Return to “Dev Logs”

Post

Re: Saturday, July 29, 2017

#16
AlexWasLike wrote:
Sun Jul 30, 2017 2:47 pm
Josh, I'm really proud of you and Adam :) Tell him I said hi and that I think the work he's doing on LT is super badass.

BFett wrote:
Sun Jul 30, 2017 1:42 pm
Is no one else impressed by how beautiful the background of that mesh is? I love those butt like nebula. That 10v10 fighter Gif is also very neat. Thank you for showing us this neat stuff Josh!
I have a browser extension called Cloud-to-Butt that changes the word "cloud" to the word "butt" that somehow still manages to catch me off guard. LOL.
This is too funny. I did laugh out loud when I read this. Thanks for sharing. :)
Image
Post

Re: Saturday, July 29, 2017

#17
JoshParnell wrote:
Sun Jul 30, 2017 8:41 pm
Interesting to hear you say that...I have always idolized surface-based CSG (i.e. not volume extraction) as the 'ultimate,' but now that I think about it, I suppose adaptive extraction does make more sense for complex CSG hierarchies. In a sense, it's like lossy compression, and will end up saving a huge amount of relatively-unnecessary tris in the case of a complex hierarchy. Cool :monkey:
surface extraction gives clean meshes and works, yes.
but badly-working unions are annoying if you want to replicate a detail you built independently into a larger mesh.
with surface extraction you have to either cut and paste the manually extracted piece of geometry which is a hassle and annoying.
or replicate the extraction process for all the places you want it to be at.

it is a very good process which i use for most of my modeling, its just not ideal for regular or repeated features on already established meshes.
Post

Re: Saturday, July 29, 2017

#18
It makes me really happy to see development speeding up thanks to the new awesome team members,

i hope you can provide tutorials on how to write blazing fast lua code for the modders then,
teach us
JoshParnell wrote:
Sat Jul 29, 2017 9:51 pm
LuaJIT Jedi Master
SDF seems to maybe also be capable of foveated rendering for VR ?
given a VR headset with eye tracking.. such things are being in development i think

that will at least save some of us from having to sell their soul to get a VBR headset and a PC that is powerful enough

Also really looking forward to see more of that Sean is doing with those AIs ..
make sure he does not accidentlly make them too smart though.. we human players have to have some chances at least

and maybe point the new team in the direction of irc / discord ? (although the discord is not linked to irc atm.. it can be done if so required)
IRC "In Josh we trust"
Post

Re: Saturday, July 29, 2017

#22
Of course I like the sound of this. Another platinum-quality update -- thank you, Josh!

Now, though, with my Very Practical Hat firmly seated, let me say that for all the reported progress, I'm still a little uncomfortable. That's because, while I can see that standards for plumbing and electrical and material quality levels are now pretty close to being agreed to, there are three construction milestones I'm waiting to hear have been met:

  • Conversion of LTSL-era PGC code is substantially complete.
  • Project AI and atomic tasks are shown working across multiple procedurally-generated star systems.
  • It feels fun. The code is meaningless if it doesn't deliver fun as a game.

The rest is polish. That's important, but it depends on accomplishing the above three milestones.

Others may have different interests; these are the keys for me to know when Limit Theory is assured of being a Real Thing that I will probably never be able to stop playing (and promoting to anyone who'll listen).

So as much as I enjoy hearing that the architecture for LT is finally looking stable and performant, the best updates from my perspective will be the ones that tell me clearly how close each of those milestones above is to being satisfied.

I remain confident that they will happen. Go, Team Procedural Reality!
Post

Re: Saturday, July 29, 2017

#23
Logging in to reply for the first time in a while.

Formations that fall below a certain numerical threshold cease to be combat effective formations, and the individual units within those formations should break off and run until they find a safe spot to regroup and rejoin the fight, if that's even what they want to do anymore. Safe could mean deep space or close to one of their larger friends with bigger guns, but hopefully that adds some ideas for how to solve the dynamic fleet structuring issue.

