Implement new auto-evo prototype algorithm #1335

TetraspaceW · 2020-05-31T23:28:47Z

Implements the auto-evo algorithm linked in #1326

hhyyrylainen · 2020-06-01T14:03:53Z

src/auto-evo/simulation/PopulationSimulation.cs

@@ -18,7 +18,7 @@ public static void Simulate(SimulationConfiguration parameters)

            while (parameters.StepsLeft > 0)
            {
-                RunSimulationStep(parameters, speciesToSimulate, random);


I would like to keep the random as a parameter as I can imagine many scenarios where the simulation step needs randomness even if it isn't currently needed,
add _ = random to stop it from being detected as unused.

hhyyrylainen · 2020-06-01T14:05:47Z

src/auto-evo/simulation/PopulationSimulation.cs

            bool lowSpecies = species.Count <= Constants.AUTO_EVO_LOW_SPECIES_THRESHOLD;
            bool highSpecies = species.Count >= Constants.AUTO_EVO_HIGH_SPECIES_THRESHOLD;


Shouldn't this be removed from here? Now that there is kinda an auto evo algorithm this hack to keep the species count reasonable shouldn't be needed.

hhyyrylainen · 2020-06-01T14:06:33Z

src/auto-evo/simulation/PopulationSimulation.cs

-            foreach (var currentSpecies in species)
+            var speciesEnergies = new Dictionary<MicrobeSpecies, float>(species.Count);
+
+            var totalPhotosynthesisScore = 0f;


Either use 0.0f or float totalPhotosynthesisScore here instead of 0f

hhyyrylainen · 2020-06-01T14:08:27Z

src/auto-evo/simulation/PopulationSimulation.cs

+            var photosynthesisScore = 0f;
+            foreach (var organelle in species.Organelles)
+            {
+                if (organelle.Definition.InternalName == "chloroplast")


This is not extensible, instead check if the organelle has a photosynthesis process, even that should probably be checked by finding a process with sunlight as the only input and making glucose.

hhyyrylainen · 2020-06-01T14:08:52Z

src/auto-evo/simulation/PopulationSimulation.cs

+            var predationScore = 0f;
+            foreach (var organelle in species.Organelles)
+            {
+                if (organelle.Definition.InternalName == "pilus")


Should check for having the pilus component in the organelle instead of by name.

hhyyrylainen · 2020-06-01T14:09:09Z

src/auto-evo/simulation/PopulationSimulation.cs

+            var chemosynthesisScore = 0f;
+            foreach (var organelle in species.Organelles)
+            {
+                if (organelle.Definition.InternalName == "chemoplast")


Same thing here as the photosynthesis check.

hhyyrylainen · 2020-06-02T08:32:04Z

There are now some changes on master regarding how the compound data is defined in the biomes, so you'll want to merge master to this PR and fix things.

hhyyrylainen · 2020-06-02T08:36:46Z

src/auto-evo/simulation/PopulationSimulation.cs

-                int currentPopulation = populations.GetPopulationInPatch(currentSpecies, patch);
-                int populationChange = random.Next(
-                    -Constants.AUTO_EVO_RANDOM_POPULATION_CHANGE, Constants.AUTO_EVO_RANDOM_POPULATION_CHANGE + 1);
+                var processesDoneByOrganelle = organelle.Definition.RunnableProcesses;


This local is not needed, right? The foreach loop could just directly use organelle.Definition.RunnableProcesses

hhyyrylainen · 2020-06-02T08:38:43Z

src/auto-evo/simulation/PopulationSimulation.cs

+            var sunlight = SimulationParameters.Instance.GetCompound("sunlight");
+            var glucose = SimulationParameters.Instance.GetCompound("glucose");


Constantly looking these up should be avoided, making these private static readonly fields should maybe work?
I think those get initialized on the first read.

hhyyrylainen · 2020-06-02T08:39:34Z

src/auto-evo/simulation/PopulationSimulation.cs

+
+            foreach (var organelle in species.Organelles)
+            {
+                var processesDoneByOrganelle = organelle.Definition.RunnableProcesses;


this local could also be inlined.

hhyyrylainen

Code-wise this is good, but this simulation has some pretty big problems in terms of gameplay.
First of, in freebuild mode basically all the random species go extinct immediately.
And the ones that don't go extinct get huge amounts of population:

The second problem is that the split off species from the player when starting a game basically always go extinct in a single editor cycle.

