Jump to content

Procedurally Generated Maps... Work


236 replies to this topic

#121 maxdest

    Member

  • PipPipPipPipPip
  • Veteran Founder
  • Veteran Founder
  • 137 posts

Posted 06 August 2015 - 11:55 AM

Whilst proceeduraly generated maps would be a great thing (if done correctly) this is a huge piece of work probably suited to an early design goal in a new game rather than patching an existing one.

Unless cryengine has some built in functions that a quick google doesnt show, then you are looking at a low level rewrite of the engine and networking. The only way around this I could think of is letting the clients build the map themselves using the seed, but am pretty sure no one would want to sit for 30mins waiting for the map to bake on the guy playing on a Pentium 4.

A better way to get lots of maps would be to open up creation to the public, then have a trusted player council whom weed out the duffers and ones with illegal assets. The best few could be put on rotation on the PTS with MC rewards and titles for contributers. Bet you could get 20 great new maps by the end of the year, maybe even enough maps to start to differentiate different planets and play modes in CW.

Edited by maxdest, 06 August 2015 - 11:57 AM.


#122 Rhaythe

    Member

  • PipPipPipPipPipPipPipPipPip
  • Philanthropist
  • Philanthropist
  • 4,203 posts

Posted 06 August 2015 - 12:19 PM

Be amusing if PGI went the SETI route and had a distributed client process that generated and vetted seeds for dynamic map generation pools.

And by amusing, I mean impressive as hell.

#123 Raggedyman

    Member

  • PipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 1,278 posts
  • LocationFreedonia Institute of Mech Husbandry

Posted 06 August 2015 - 12:35 PM

View PostMister Blastman, on 06 August 2015 - 06:39 AM, said:

More upsetting than when you take a SRM+Small Laser 'mech and draw Alpine in the random map selector?

If anything random maps will make less meta mechs potentially more useful since nobody will know where the sniper nest is or the most advantageous positions are to utilize their longer range weapons. Brawlers will more easily be able to sneak up behind players because nobody will know the map.


The thing is that people know how many Alpine maps there are at the moment (one.... possibly one to many...) so they are able to make a judgement call in their head as to how likely it is to turn up in their rotation and then respond accordingly. They have a relatively constant value of Alpine (or any other map) to gamble their loadout with.

Yes, this does mean that people will tend towards more "meta" type builds, because they want to win the round. See my comments about Killers and Achievers, a lot of people like meta/min-max/most-viable loadouts, its part of their enjoyment of the game.

View PostMister Blastman, on 06 August 2015 - 06:39 AM, said:

I know folks like comfort, but soldiers on a modern battlefield or true Battletech warriors had to/have to accept that every time they go to battle, it will probably be in unfamiliar surroundings.


The thing is that Soldiers and Mechwarriors alike, whilst they won't know the exact layout, will have enough of an idea of what they are up against to adjust accordingly. If you want to bring in "real world" realism to the game then you would need to include a 'see the map and then pick your mech/loadout' approach.
Which might not actually be a bad thing in many ways, but it would mean a new type of "meta" as people learned how to min-max their post reveal choices accordingly.

View PostMister Blastman, on 06 August 2015 - 06:39 AM, said:

Relying on familiar maps is a crutch. There's a reason there are map compilations for older shooters like Duke Nukem 3D or Doom that number in the hundreds or even thousands. Players did and still do enjoy testing their abilities in unfamiliar environments.


Okay, I'm going to have to disagree on the use of the word "crutch" on two fronts:
1 - Making a solid, playable, visually appealing map that is enjoyed by a large portion of the player base (ie what the developer has to do) is actually very hard. Making a procedural system that keeps that same player base happy is even harder.
2 - Predictability in playing field is something that has been around in gaming/game playing for a very long time. People/players really like it, beyond any technical limitations. This is one of (though obviously not the only) reasons Settlers of Catan is outsold by Monopoly and that major gaming tournaments tend to be fought on known maps.

With regards to mappacks - yes, there are games with thousands of maps out there. However most games are doing amazingly if they have more than 20 which are considered "classic" and that a majority of people want to play. Even with map-packs most people will end up cycling through a small selection of the overall amount available, often because it takes so long to exhaust all the play-ability from a good map.