I can describe a few scenarios from my time commanding fleets in EVE, from most fragile to least fragile compositions:
1) You're flying a fleet comp where formation and composition is critical, like "snatchfleet" or "blink fleet," both of which make use of micro-jump fields with a small (6km) radius. All ships need to be within that radius of the ship doing the micro jump. Any ship out of that radius gets left behind when the jump occurs, and must then warp out, burn off, or be destroyed like the sitting duck it is. In this case, as long as the ships stay close together and don't die too quickly, they can stay on field and apply damage for quite a while. Drop below a certain number of micro jumpers or dps ships, however, and they have to retreat and regroup collectively, at which point decisions about re-engagement can be made.
2) You're in a kiting comp, like Orthruses, Omens, 100MN Tengus, etc. The ideal conditions for this comp have each ship speeding in roughly the same direction - say, towards the sun or a stargate. Any quick ships trying to give chase and pin these kiters down have to spend a good deal of time catching up while in range of the kiters' guns, and the kiters love nothing more than for people to chase in a straight line. As kiters are caught or forced into bad positions, they individually have to warp out - but as long as the fleet is still kiting, individuals can warp back in to a kiter and rejoin the formation. Once the kiters' escape vectors begin to narrow, or the kiters lose too many ships, the remaining kiters will scatter and regroup.
3) You're in a ship-of-the-line comp. In EVE, once a battleship fleet commits to a fight, it'll get bubbled and held down for quite a while. With long range guns, the battleships intend to stick around and apply damage across the field until they can't break enemy ships or until objectives are won or lost. Individual ships that attempt to break ranks and flee, or that get caught out in a bad position, don't usually have much of an opportunity to rejoin the fight unless the entire fleet breaks free to reposition and re-engage. These formations, however, can spread out in lines, balls, arcs, etc. and still have the range to cover each other and coordinate fire on targets, so the formation can get pretty loose and still be very effective.

Battleship fleets almost always have groups of support ships that behave independently. The healers, for example, will usually follow a healing anchor, who keeps the healers at a nice distance from the friendly battleships, and tries to keep enemies as far away as possible. Tacklers, such as interceptors and interdictors, are fairly autonomous and don't form up all that much, though I'm sure LT will look a bit different. You can also have electronic warfare support wings, who look to dance in and out of battlefields and cause as much chaos as possible until someone looks at them funny, and other disruptors like interdictors and command destroyers, who want to break apart enemy formations or keep them from retreating - they are often fairly suicidal and individual or paired-up roles. In each case, the smaller support groups rely on the presence of the ships of the line. The healers almost never cut and run because they're useless at defending themselves anyway; the tacklers and disruptors, on the other hand, are happy to run as soon as something big comes to chew on them. In almost every case, anything that runs from a smaller fight that's happening in the context of a large battle will try to communicate with its fleet to choose a new re-engagement point, where it can meet up with its friendlies and re-establish its formation in a relatively safe part of the field.
Spacecredentials: looks at stars sometimes, cheated at X-Wing vs TIE Fighter, killed a titan once.
Post

Re: Saturday, July 29, 2017

#24
fatmop wrote:
Wed Aug 02, 2017 10:38 pm
Logging in to reply for the first time in a while.

Formations that fall below a certain numerical threshold cease to be combat effective formations, and the individual units within those formations should break off and run until they find a safe spot to regroup and rejoin the fight, if that's even what they want to do anymore. Safe could mean deep space or close to one of their larger friends with bigger guns, but hopefully that adds some ideas for how to solve the dynamic fleet structuring issue.