So there are two choices:
You either make changes to fix the gameplay side for this new population simulation (or get someone to fix it).
Or you make a duplicate SimulatePatchStep method that has the changed version in it, but it isn't used anywhere. So we would have the logic in thrive compatible format for when someone wants to go through the needed steps to fix this up.

hhyyrylainen

Despite trying really hard to get someone to review this as well, I couldn't get anyone to do that.
So I'll comment more. Now it's definitely better that the new species don't instantly go extinct, but the glucose energy calculation should be made more accurate. Also iron should be added as an energy source, otherwise species that use iron won't appear.
Also the energy needed for population needs to be increased by about 10 as otherwise the population numbers get way too high, which will break the spawn system.
I hope you understand me asking a bunch of changes constantly as this PR will have huge gameplay implications.

hhyyrylainen · 2020-06-03T14:26:33Z

src/auto-evo/simulation/PopulationSimulation.cs

+                currentSpeciesEnergy += glucoseInPatch
+                    * 1 / species.Count;


I think that the glucose score should be done similarly to the other scores by calculating how much glucose the species can turn into ATP, doesn't make sense if a species that relies on iron gets a ton of glucose score.

ItsStolas · 2020-06-03T23:27:38Z

Despite trying really hard to get someone to review this as well, I couldn't get anyone to do that.

Hey, sorry I don't think I am a good person to review intricate gameplay changes as I haven't been following the project long enough.

athariqk

One thing I noticed is that after you play for long enough, the number of species grows large and that will probably cause a performance hit especially when the player respawns since lots of different species microbes will be spawned at once. I'm not sure, perhaps it's just the spawn system needing a rework, or maybe #1361 will fix this.

TetraspaceW · 2020-06-05T14:45:00Z

One thing I've noticed about the new algorithm is that there's no real way for species to go extinct naturally; if a species is capturing any energy at all from the environment, it's going to get a strictly positive 1/X share of the total. So a high species penalty is maybe still needed to keep species counts reasonable.

TetraspaceW · 2020-06-06T14:50:17Z

Removing flat 200 population causes discontinuous mass extinctions still (e.g. patch has capacity for 2000 population, 10 species mean average species has 200 population, so when the high species penalty fires they all go extinct), so I'll need to do more tweaking to keep species numbers in a reasonable range.

hhyyrylainen

I have just that one comment about the code. Otherwise it looks good.
@buckly90 and/or @Kasterisk could you do some game balance testing with this?

hhyyrylainen · 2020-06-10T17:10:21Z

src/microbe_stage/Patch.cs

@@ -111,6 +111,23 @@ public int GetSpeciesPopulation(Species species)
        return SpeciesInPatch[species];
    }

+    public float GetPatchChunkTotalCompoundAmount(Compound compound)


Rename this to GetTotalChunkCompoundAmount

hhyyrylainen

I didn't reread this whole thing, but I suppose now that my last code related comment was fixed, all that's left is get this approved from game balance perspective @buckly90 ?

buckly90 · 2020-07-13T00:37:20Z

From what Tjwhale has told me and what I got from testing, it seems this implementation mostly works as intended so that's good!
However, there's just a few small issues.

It apparently includes no means of extinction, so species wont die out, ever. So we obviously need to include that eventually.
Species population sizes tend to average at some really small amounts I personally feel they could be bigger but it is no pressing matter.
Large populations can suddenly vanish within one generation, I don't think that is intended and we dont exactly have any extinction events yet. From what I understood from what Tjwhale said, it could be a quirk to how the algorithm calculates energy production.
Fairly often population counts dont change at all, not quite sure what causes it.