And I do agree that some players "did and still do enjoy testing their abilities in unfamiliar environments". However it is not a majority, and that's not just because they haven't had the experience of playing in them.

#124 Mister Blastman

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Survivor
  • Survivor
  • 8,444 posts
  • LocationIn my Mech (Atlanta, GA)

Posted 06 August 2015 - 01:25 PM

View PostSatan n stuff, on 06 August 2015 - 11:18 AM, said:

He needed 100 lines to code that mess? Those are just randomly sized blocks with randomized positions, there's nothing even remotely intelligent about it. If you're calling that a city generator you really don't have a clue how much work would be needed to make that look believable.


You're really bad at reading comprehension.

#125 Satan n stuff

    Member

  • PipPipPipPipPipPipPipPipPip
  • Urban Commando
  • Urban Commando
  • 3,508 posts
  • LocationLooking right at you, lining up my shot.

Posted 06 August 2015 - 03:22 PM

View PostMister Blastman, on 06 August 2015 - 01:25 PM, said:


You're really bad at reading comprehension.

Maybe you're just bad at software comprehension.

#126 Johnny Z

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Knight Errant
  • 9,942 posts
  • LocationDueling on Solaris

Posted 06 August 2015 - 03:28 PM

View PostCygnusX7, on 06 August 2015 - 10:22 AM, said:

Hey lets come up with every reason not to.. rather than every reason it should be done.


Seen the new River City map? Its awsome. Like I said before it may be among the best maps of any game out all things considered. If procedural maps could do that, then by all means bring it on. Procedural map gen cant even come close remotely to that. Not even close. I can asure anyone that map was hand done to the last millimeter. Literally.

It was probly limited in some ways by what they wanted to use for memory on it and thats about it.

They keep all this up they will be putting other games in the dark ages. Really. Except a very few.

Edited by Johnny Z, 06 August 2015 - 03:35 PM.


#127 Ordin Hall

    Member

  • PipPipPipPipPip
  • Survivor
  • Survivor
  • 123 posts

Posted 06 August 2015 - 05:38 PM

View PostZeusus, on 04 August 2015 - 11:26 AM, said:

Second won't work. I'm not waiting 20 minutes to DL/generate a map each match...


What exactly makes you think it will take 20 minutes to download some simple map geometry? Are you still using dial-up?

#128 Iqfish

    Member

  • PipPipPipPipPipPipPipPipPip
  • 3,488 posts
  • Google+: Link
  • LocationGermany, CGN

Posted 06 August 2015 - 05:52 PM

Changing a game to procedurally generated maps isn't just flicking a switch. To do this, you would need to rework a game like MW:O from it's core structure. We all know this won't happen, so stop dreaming about it.

Also, please recognize the facts here.
You show a cool procedurally maps generator and there are a few more out there but they are LOCAL.
To be abled to do this in an online game, you would need to either have the generator local on every connected machine, to generate the map locally or you'd have to stream a new map everytime a match starts.
Both options result in horrendous waiting times. That's the second elimination for the idea.

Next we have Hitboxes. Hitboxes do not take models into account since the physics simulation cannot be that detailed, that's why there are inaccuracies. Having auto-generated hitboxes for procedurally generated maps wouldn't work or would need a lot of attention and maintenance. Third elimination.

Next up is the topic of other assets. Generating a landscape is fairly easy. Everyone who ever toyed around with a landscape generation tool like the CE3 sandbox knows what I am talking about. It's a matter of rising and lowering terrain and then applying textures like sand, rocks or water to it.
Buildings however is a different story. Do you want to play on empty maps all the time? No, so that's why we need buildings. These assets need to be built and you need a lot of them if you want to place them randomly. If you want algorythms to place them in particular orders, you need to have those written. Oh and by the way, the XYZ orientation, rotation and scale of these assets needs to be controlled too. 4th elimination.

Next up we have playability. Do you know how long a map is being reviewed until its puiblished? raising the terrain, placing a few buildings and trees in there and watching the polycount is the easiest step, you can do that in a week by yourself.
The CONCEPT behind these maps, the corridors and keypoints for gameplay are hard. Observe any map in any game and you'll notice a concept. That's something which would be impossible for procedurally generated maps, which renders them unplayable.

