Procedural Map Generation - Even Possible?
#21
Posted 23 September 2014 - 08:42 AM
Problem is the bounding (hitbox) mesh is an overlay mesh and would be game-breakingly inaccurate and would be the source of so many wails of indignation it would be deafening...
#22
Posted 23 September 2014 - 08:44 AM
DaZur, on 23 September 2014 - 08:42 AM, said:
Problem is the bounding (hitbox) mesh is an overlay mesh and would be game-breakingly inaccurate and would be the source of so many wails of indignation it would be deafening...
Yep auto nav mesh sucks, but they may have made improvements.
Edited by Johnny Z, 23 September 2014 - 08:46 AM.
#23
Posted 23 September 2014 - 08:46 AM

Really. Even assuming issues with map balance and spawn points can be worked out and you have an ideal system, map delivery becomes the killer. Before the match can begin, the map has to be generated, downloaded to 12 clients, then loaded and rendered on the clients. Even using modular resources like common textures, etc. that would reside client-side, that's a significant amount of data being pushed around before matches.
This game badly needs more maps, but that's not the solution.
#24
Posted 23 September 2014 - 08:46 AM
#25
Posted 23 September 2014 - 08:48 AM
Heffay, on 23 September 2014 - 08:23 AM, said:
We have this already...
Another thought, an auto generating map tool that makes a selection of maps every few weeks or each month.
out of these generated maps a selection is chosen by Piranha and put up in a monthly map contest, whereby the community votes for the map they'd like to see.
PGI can then go over the map and work out any issues and produce a map that people want, built to PGis standard with minimal effort.
Also gets rid of the loadtime issue there as well...
Edited by Sadist Cain, 23 September 2014 - 08:57 AM.
#26
Posted 23 September 2014 - 08:51 AM
They are working on destructible enviroments and stuff so who knows how much is is being held back etc.
They have knock down trees for instance. Why isnt it in game yet? Same goes for knock down mechs, we are not building this game so who knows what the score is. Players can only guess.
Edited by Johnny Z, 23 September 2014 - 08:54 AM.
#27
Posted 23 September 2014 - 08:52 AM
DaZur, on 23 September 2014 - 08:42 AM, said:
Problem is the bounding (hitbox) mesh is an overlay mesh and would be game-breakingly inaccurate and would be the source of so many wails of indignation it would be deafening...
I believe in CE3, that is only true for brushes (objects/props/vegetation) and entities (actors/characters/animated objects).
Its a good point to bring up because the server would have to also generate every map being played on as collision is also controlled server side..
Edited by Ghogiel, 23 September 2014 - 08:54 AM.
#28
Posted 23 September 2014 - 08:54 AM
#29
Posted 23 September 2014 - 08:55 AM
As for Faction play and the new CW Invasion mode coming later this year, this would not be feasible as the maps are being custom designed specifically to the game mode.
#30
Posted 23 September 2014 - 08:55 AM
EvilCow, on 23 September 2014 - 08:46 AM, said:
Well MW2 mercs was also a *bleeep*ton less complex mapwise.

But if that is the kind of maps you like... I am sure someone made a mod to MW2 so you can play that.
#31
Posted 23 September 2014 - 09:00 AM
Johnny Z, on 23 September 2014 - 08:51 AM, said:
Actually we do, to an extent. You can see a lot of model reuse; the ships, the cranes, buildings, cars, stuff like that. River City is a treasure trove of models if you know how to unpack the game files, and Crimson Straits reused quite a number of elements. HPG introduced yet more.
So yeah, they have really enough assets to turn out maps faster than they used to. But map creation is not just about assets. They have to ensure that maps are not overly dominated by one style of play, and to include enough variety of terrain and cover to promote play; they have to ensure that the maps have multiple routes and possibilities, to avoid games from stagnating into a single playstyle. And of course they have to QC the maps to make sure mechs don't get trapped in terrain.
The recent PTS of the upcoming Mining/ Mech Factory is a good example. The map obviously drew on existing assets, and had multiple attack routes, and despite being heavily urban, LRM attacks are still possible, and there are long range sniping spots. The diversity in play during the PTS period was affected, but not overly in favour of one playstyle or another. The design thinking going into maps is non-trivial.
And of course the PTS had its share of trapped mechs. Had mechs dropping in between buildings and getting caught in there, some times irrecoverable and protected heavily by the surrounding buildings. That kind of problems needed to be weeded out, and that simply just take a lot of time.
