platform game enemies you could probably do it in a week. The basic idea is to make a state machine (for example, idle, attacking, retreating, etc.). Then just create situations that change states. Like if another player is within 200 pixels, then move toward them. If they are being chased and there is a platform above, then jump. If their life is low, then instead of running toward a player, they run away. Just a few if-else statements and a simple finite state machine. I like I said, not more than a week of work.

casual-garage-coder

I know. But I'm doing this project one my spare time. It'll certainly take more time (or not if I'm not reasonable with sleeping).

casual-garage-coder

Small dev-log-ish : - Still working on AI. I'm trying to re-create what i've done some decades ago. Using tilemaps and collision shapes helps a lot. I have to build movement segments based on normals. It identifies were an entity can walk/wall grab/wall jump/hang on. Hopefully, in my game, there's only walk for the moment. Other movement might appears as passive special abilities of some characters - Once you've some the movement segments, you place navigation node accordingly. Each node has a purpose (walk-to, jump-to, etc.). Finally, there's some dark magic to take into account dynamic environement (moving platform, closing doors, for example). - Then, the fun begins : fuzzy-pathfinding using the graph. BUT, it's only for common purpose to travel from A to B. It doesn't take into account dog-fighting. - Dog-fighting is the fast-pacing movement chain when you encounter your opponent. It is restricted to empty sub-areas (in godot, Area2D), the AI movement computation is more realtime (in opposition to planned when using the navigation graph). But at some point, either the fight concludes to a death, else, it transitions to fleeing/following using the classic pathfinding. More fun ? 3-way/4-way dogfighting. - Dynamic environment (moving platforms). It can be painful in numerous way : First, you have to use a flexible navigation supervisor allowing graph local reconstruction in case of obstruction. The movement segment may or may not change (collapsing platform). But, it can be managed as another kind of movement. But the key for dynamic environment is "waiting". Your entity computes the fact that it has to wait for a navigation node to be accessible. - Other 'fun stuff' : additional gameplay (aka. arena puzzle). I plan to had puzzle and traps in th arenas. It's additionnal gameplays that are to be taken into account. Else, the "human" players will have a big advantage over the bots.

Can't wait to see an 'all-bot' brawling session :scream: But it will definitly take me more than one week :)

cybereality

Oh yeah, well you're doing it the correct way which will take more than a week.

casual-garage-coder

So, I've been busy with a new project at my "real" job. For now, i'm 20+ hours into the AI. And while searching the net for an algorithm to compute jump trajectory (it's easy in fact, but i was lazy), I found THIS:

https://github.com/SnoringCatGames/surfacer

They are working on the exact same thing than me ! But with more elaborate moves including wall jump, wall grip/climb, etc.

I considered using this at one moment, but ... well ... My case is trivial and the integration of the Surfacer/Scafolder doesn't seem to be really easy (and I've not talk about the Google reporting thing ...).

If any of you is curious, you might take a look.

I'm 20 hours and 800LoC into it. They are way above that :)

cybereality

Cool. I do like to write everything myself, I don't use plug-ins or add-ons. But I will search for algorithms and rewrite them in my code style.

casual-garage-coder

Exactly !

But I always perform a benefit/loss analysis for using COTS. An habit from my actual job where sometimes, time is running out and you should re-use stuff. The condition is that the degree of reusability must be high enough to justify the integration.

In this specific case, I won't use the plug-in cause it's like using a flamethrower against mosquitos.

But I thought it was a good occasion to shoutout about the massive amount of work that has been done for the godot community.

cybereality

Well it depends, for really large and complex systems I will use libraries, provided they are specific for a feature. Like when I was writing my own 3D engine, I wrote the model importers myself for COLLADA (this was before glTF came out) and COLLADA uses XML, so I wrote the XML parser by hand too. This was a huge mistake. I did get it working, but the code was horribly slow and just bad. Then later I downloaded an XML library and it took like 5 minutes and it was up and running and 100x faster than my ghetto-parser. Still learned something, but it was a waste of time and a worse solution.

« Previous Page