There is a lot more which makes these generators impossible for current online games and these are good reasons why nobody uses them. Even big companies which are far bigger than anything PGI has ever dreamed to be like cloud Imperium games are struggling with procedurally generated content and only utilize it for smaller components of their game Star Citizen.

So please, stop giving people the idea that it's only a matter of goodwill and some cool generator. We are not there yet.

#129 Mister Blastman

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Survivor
  • Survivor
  • 8,444 posts
  • LocationIn my Mech (Atlanta, GA)

Posted 06 August 2015 - 07:33 PM

View PostJohnny Z, on 06 August 2015 - 03:28 PM, said:

Seen the new River City map? Its awsome. Like I said before it may be among the best maps of any game out all things considered. If procedural maps could do that, then by all means bring it on. Procedural map gen cant even come close remotely to that. Not even close. I can asure anyone that map was hand done to the last millimeter. Literally.

It was probly limited in some ways by what they wanted to use for memory on it and thats about it.

They keep all this up they will be putting other games in the dark ages. Really. Except a very few.


It is awesome but people still go the same stupid spot on it. Citadel. Over and over again. It is dumb. We need procedural maps to force players to stop doing the same crap over and over.

I do not care if they do not like it. I do not care if it makes them uncomfortable and I certainly don't care if more than half the MWO population leaves because they can't handle using their brains rather than repeating the same ritual every single drop. The game will be better for it in the end.

#130 Raggedyman

    Member

  • PipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 1,278 posts
  • LocationFreedonia Institute of Mech Husbandry

Posted 07 August 2015 - 03:35 AM

View PostMister Blastman, on 06 August 2015 - 07:33 PM, said:

I do not care if they do not like it. I do not care if it makes them uncomfortable and I certainly don't care if more than half the MWO population leaves because they can't handle using their brains rather than repeating the same ritual every single drop. The game will be better for it in the end.


I care if more than half the population were to leave after a highly-expensive-to-develop-technology gets put into the game, because the game would end.

#131 Joseph Mallan

    ForumWarrior

  • PipPipPipPipPipPipPipPipPipPipPipPipPipPipPip
  • FP Veteran - Beta 1
  • FP Veteran - Beta 1
  • 35,216 posts
  • Google+: Link
  • Facebook: Link
  • LocationMallanhold, Furillo

Posted 07 August 2015 - 06:05 AM

If it is cost effective for this game do it. If not... wait.

#132 Rhaythe

    Member

  • PipPipPipPipPipPipPipPipPip
  • Philanthropist
  • Philanthropist
  • 4,203 posts

Posted 07 August 2015 - 06:09 AM

View PostJoseph Mallan, on 07 August 2015 - 06:05 AM, said:

If it is cost effective for this game do it. If not... wait.

It's not, ultimately. Nor will it be any time soon. The up-front cost is too extreme, and it offers no direct path of profit. So won't happen.

#133 Mister Blastman

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Survivor
  • Survivor
  • 8,444 posts
  • LocationIn my Mech (Atlanta, GA)

Posted 07 August 2015 - 06:31 AM

View PostRhaythe, on 07 August 2015 - 06:09 AM, said:

It's not, ultimately. Nor will it be any time soon. The up-front cost is too extreme, and it offers no direct path of profit. So won't happen.


Well, see, that's the thing--and we know this is true, PGI has been extremely reluctant to produce new maps because they are cost centers. The procedural map generator doesn't even need to be done in the cry-engine, it can be an external application that produces the map which is then placed in the map folder and the game code picks it up from there. If we think of all the man hours required to create two or three maps and in those same amount of hours create a procedural generator that gives limitless maps...

To PGI that might actually look appetizing. In the long run it will lower costs tremendously and give players what they want--variety in maps. It would revitalize the game.

PGI has said it takes 200k to make a single map. You hire a coder who's skilled in procedural stuff, stick him in an office by himself and that is his only job. A single coder costs way less than 200k. From a business perspective something like this makes sense.