I can describe a few scenarios from my time commanding fleets in EVE, from most fragile to least fragile compositions:
1) You're flying a fleet comp where formation and composition is critical, like "snatchfleet" or "blink fleet," both of which make use of micro-jump fields with a small (6km) radius. All ships need to be within that radius of the ship doing the micro jump. Any ship out of that radius gets left behind when the jump occurs, and must then warp out, burn off, or be destroyed like the sitting duck it is. In this case, as long as the ships stay close together and don't die too quickly, they can stay on field and apply damage for quite a while. Drop below a certain number of micro jumpers or dps ships, however, and they have to retreat and regroup collectively, at which point decisions about re-engagement can be made.
2) You're in a kiting comp, like Orthruses, Omens, 100MN Tengus, etc. The ideal conditions for this comp have each ship speeding in roughly the same direction - say, towards the sun or a stargate. Any quick ships trying to give chase and pin these kiters down have to spend a good deal of time catching up while in range of the kiters' guns, and the kiters love nothing more than for people to chase in a straight line. As kiters are caught or forced into bad positions, they individually have to warp out - but as long as the fleet is still kiting, individuals can warp back in to a kiter and rejoin the formation. Once the kiters' escape vectors begin to narrow, or the kiters lose too many ships, the remaining kiters will scatter and regroup.
3) You're in a ship-of-the-line comp. In EVE, once a battleship fleet commits to a fight, it'll get bubbled and held down for quite a while. With long range guns, the battleships intend to stick around and apply damage across the field until they can't break enemy ships or until objectives are won or lost. Individual ships that attempt to break ranks and flee, or that get caught out in a bad position, don't usually have much of an opportunity to rejoin the fight unless the entire fleet breaks free to reposition and re-engage. These formations, however, can spread out in lines, balls, arcs, etc. and still have the range to cover each other and coordinate fire on targets, so the formation can get pretty loose and still be very effective.

Battleship fleets almost always have groups of support ships that behave independently. The healers, for example, will usually follow a healing anchor, who keeps the healers at a nice distance from the friendly battleships, and tries to keep enemies as far away as possible. Tacklers, such as interceptors and interdictors, are fairly autonomous and don't form up all that much, though I'm sure LT will look a bit different. You can also have electronic warfare support wings, who look to dance in and out of battlefields and cause as much chaos as possible until someone looks at them funny, and other disruptors like interdictors and command destroyers, who want to break apart enemy formations or keep them from retreating - they are often fairly suicidal and individual or paired-up roles. In each case, the smaller support groups rely on the presence of the ships of the line. The healers almost never cut and run because they're useless at defending themselves anyway; the tacklers and disruptors, on the other hand, are happy to run as soon as something big comes to chew on them. In almost every case, anything that runs from a smaller fight that's happening in the context of a large battle will try to communicate with its fleet to choose a new re-engagement point, where it can meet up with its friendlies and re-establish its formation in a relatively safe part of the field.
This looks like a fantastic resource for extracting tactical concepts from, coming from an experienced tactician. If this understanding could be turned into code, that would be amazing.
Image
Challenging your assumptions is good for your health, good for your business, and good for your future. Stay skeptical but never undervalue the importance of a new and unfamiliar perspective.
Imagination Fertilizer
Beauty may not save the world, but it's the only thing that can
Post

Re: Saturday, July 29, 2017

#25
The question is: how can those fleet mechanics be implemented in a singleplayer game.
In Eve there are dozents to even hundrets of actual players concentrating on their specific role in the battle.

In a singleplayer game there is just the single player, who cannot possibly put that much attention to each unit. So the AI has to have autonomous responses to changes in a larger battle.
Or: its played in a classic strategy game mode, with a top-view of the battlefield and groups that get their general commands. (move there, attack this, apply this ability)

It also depends on the speed of a battle (basically determined by damage per second vs damage-protection). If a conflict is over after 30 seconds, its a different scenario than a conflict lasting for 20 minutes.
Post

Re: Saturday, July 29, 2017

#26
That is a very good question. I'd say the easy way to implement it would be that fleets default to their normal AI behavior without any player intervention, and the player can choose to intervene by using presets - a series of sliders that dictate, for example, the min and max size of a fighter squadron. Max = new fighters are shoved into a new squadron, min = squadrons below this threshold do not engage enemies until they can collect more allied fighters. Taking the time to issue real-time commands to your fleet probably isn't possible in-flight, but the right presets give you a lot of control over your fleets.
Spacecredentials: looks at stars sometimes, cheated at X-Wing vs TIE Fighter, killed a titan once.
Post

Re: Saturday, July 29, 2017

#27
In X3 there where those ship ai plugins (for trading or combat), that could be purchased.