That being said, I think Holomanga did a nice job with incorporating this into Thrive, these issues are more with the prototype than what he did I think.

hhyyrylainen

I'll mark this as needing changes so that this doesn't get merged before Buckly's comments are addressed.

hhyyrylainen · 2020-08-20T19:18:21Z

A more thought out update to my earlier comment:
The problem with freebuild (all the other species dying out quickly) is not necessarily the fault of the algorithm. But it still causes a problem where you can't easily test a cell design as all the other species die out, so you can't quickly test drive.
It would be possible to separately improve the random species generation to run auto-evo on them to select the best mutations to make the random initial species fare better.

hhyyrylainen · 2020-08-27T11:18:29Z

src/auto-evo/simulation/PopulationSimulation.cs

@@ -107,7 +113,11 @@ private static void RunSimulationStep(SimulationConfiguration parameters, List<S
            {
                // Simulate the species in each patch taking into account the already computed populations
                SimulatePatchStep(parameters.Results, entry.Value,
+<<<<<<< HEAD


This looks like a partly done merge artifact.

TjWhale · 2020-09-10T18:43:12Z

@Holomanga Hi. I had a look at your code and you've made a load of nice progress with this system, which is really cool. The prototype was quite incomplete which I think explains quite a lot of the odd behaviour.

I wanted to ask if you have any interest in joining the team and working on it with me? I made a roadmap here for how to proceed with it which I think will work.

https://forum.revolutionarygamesstudio.com/t/a-roadmap-for-implementing-auto-evo/703?u=tjwhale

And I'm around to give help and theoretical support (though unfortunately not much programming stuff) to help get it done.

I totally understand if you're busy or not interested and it's a major part of the project so it would mean a lot to us if we could get it done.

…inctions.

TetraspaceW · 2020-09-10T19:54:58Z

This commit finishes the rebasing, so it's in a state where I can start working on it again.

Over discord, IceDjuro proposed to have predation subtract energy from a species' pool of energy - this allows energies to go to zero, which would reduce a big problem (energy and hence populations are always positive so extinctions never occur).

… to compensate for the lost energy.

TetraspaceW · 2020-09-11T20:36:32Z

The philosopy behind this change is that before, predation was magic extra energy coming into the system, but now the new energy is removed later which means that it's just a redistribution of energy in favour of predators.

…ants and added checks to make sure that sent population amount is reasonable

the energy amounts a bit

now it just prints out the exception message and quits the task thread

hhyyrylainen · 2020-09-25T16:33:55Z

I added extinction by a minimum population, and did some general tweaks on this.

mine

hhyyrylainen requested changes Jun 1, 2020

View reviewed changes

hhyyrylainen requested changes Jun 2, 2020

View reviewed changes

hhyyrylainen requested review from athariqk and ItsStolas June 2, 2020 18:24

hhyyrylainen requested changes Jun 3, 2020

View reviewed changes

ItsStolas removed their request for review June 3, 2020 23:27

athariqk reviewed Jun 4, 2020

View reviewed changes

hhyyrylainen requested changes Jun 10, 2020

View reviewed changes

hhyyrylainen approved these changes Jul 12, 2020

View reviewed changes

hhyyrylainen previously requested changes Jul 13, 2020

View reviewed changes

hhyyrylainen added this to the Release 0.5.2 milestone Aug 1, 2020

hhyyrylainen reviewed Aug 27, 2020

View reviewed changes

TetraspaceW added 6 commits September 10, 2020 20:51

Count up total number of organelles of each time in biome.

2739ff0

Actually count up number of organelles of each type in patch.

1333f73

Create dict of organelles in members of that species.

c15b7dc

Get amount of sunlight and hydrogen in biome.

8e3bb02

Loop over species four times.

7115b7f

Remove old code.

7e781b2

TetraspaceW added 9 commits September 10, 2020 20:51

Get chunk compound amount from inside the Patch.

c79bcf2

Remove leftover blank lines.

dbe9df8

currentMicrobeSpeceis -> currentSpecies

477784d

Add high species penalty.

6d8c7ad

Increase energy needed for population by factor of 10.

4f49497

Decrease size of high species penalty to avoid discontinuous mass ext…

32d88ae

…inctions.

Remove high species penalty for now.

8230076

Rename to GetTotalChunkCompoundAmount.

f35718f

Remove partly done merge artifact.

babcefe

TetraspaceW added 2 commits September 11, 2020 21:16

Remove energy gained from predation; increase compound energy amounts…

66d5351

… to compensate for the lost energy.

Revert unused changes to high species penality.

25d7642

TetraspaceW and others added 9 commits September 11, 2020 21:39

Uncomment the actual change, whoops.

457759a

Add energy for predation too, though.