#32
Posted 23 September 2014 - 09:00 AM
2 major problems tho;
1) The fact this is a 3D game, unlike Diablo and other games with isometric point of view, it will be harder to "hide the misery" and all the imperfections of the generated maps, however a fixed ground that would only change its elevation and relief would simplify things up. Every match the terrain is randomly "bumped" then all the fixed elements are simply dropped on it. The big problem with this method is the ground texture that would have to modify itself with the changes on the map.
2) the other way is to segment the map with the fixed elements on the moving bits like this

The big problem with this will be the elevation of every single parts, the computer will have to compensate for these and so will have a huge potential for "falling into the void" for the players.
Procedural maps would be nice and i think the first method is safer even tho it wont be the prettiest. In any ways it would make a awesome addition to the game.
Edited by Bacl, 23 September 2014 - 09:02 AM.
#33
Posted 23 September 2014 - 09:03 AM
Lynx7725, on 23 September 2014 - 09:00 AM, said:
Actually we do, to an extent. You can see a lot of model reuse; the ships, the cranes, buildings, cars, stuff like that. River City is a treasure trove of models if you know how to unpack the game files, and Crimson Straits reused quite a number of elements. HPG introduced yet more.
So yeah, they have really enough assets to turn out maps faster than they used to. But map creation is not just about assets. They have to ensure that maps are not overly dominated by one style of play, and to include enough variety of terrain and cover to promote play; they have to ensure that the maps have multiple routes and possibilities, to avoid games from stagnating into a single playstyle. And of course they have to QC the maps to make sure mechs don't get trapped in terrain.
The recent PTS of the upcoming Mining/ Mech Factory is a good example. The map obviously drew on existing assets, and had multiple attack routes, and despite being heavily urban, LRM attacks are still possible, and there are long range sniping spots. The diversity in play during the PTS period was affected, but not overly in favour of one playstyle or another. The design thinking going into maps is non-trivial.
And of course the PTS had its share of trapped mechs. Had mechs dropping in between buildings and getting caught in there, some times irrecoverable and protected heavily by the surrounding buildings. That kind of problems needed to be weeded out, and that simply just take a lot of time.
Have to agree with all of that. But they should be getting faster with practice. Some of those spots players were getting stuck were nooby map maker mistakes
#34
Posted 23 September 2014 - 10:25 AM
The "balance" issue (or rather, the illusion of "balance") kind of makes me laugh. The MM system as it stands does not create balanced matches. It can only try and get close, and we can only hope it gets better. Couldn't a procedurally generated map be coded to try and avoid some of the major "balance" pitfalls as well - at least to the level we see through the matchmaker?
If not, I would be happy to see much larger maps with objectives that changed enough to keep them interesting. The cost of making maps has apparently been reduced, but we are still looking at only a handful of maps after more than two years of development, and most of them feel way too small for 12 vs.12 and too repetitive. What has PGI said about opening up map development to the community?
As an aside. I didn't play a lot of Mechwarrior Living Legends, but I remember the large maps opening up a greater range of tactics and play, and being a lot more entertaining, which is kind of why I started this topic.
#35
Posted 23 September 2014 - 10:45 AM
DROPSHIP, on 23 September 2014 - 10:25 AM, said:
The "balance" issue (or rather, the illusion of "balance") kind of makes me laugh. The MM system as it stands does not create balanced matches. It can only try and get close, and we can only hope it gets better. Couldn't a procedurally generated map be coded to try and avoid some of the major "balance" pitfalls as well - at least to the level we see through the matchmaker?
The answer then becomes: "Yes, but it's hard to say whether the effort to work out the parameters, vs. the result that comes out, is better than just flat out doing maps the old way."
I suppose if you parameterize the necessary stuff, it can be possible to do it. At a certain point though, locking in too many parameters would result in functionally similar maps. So if developers want to approach it that way, they have to figure out what are important parameters, what can be randomized, get the code done, and then run the code enough times in test to be assured that the stuff works well enough.
By which point, a traditional art team might have already finish a map, or two.
At that point, what we may have is one semi-random map (with attendant balance issues), and one crafted map. It really boils down to how much value the additional random map brings to the gameplay. I can't answer either way, really.. the only thing I can say is that the procedural map generation carries more risks to gameplay than the crafted map, without necessarily guaranteeing sufficient value.