The autonomous AI capabilities of owned and controlled ships could depend on the tech level their ai cores have / or how well trained the pilot is.

Using different roles in a larger battle still requires the player to assign those roles. (healer, dps, tank, special ...), and how to group ships into those roles (eg healer/repair squadron)
Post

Re: Saturday, July 29, 2017

#28
Wow. That's a very nice old school dev log right there. After reading this and most of people's comments I already feel a few years younger. ^.^

I almost feel sorry for all those little AI dots condemned to so bravely died in battle over and over in the name of science. ;D
Image
"Playing" is not simply a pastime, it is the primordial basis of imagination and creation. - Hideo Kojima
Post

Re: Saturday, July 29, 2017

#29
Damocles wrote:
Thu Aug 03, 2017 9:09 am
In X3 there where those ship ai plugins (for trading or combat), that could be purchased.

The autonomous AI capabilities of owned and controlled ships could depend on the tech level their ai cores have / or how well trained the pilot is.

Using different roles in a larger battle still requires the player to assign those roles. (healer, dps, tank, special ...), and how to group ships into those roles (eg healer/repair squadron)
The AI needs some way to recognize and assign those roles for its own fleets, though. Depending on what types of ships and modules are ultimately released, there will probably have to be some hardcoded roles. How those roles are used can be tweaked and experimented with, but AI needs some way to understand "This ship deals long range damage and is tanky, while this ship will die in two shots but is fast, and then this ship needs to sit close to its fleet mates."
Spacecredentials: looks at stars sometimes, cheated at X-Wing vs TIE Fighter, killed a titan once.
Post

Re: Saturday, July 29, 2017

#30
Great description, fatmop. It's a testimony to EVE's ship designs that this description, while it's somewhat EVE-specific, is really not all that far off from generic fleet combat battle plans.

The part that may not transfer to LT is the "role" model, which sort of abstracts the old tank/ranged/healer/crowd-control notion from fantasy MMORPGs.

The part that almost certainly will apply in LT is the very basic distinction between two fleet composition models:

  • 1 big ship + lots of smaller ships (e.g., carrier battle group)
  • collection of roughly equal-sized ships (e.g., strike force)

At an absolute minimum, good fleet tactics AI in LT should recognize when either of these two models is applicable and offer appropriate options. If LT (or modded LT) is more complex, cool, but this seems like the basic requirement.

(VERY important note here, though: if this -- formations -- is all there is to fleet tactics in LT, I'll be disappointed. It's the equivalent of saying that tactics in real-world combat is nothing more than how big and how many your guns are. Yawn. As I've said before, what makes tactics interesting is "terrain": in LT's case, the space equivalent of trees and hills and ditches and sheer mountainsides and rivers and marshes. Basically, the world needs to have "stuff" in it to turn to your advantage and/or your opponent's disadvantage. Otherwise, every fight is just a boring repetition of who has the most pew-pews. I went into this point in a little more detail here.)

My go-to source for the theory and practice of fleet tactics is... well... Fleet Tactics: Theory and Practice by Capt. Wayne P. Hughes, USN (ret.). It's not 100% applicable to LT, but the big-picture stuff could reasonably inspire the core design concepts for fleet tactics AI in LT.

(Of course this leaves open the question of player-directed fleet tactics. "What can I tell the ships I control to do?" I leave that to others. :D )

Online Now

Users browsing this forum: No registered users and 11 guests

cron