0658f7d

Remove accidental changes to project.godot.

24879fd

Merge branch 'master' into 1326_auto_evo_population_changes

6158c71

Did some minor cleanup on the auto-evo changes

bf42031

Added extinction to the auto-evo algorithm, cleaned up some old const…

26668b0

…ants and added checks to make sure that sent population amount is reasonable

Made toxins also add predation score

ca48e19

Made the auto evo algorithm new numbers into constants and tweaked up

ed2c7b8

the energy amounts a bit

Fixed crash when quitting the game while a background task is being ran

ddde824

now it just prints out the exception message and quits the task thread

hhyyrylainen added the review label Sep 25, 2020

hhyyrylainen self-requested a review September 25, 2020 19:07

hhyyrylainen approved these changes Sep 26, 2020

View reviewed changes

hhyyrylainen merged commit b38d406 into Revolutionary-Games:master Sep 26, 2020

hhyyrylainen mentioned this pull request Sep 26, 2020

Make initial freebuild species more viable #1650

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Implement new auto-evo prototype algorithm #1335

Implement new auto-evo prototype algorithm #1335

TetraspaceW commented May 31, 2020

hhyyrylainen Jun 1, 2020

hhyyrylainen Jun 1, 2020

hhyyrylainen Jun 1, 2020

hhyyrylainen Jun 1, 2020

hhyyrylainen Jun 1, 2020

hhyyrylainen Jun 1, 2020

hhyyrylainen commented Jun 2, 2020

hhyyrylainen Jun 2, 2020

hhyyrylainen Jun 2, 2020

hhyyrylainen Jun 2, 2020

hhyyrylainen left a comment

hhyyrylainen left a comment

hhyyrylainen Jun 3, 2020

ItsStolas commented Jun 3, 2020 •

edited

Loading

athariqk left a comment •

edited

Loading

TetraspaceW commented Jun 5, 2020 •

edited

Loading

TetraspaceW commented Jun 6, 2020

hhyyrylainen left a comment

hhyyrylainen Jun 10, 2020

hhyyrylainen left a comment

buckly90 commented Jul 13, 2020 •

edited

Loading

hhyyrylainen left a comment

hhyyrylainen commented Aug 20, 2020 •

edited

Loading

hhyyrylainen Aug 27, 2020

TjWhale commented Sep 10, 2020

TetraspaceW commented Sep 10, 2020

TetraspaceW commented Sep 11, 2020

hhyyrylainen commented Sep 25, 2020

		bool lowSpecies = species.Count <= Constants.AUTO_EVO_LOW_SPECIES_THRESHOLD;
		bool highSpecies = species.Count >= Constants.AUTO_EVO_HIGH_SPECIES_THRESHOLD;

		var sunlight = SimulationParameters.Instance.GetCompound("sunlight");
		var glucose = SimulationParameters.Instance.GetCompound("glucose");

Implement new auto-evo prototype algorithm #1335

Implement new auto-evo prototype algorithm #1335

Conversation

TetraspaceW commented May 31, 2020

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

hhyyrylainen commented Jun 2, 2020

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

hhyyrylainen left a comment

Choose a reason for hiding this comment

hhyyrylainen left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ItsStolas commented Jun 3, 2020 • edited Loading

athariqk left a comment • edited Loading

Choose a reason for hiding this comment

TetraspaceW commented Jun 5, 2020 • edited Loading

TetraspaceW commented Jun 6, 2020

hhyyrylainen left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

hhyyrylainen left a comment

Choose a reason for hiding this comment

buckly90 commented Jul 13, 2020 • edited Loading

hhyyrylainen left a comment

Choose a reason for hiding this comment

hhyyrylainen commented Aug 20, 2020 • edited Loading

Choose a reason for hiding this comment

TjWhale commented Sep 10, 2020

TetraspaceW commented Sep 10, 2020

TetraspaceW commented Sep 11, 2020

hhyyrylainen commented Sep 25, 2020

ItsStolas commented Jun 3, 2020 •

edited

Loading

athariqk left a comment •

edited

Loading

TetraspaceW commented Jun 5, 2020 •

edited

Loading

buckly90 commented Jul 13, 2020 •

edited

Loading

hhyyrylainen commented Aug 20, 2020 •

edited

Loading