#134 Joseph Mallan

    ForumWarrior

  • PipPipPipPipPipPipPipPipPipPipPipPipPipPipPip
  • FP Veteran - Beta 1
  • FP Veteran - Beta 1
  • 35,216 posts
  • Google+: Link
  • Facebook: Link
  • LocationMallanhold, Furillo

Posted 07 August 2015 - 06:35 AM

View PostRhaythe, on 07 August 2015 - 06:09 AM, said:

It's not, ultimately. Nor will it be any time soon. The up-front cost is too extreme, and it offers no direct path of profit. So won't happen.

So we wait.

#135 Mister Blastman

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Survivor
  • Survivor
  • 8,444 posts
  • LocationIn my Mech (Atlanta, GA)

Posted 07 August 2015 - 06:35 AM

View PostIqfish, on 06 August 2015 - 05:52 PM, said:


Next up is the topic of other assets. Generating a landscape is fairly easy. Everyone who ever toyed around with a landscape generation tool like the CE3 sandbox knows what I am talking about. It's a matter of rising and lowering terrain and then applying textures like sand, rocks or water to it.
Buildings however is a different story. Do you want to play on empty maps all the time? No, so that's why we need buildings. These assets need to be built and you need a lot of them if you want to place them randomly. If you want algorythms to place them in particular orders, you need to have those written. Oh and by the way, the XYZ orientation, rotation and scale of these assets needs to be controlled too. 4th elimination.



We already have building assets. We just plug those in through the algorithm. The advantage is we already know their dimensions which can be categorized in a small database.

The terrain hitboxes are dependent on the resolution refresh done for those hitboxes. I think this is trivial at best. It is not a big deal.

As for playability? They are infinitely more playable! There's no more ring-around-the caldera nascar or cling to the side of the citadel because nobody knows what to do or where to go. Chaos ensues and talent wins. It beats the hell out of the random teammate mess we have right now. Hell, it might even give the good players who are teamscrewed a chance.

#136 Rhaythe

    Member

  • PipPipPipPipPipPipPipPipPip
  • Philanthropist
  • Philanthropist
  • 4,203 posts

Posted 07 August 2015 - 06:46 AM

View PostMister Blastman, on 07 August 2015 - 06:31 AM, said:

The procedural map generator doesn't even need to be done in the cry-engine, it can be an external application that produces the map which is then placed in the map folder and the game code picks it up from there.

No.

Speaking as a coder, you *don't* buy off-the-shelf when you have modified your game engine to do so much on its own. I've tried to explain this already. It's cool. Procedural stuff is awesome. But that is a design decision you make *early* in a project. It is *very* expensive to implement later in the life-cycle.

Even worse than generating the map is vetting it, and that might be even more expensive to automate. You have to account for pathing. Traversal. Render faces (how much of the map does a client see at any angle at any given time). Competitiveness. Visibility.

You don't just "code a procedural engine" and shove it into your system. Your system has to be built around it.

We will not see procedural maps in this game. To imagine otherwise is fun, but it's silly to think it'll be acted upon. The up-front cost is *severe*, and PGI will see no profits from it. That is, ultimately, the deciding factor. They can't sell it, so why would they fark up their hole game to try and shove it in?

#137 Johnny Z

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Knight Errant
  • 9,942 posts
  • LocationDueling on Solaris

Posted 07 August 2015 - 06:49 AM

View PostRaggedyman, on 06 August 2015 - 03:53 AM, said:


Question: do you mean randomly generated maps for that planet or each planet getting a random sequences from a set selection of prebuilt maps?

Either is cool, counts at Procedural Generated Content, and adds to the fun IMO, I just wanted to be clear on what you meant before I get all excited.


Well looking at the galaxy map as a whole, then them linking map sets to each planet is in a way procedural. As they finish new map sets they would be added of course also. Thats sort of my point.

Procedural maps cannot be compared to hand built maps. Like I said before. Since these products wont degrade over time in some ways and everything else taken into consideration like each piece/map will be so widely used so many times, going with hand built is without any doubt the best way. Which Mechwarrior Online is doing if anyone doesnt know.

By the way, this is the Art Deco debate all over again. :) Art deco isnt actually a style, its anything that includes form with function, not just function. It was general more expensive and more labour intensive, also was reliant on higher skilled labour and basically fell by the way side. But those items and buildings are worth far more than anything else from their time.

