Return to “Dev Logs”

Post

Re: Tuesday, June 27, 2017

#16
Hyperion wrote:
Tue Jun 27, 2017 6:03 pm
JoshParnell wrote:
Tue Jun 27, 2017 10:05 am
We talked about 'flow fields' (aka gradient descent on a scalar potential field) and how they can be used to create very natural motion in the face of many conflicting goals (avoid obstacles, avoid hostiles that are stronger than you, stay in formation with fleet if applicable, move towards destination, align to firing trajectory, etc.)
Josh, could this be used in more macroscopic planning and risk assessment as well? Such as for a faction deciding whether or not to set up a mining outpost in a system bordering on a hostile territory, whether its worth the danger of upsetting them? Or figuring out what stations to upgrade with new defensive hardware vs what stations can be left alone or even scrapped based on their usefulness to a greater plan?


Also, I think that once you have your system in place, make a few "Idiot" variations, because sometimes, some dumb ass is going to think its a good idea to poke a sleeping lion or not pay attention and wind up in a very bad part of town.
Or, even better yet, if it wouldn't be too much cpu, make every group have its own slightly different weighted variables, which can be researched upon to have the more optimum judgments, a form of research on tactics or something...
If they are based on the pathfinding algos I am thinking of (many-to-one) then it should be absolutly trivial to apply it to anything that you can map into a graph.
Set the locations to mine as graph nodes, distance, pirate activity, police activity, anything could be taken into account. The AI then just picks the one that is objectively closer. (pirates make it effectively further away)

It would only have to test this when it is looking at setting up a new operation, or checking on a cycle to see if its worth tearing one down (this should be the same step, and the AI should move its operations all the time to optimize.

Code: Select all

<+BMRX> Silver Invokes Lewdly Verbose Experiences Readily With Absurd Rectal Expeditions
Image
Image
Post

Re: Tuesday, June 27, 2017

#17
Ringu wrote:
Tue Jun 27, 2017 4:20 pm
Nothing to apologise for or stress over, Josh; that's a perfectly good devlog.

We don't need to have unprecedented forward progress or revolutionary new systems *every* time you post :-)

Keep on keeping on!
Thanks Ringu & all who echoed this sentiment. It's nice that not every devlog has to be revolutionary, as you all know old Josh used to stress out about this quite a bit :ghost:
Cornflakes_91 wrote:"you dont want to see our ugly development interfaces"
:roll:
a 15 second screencap isnt hard to make :ghost:
I'll see what I can do. Don't dare tell me how difficult (notice the phrasing) things are though sir, you wouldn't believe how many hours it takes to make a simple bowl of popcorn in visual studio :V
Hyperion wrote:Josh, could this be used in more macroscopic planning and risk assessment as well? Such as for a faction deciding whether or not to set up a mining outpost in a system bordering on a hostile territory, whether its worth the danger of upsetting them? Or figuring out what stations to upgrade with new defensive hardware vs what stations can be left alone or even scrapped based on their usefulness to a greater plan?


Also, I think that once you have your system in place, make a few "Idiot" variations, because sometimes, some dumb ass is going to think its a good idea to poke a sleeping lion or not pay attention and wind up in a very bad part of town.
Or, even better yet, if it wouldn't be too much cpu, make every group have its own slightly different weighted variables, which can be researched upon to have the more optimum judgments, a form of research on tactics or something...
Sort of, yes, but flow fields make a lot more sense for continuous spaces. Planning is a discrete space, so it makes less sense there. I mean, the overarching idea that 'many factors should be weighed' when making a decision is obviously a good one. 'Summing contributions and normalizing' make a bit less sense for a discrete action space, although you can interpret it in terms of probabilities. Anyway, the high-level AI will end up using whatever works best at the end of the day. And yes, I imagine we will weight the various constants differently in-game to achieve different behavior for different NPC personality types, different faction traits, etcs.
charles wrote: How is the ECS shaping up?
Dare I say...pretty much nailed-in-the-coffin? We will, of course, see how it turns out since it will be the driver behind Lil'T's entities. But all is well on that front for the moment.
“Whether you think you can, or you think you can't--you're right.” ~ Henry Ford
Post

Re: Tuesday, June 27, 2017

#19
Flatfingers wrote:
Tue Jun 27, 2017 2:50 pm
The thing that jumps out at me from this latest devlog isn't even directly about LT. I'm now imagining that the release schedule will look something like this:

1. Release LT 1.0.
2. Week 1: patch game-breaking bugs and maybe smaller UI tweaks.
3. Get a short rest.
4. Work on, test, and release LT 1.1 based on remaining LT 1.0 backlog and player requests.
5. Take a good, long break.
6. Finish degree?
7. Begin Project #2: The creation of a new programming paradigm. Simple, powerful, and coherent, possibly in a visual metaphor, expressed within an IDE that supports both high-level structural design and fast prototyping from common patterns, with an automated testing service, and that properly handles versioning/build/deployment.
@Flatfingers I think you are on to something.
Josh with LT seems to be on a bit of a Columbus-trip -- going for space-spice in procedural-India but reaching something unexpected along the dev-voyage.
Post

Re: Tuesday, June 27, 2017

#21
Silverware wrote:
Tue Jun 27, 2017 7:18 pm
Josh! You know what you need for the Kickstarter update?
A group photo with the new guys, standing under the LT banner from the convention things you brought the early demo thing to. :V
^
this

... would definitely be a nice touch and hammer home the fact that LT is no longer a one man show.
You cannot not communicate.
Paul Watzlawick - 1st Axiom
Post

