What eagle dynamics can learn from "Liberation Dynamic Campaign"

[kunterbunt]

Wags has mentioned several times, that they are working on a dynamic campaign engine. We have not seen any outcome of that work yet.

I have seen a lot of positive remarks from people who have experienced the "Liberation Dynamic Campaign", which is a tool written by 2 people in their spare time.They seem to be doing a lot of things right.

It seems to be a game changer for DCS single player experience.

 

I believe that it would make sense for eagle dynamics to have a look at what "Liberation Dynamic Campaign" is doing right and try to learn from that.

They need to capture that feeling of flying in a world that is life no-matter if I do something or not. A world that I can enjoy even if I am not experienced and cannot help, but just fly and watch. It wold make the simulation much more interesting to newcomers too.

 

A dynamic engine does not have to be sophisticated to change our experience in the simulation.

 

@Wags: did you have a look at the "Liberation Dynamic Campaign"? I believe this is a very positive step in the right direction and should not be ignored by eagle dynamics.

[evilnate]

IMO dynamic campaign can’t happen until the computing resources of flying munitions and AI assets are significantly reduced (even RNG modeled).

[kunterbunt]

IMO dynamic campaign can’t happen until the computing resources of flying munitions and AI assets are significantly reduced (even RNG modeled).

 

What do you call „Lineration Dynamic Campain“?

[DonCalzone]

I really like the Liberation Dynamic Campaign and since it works more or less fluently on my budget rig (only 3GB VRA and, 3GHz CPU), I'm pretty sure they don't have to reduce computing resources of anything to make it happen. 2 random dudes basically made it happen. They could just pay them some Rubels and integrade their code into DCS. It would probably be less work than coding everything on their own but who am I to judge?

[evilnate]

What do you call „Lineration Dynamic Campain“?

 

Depends on what you consider as ideal dynamic campaign. I could provide example of which sim sets the bar, but I think that is against forum policies.

[Silver_Dragon]

Wags has mentioned several times, that they are working on a dynamic campaign engine. We have not seen any outcome of that work yet.

 

ED has talked they working on a RTS Dynamic Campaing.

https://forums.eagle.ru/showpost.php?p=4168367&postcount=233

Dynamic Campaign

Work on the dynamic campaign has been underway for over a year, and the progress has been steady on this very complex task. We understand that this is one of the most awaited features in DCS World and we are giving it the attention it deserves.

 

Much of the work is focused on the dynamic campaign creation tool layers as part of the Mission Editor, developing machine learning mechanics, and creating an AI neural network for the RTS element of the system. The machine learning and neural network tasks are particularly challenging, but we believe they should provide a unique dynamic campaign experience.

 

We expect to start testing this feature in the latter part of 2020.

In fact, Chizh has talked about them on some times on russian forums, the last, yesterday

https://forums.eagle.ru/showpost.php?p=4439392&postcount=1741

[jasonbirder]

To be fair if ED themselves pumped out a selection of free and paid-for content for the modules (a few SP missions with new updates in the way they add additional functionality and a regular release of paid for DLC campaigns (official ones would be better - as then ED would be obligated not to trash it a few weeks after you've purchased it) then they'd start adressing the content gap...

Dynamic Campaigns might be a bit of a stretch until ATC, WIngment and AI start working better - as you're reliant on assets acting in appropriate ways without extensive scripting...

[MacEwan]

I just hope we dont get the first release of the Dynamic Campaign the same way we got Integrated Voice Comms. Basically, barebones and barely functioning, missing tons of features, and slapped together haphazardly.

 

ED please take the time to get this working properly as much as possible before releasing it. Let it be a good representation of where the future of DCS is headed.

[SUBS17]

Do you want to know the truth about Falcon 4's Dynamic Campaign? Have a guess how long it took to code? Was it 412 000 years?:book:

[Fri13]

What i wish more than anything, is that ED rebuilds the whole DCS core of units handling. No more "Groups" but "units". It starts from individual soldier. Then pair, team, squad, platoon, company etc etc.

 

You give a single soldier parameters as moral (calm, panic, furious, frightened etc), honor, obedience, health (injuries) etc. You give each Soldier as well vision, hearing and reaction time.

 

Now you can start to have more realistic simulations. A vehicle crew can be more or less handled as one as long they stay inside vehicle. But each crew member is unique "part" of vehicle, so damage modeling applies them too.

 

You handle each unit as group, until someone happens to individuals, like a squad has three teams, and one is under fire, so you change the effected team statics from default to something negative (panicking, stressed, fear, obedience lowers, reaction times gets longer etc). Meanwhile two other teams get slight boost (honor, moral) as friends are in trouble and they perform better as start.

 

These values change the basic unit elements (speed, scouting, observation, stance etc) that are taken counted in the most important factor of all, communication.