I'm not necessarily against the idea of utilizing procedural generation. It's just what to generate. I think it can be great as a development tool to generate the first cut map, but I appreciate the care that PGI goes into making the maps to be as thematic and varied as they are. I'm not sure procedural generation can do that "safely".
I do think some of the other suggestions on utilizing procedural generation has merit. Things like using it to move spawn points or bases around, or to generate one-off terrain effect like rock slides, may not be a bad idea. The idea is to keep the landscape similar, so that teams can form strategies around known certainties, but introduce sufficient randomness to keep everyone on their toes, can bring some variety to games, while preserving some elements of team play.
#36
Posted 23 September 2014 - 10:55 AM
DROPSHIP, on 23 September 2014 - 10:25 AM, said:
As an aside. I didn't play a lot of Mechwarrior Living Legends, but I remember the large maps opening up a greater range of tactics and play, and being a lot more entertaining, which is kind of why I started this topic.
MWLL had many wide open maps, Which would no doubt promote stale camping or certain boring ass samey tactics in MWO, however it's not just the maps in MWLL that made it interesting, it was also the way things were balanced, fun game modes (TC), 45min matches and respawn gameplay that made MWLL games as dynamic as they were on those maps. Death Valley, one of the best mech maps ever created imo was lots of flat desert stretches, but gameplay was so versatile that I as never bored
I still can't put my finger on why and how the map was as good as it was and I spent a lot of time in dev mode messing around in it and playing the every loving **** out it as a player.
Though I would certainly like to try some MWLL maps in MWO...
Edited by Ghogiel, 23 September 2014 - 10:56 AM.
#37
Posted 23 September 2014 - 10:57 AM
#38
Posted 23 September 2014 - 10:59 AM
DROPSHIP, on 23 September 2014 - 10:57 AM, said:
That would make more sense, but it really wouldn't save too much time. Even if you have it place everything for you, you still need to create the assets that go in, create the conflict zones/lines of fire, meshes, testing, testing, and a little more testing...
The amount of time you spend generating a PG system that can get you started would probably dwarf the time it would take to just manually do 4 maps a year.
#39
Posted 23 September 2014 - 11:06 AM
DROPSHIP, on 23 September 2014 - 10:57 AM, said:
Well most level designers including myself use procedural tools to various degrees to do just that.
#40
Posted 23 September 2014 - 11:21 AM
EDIT: Actually, here's my post that initiated my thread ... even more relevant now given we had to wait 8 months for one map. In particularly, I invite you to think about the Community Warfare aspect of this, and that during an invasion, battles are definitely not fought on "balanced" battlefields.
It takes a lot of time, effort and skill to create a great multiplayer map. That's why we see relatively few of them in most multi-player games.
However, playing on the same old maps over and over gets dull. And dull is the enemy of a game's longevity.
That's why I'd like the devs to reconsider and use procedurally-generated random maps (''PGRM'') in Community Warfare.
For who don't know what a PGRM is: basically, your computer builds a map and terrain using a set of 'terrain building'' rules which randomly allocate hills, valleys and other terrain features. The map building is seeded by a single random number, so that anyone who puts the same number into the map creation engine will end up with exactly the same map. It's how Minecraft maps are created.
MWO could use the same approach, with modular ''terrain sets'' which determine the colour palette, environmental conditions, and basic type of terrain. A PGRM could be replicated client-side simply by telling each player's PC what tile set to use, and which random number to seed the PGRM engine.
Now, I do realise that PG maps wont be as good as hand-crafted multiplayer maps. Most will give an unfair advantage to one side or the other. But when you're simulating a galaxy, across hundreds of worlds and hundreds of games, the unfair advantages will even out for everyone over time. Further, that's what fighting in the IS invasion would be like... MechWarriors encountering different terrains on alien worlds, and having to do the best they can to fight from disdavantaged positions when they encounter them.
But most importantly of all, the game would never get dull. Each drop would be a new adventure. Instead of 'Yawn, Tourmeline again, all trudge to wreckage hill at d4 ... again'' your team would have to decide on the fly which way to go, which terrain features to use to advantage, etc.
Edited by Appogee, 23 September 2014 - 11:26 AM.
1 user(s) are reading this topic
0 members, 1 guests, 0 anonymous users
