Art Deco is a seriously deep subject actually. :)

Edited by Johnny Z, 07 August 2015 - 07:05 AM.


#138 Mister Blastman

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Survivor
  • Survivor
  • 8,444 posts
  • LocationIn my Mech (Atlanta, GA)

Posted 07 August 2015 - 07:28 AM

View PostRhaythe, on 07 August 2015 - 06:46 AM, said:

No.

Speaking as a coder, you *don't* buy off-the-shelf when you have modified your game engine to do so much on its own. I've tried to explain this already. It's cool. Procedural stuff is awesome. But that is a design decision you make *early* in a project. It is *very* expensive to implement later in the life-cycle.

Even worse than generating the map is vetting it, and that might be even more expensive to automate. You have to account for pathing. Traversal. Render faces (how much of the map does a client see at any angle at any given time). Competitiveness. Visibility.

You don't just "code a procedural engine" and shove it into your system. Your system has to be built around it.

We will not see procedural maps in this game. To imagine otherwise is fun, but it's silly to think it'll be acted upon. The up-front cost is *severe*, and PGI will see no profits from it. That is, ultimately, the deciding factor. They can't sell it, so why would they fark up their hole game to try and shove it in?


I think you misunderstand--and I'm an ex-coder who got out of that racket eleven years ago--the procedural map generator is an external application--it makes the map, plops the file into the filesystem. The generator is written from scratch. It isn't off the shelf.

I have great faith in what can be done through good coding. There are a lot of **** coders out there--more than doctors! (and there are a lot of bad doctors), but the good ones can produce neat things such as that.

You aren't farking up the game by making an external map generator. The game doesn't know the difference. The game loads the map file and you play on said map. To the game, it is just a map, nothing more.

#139 Triskelion

    Member

  • PipPipPipPipPipPip
  • The Devoted
  • The Devoted
  • 226 posts
  • Google+: Link
  • Twitter: Link
  • Twitch: Link
  • LocationFort Collins, CO

Posted 07 August 2015 - 07:32 AM

View PostMister Blastman, on 04 August 2015 - 12:29 PM, said:

I beg to differ. I'm sure there are many competitive players that would prefer to play on random maps. Thinking on your feet is the ultimate test of skill and prowess.


My sides.

Procedural generation is all well and good, but trying to force it into an environment that wants to be competitive on a team level is nonsense. Do you want to do spawn swapping EVERY match? Do you want to look over every map to make sure a weight class isn't getting a serious advantage? This is already a game where some mechs are naturally punished by map choice. We can't directly pick maps either - this gets worsened if maps are universally inconsistent and far less tested.

Not to mention, you're clearly spouting off on something that you have absolutely no knowledge of. What happens to the map costs for the current map set? Do those get written off? Do you think PGI makes enough money to even get to make that choice? I don't have a programming background, but reading through responses by actual programmers and my current, exceedingly basic knowledge of the downsides involved makes me think that this is pointless so long as players have inconsistent hardware, and PGI doesn't have more income.

Furthermore, if updates are per-patch, what are you solving? You're not learning new ways to play every match, you're just replacing your current knowledge of other maps with less time to practice new ones. This makes absolutely no sense in a competitive environment. I think you're confusing "Oh, this would be pretty and interesting" with "Oh, this could make this game more competitve". Pro FPS teams spend countless hours learning maps in and out. What are you actually solving by attempting to remove this facet of competitive play?

I can only assume this thread was a joke, I got a good laugh.

Edited by Triskelion, 07 August 2015 - 07:35 AM.


#140 Rhaythe

    Member

  • PipPipPipPipPipPipPipPipPip
  • Philanthropist
  • Philanthropist
  • 4,203 posts

Posted 07 August 2015 - 07:36 AM

View PostMister Blastman, on 07 August 2015 - 07:28 AM, said:

I have great faith in what can be done through good coding. There are a lot of **** coders out there--more than doctors! (and there are a lot of bad doctors), but the good ones can produce neat things such as that.

I feel like you keep cherry-picking my arguments and ignoring my overall point. Which is: ultimately, it still comes down to $$$$$. So as much as we love the idea, won't happen.





5 user(s) are reading this topic

0 members, 5 guests, 0 anonymous users