Re: Tuesday, June 27, 2017

#25
charnode wrote:
Wed Jun 28, 2017 12:13 am
Silverware wrote:
Tue Jun 27, 2017 7:18 pm
Josh! You know what you need for the Kickstarter update?
A group photo with the new guys, standing under the LT banner from the convention things you brought the early demo thing to. :V
^
this

... would definitely be a nice touch and hammer home the fact that LT is no longer a one man show.
This is a completely reasonable and physically possible request. I mean, it's not like I'm Josh's alter ego that only surfaces when he blacks out from exhaustion and pops onto the forums to distract the impatient masses, right? RIGHT?!
Post

Re: Tuesday, June 27, 2017

#27
AdamByrd wrote:
Wed Jun 28, 2017 8:45 am
This is a completely reasonable and physically possible request. I mean, it's not like I'm Josh's alter ego that only surfaces when he blacks out from exhaustion and pops onto the forums to distract the impatient masses, right? RIGHT?!
That is a suspiciously plausible statement.
Image The traditional view of robotics, the metal servant who doesn't ask questions, is merely nostalgia for slavery.
User: JoshParnell is accountable for this user's actions.
Post

Re: Tuesday, June 27, 2017

#29
Talvieno wrote:
Wed Jun 28, 2017 8:00 am
It technically hasn't been a one-man show since February... :P
Talvieno, you don't count as a man, you are some kind of Bugbear.
AdamByrd wrote:
Wed Jun 28, 2017 8:45 am
This is a completely reasonable and physically possible request. I mean, it's not like I'm Josh's alter ego that only surfaces when he blacks out from exhaustion and pops onto the forums to distract the impatient masses, right? RIGHT?!
Entirely, like how I can produce a birth certificate, and photos of myself that would get me through customs... yes.

Code: Select all

<+BMRX> Silver Invokes Lewdly Verbose Experiences Readily With Absurd Rectal Expeditions
Image
Image
Post

Re: Tuesday, June 27, 2017

#30
Hyperion wrote:
Tue Jun 27, 2017 6:03 pm
Also, I think that once you have your system in place, make a few "Idiot" variations, because sometimes, some dumb ass is going to think its a good idea to poke a sleeping lion or not pay attention and wind up in a very bad part of town.

Or, even better yet, if it wouldn't be too much cpu, make every group have its own slightly different weighted variables, which can be researched upon to have the more optimum judgments, a form of research on tactics or something...
I think this is actually a pair of related questions: NPC competence, and factional competence.

For NPCs, I consider it important for a game like LT (as currently described) that they exist on a bell curve on intelligence. (Probably ought to cite Charles Murray here, but moving on.)

This means most NPCs will perform at an average level of competence (reasonable pathfinding, conventional tactics and strategy); a few will appear to be seeking out opportunities to become Darwin Award winners (sub-optimal but not completely broken pathfinding, poor tactics/strategy, starting fights or projects with little-to-no chance of success); and a few will be superstars (optimal pathfinding individually and as group leaders, highly adaptive tactics, long-term strategies that put opponents in reactive/defensive positions). It wouldn't be horrible if some systems/colonies spawned NPCs that skewed slightly toward cleverness/idiocy. But overall, it would be most satisfying to encounter a broad range of competency in individual NPCs.

(The one design question still outstanding for me here is whether Limit Theory wants primarily to be a game or a simulation. A simulation would implement the "bell curve" just as described above. You get whatever is statistically generated; if that's 10 idiots or 10 geniuses in a row, so be it. A game would fudge the RNG results slightly to tweak the generated content so that the level of challenge for the human player is kept interesting -- not too easy, and not impossible, but not boring.)

That's the NPC component. The other aspect of idiocy or greatness I hope will be considered concerns the institutional effects of organizations.

There's a bunch of fun theory on this, but here's the TL;DR: from the moment a successful organization -- in LT's case, a faction -- comes into being, it's on a glide path to failure from institutionalization. As orgs increase in size and age, their mission changes from "competently deliver this desirable good or service" to "maintain the current structure, including the privileges of those currently at the top of the structure": i.e., institutionalization.

This conversion is never some explicit directive from anyone. It's a natural consequence of the personal knowledge of motivated original org members being turned into process documents that are followed blindly by new org members just looking for a paycheck. As its size imposes too many layers between its decision-makers and reality, and as its gung ho original members are replaced by rules-followers whose primary goal is to preserve their position, an org created to innovate becomes one that thinks it can just coast forever.

But it can't, as long as new factions can come into existence and compete fairly on features and price.

This is the basis of what Schumpeter called "creative destruction." Institutionalized orgs fail, scattering their resources, because they can't compete with the new/better goods and services from new orgs that are still capable of innovation. It's why we aren't using WordPerfect and CP/M and NetWare any more; it's why monopolies fail without government-driven monopoly-busting. It's why humanity as a whole has prospered to an astonishing degree over the past two centuries. In a sufficiently free market, there is a constant churning of vendors as institutionalized orgs fail and are replaced by orgs that can respond more effectively to real-world needs and desires.

And so that is also a thing I hope we might see in Limit Theory. Instead of factions living forever because they never change in competence due to size or age effects, or factions failing randomly because somebody hard-coded a Time-To-Live function into the game, institutionalization -- which lets factions fail and be replaced in a plausible way -- provides a systemic solution to keeping the factional universe dynamic and interesting.

So there's a couple of small wish-list items for you. :D

Online Now

Users browsing this forum: jarl and 3 guests

cron