 

Units to maintain commands structure is important. Radio calls, visual signals, audible commands etc. The command structure can not work if one can not transmit command, and other can not receive it. So you have importance to maintain that.

 

Be it a submarine that goes silent once it goes to mission, as it operates by the commands given before going to mission. Same way Sniper team goes silent once they go on mission. You don't know where they are, what they do, until they return.

A recon team might be required to break silence and report situation, like call a airstrike or report something.

A patrol team follows known route, keeps up the schedule and makes reports by the schedule.

 

What is the idea of this?

 

In real battlefield, one doesn't know all the time the exact troops positions or status. You know their commands, and last reported situations.

 

So when you command a specific unit to move from X to B, you know their commands so their route, time of departure and expected arrival times. But you do not know where they exactly are, what is happening. So you are as commanding officer, blind to any minor details that happens between those two. You can not open map and see where troops move, only their commanded locations and tasks.

 

So if something happens, you might need to revise other troops commands, depending can you contact them. They report more accurately their location, and you can issue new orders.

 

So the battlefield strategic and tactical map should be all based to delayed/old information based commands and last reports, not to exact unit positions and status in realtime.

 

This as well change requirement for heavy processing, as lots of simulation can be delayed and processed on a simpler form. Because "if no one see tree falling, do trees fall?".

 

This automatically creates new kind combat, more realistic one. Where a pilot needs to report their flight plan, as otherwise none knows where they are, and IFF can't be trusted enough alone. As well pilots can't know who they need to engage or defend as they can't go just fly and try to find visually the objects. So pilots needs to know where to go, and to whom report over radio or other means to assist them.

 

Lots of reports from enemy locations are old, obsolete or very inaccurate. So one needs to spend far more time to recon enemies and find their current locations, types, amount etc. Then get it reported back to own troops.

Some people can see you have logged a patrol flight with flight plan, but they can't see anything you see unless you report something. And reporting something just place them to general area, instead exact location.

(Example: four gun artillery, one klik on east side of town A).

 

So you get general area, unit types and count, but no other info.

Now you send a recon team on area, that needs to perform their commands and spotting, and they can report back to you more accurate data, like map coordinates etc at given accuracy. But not perfect information.

 

So how does a dynamic campaign work? Lots of it is simulation, without simulation. So in it's original idea "do something without doing it". So if two units meet, they are calculated to perform given things. It becomes like a card game where values are thrown against each others in step by step order.

 

If a two platoons makes contact, the engagement is not over in 1 second. It is long and furious combat where lots of waiting is usually happening. And all that real time can be done in real time, no need to resolve end game in a second. The CPU can be performing a lot of things meanwhile as simple, but deeper simulation is performed between these two platoons. A one engagement can last 30-180 minutes or even days depending scene, commands and many other factors.

Stalemate situation is plausible and accepted, so reinforces can be requested to get the upper hand, like an air strike or a close air support, an artillery fire mission, more troops to be transported as enforcement etc etc.

 

Logistics and commands becomes important simulation of dynamic campaign. Where going around, ambushing etc are primary means than just direct frontal assault with "to the last man" attitude like right now.

 

In real world war tactics can be used to make the simulation rules. Like 3:1 ratio in forest etc to engage and expect a victory. 6-9:1 ratio in urban areas for same.

When each unit is modeled with basic expected logic, you can only issue orders and those basic doctrines fills the time between. Like if you order unit to move given place, depending the unit it takes time to prepare for move, and then process given phases as doctrine states. There is no need for scripting and micromanagement (unless wanted) as the AI follows training. Now add there the moral, obedience, fatigue etc factors, and you can have units performing better or worse than expected. So unit that has fought whole night is not so great shape as just fresh reinforcement is. But old unit can have information that is their advantage (enemy locations, types) as new doesn't, until old informs via communication the situation , and even then new unit has just general information.

 

In dynamic campaign we don't talk 2-15 hour playtime, more like a 6-30 days.

You could speed up simulation, but there comes heavier processing and inaccuracies as some factors are simplified more.

But let's say one wants to advance campaign between 1 hour flights by a 24 hours to get a new mission, it is possible but less expected by what could happen in slower pace simulation.

So instead "card X vs card Y" as in card games, it becomes more like a "flip a coin" kind experience.

 

In realtime or almost realtime mode, combined arms and aircrafts becomes very important modules. But in fast speed mode, it is more and more like a let's see what happens if...".

[MAXsenna]

Do you want to know the truth about Falcon 4's Dynamic Campaign? Have a guess how long it took to code? Was it 412 000 years?:book:

What I don't understand is why ED wants to reinvent the wheel, and didn't aqqire the Falcon 4.0 license when it was up for grabs. I guess I'm "stoopid"!

 

Sent from my ANE-LX1 using Tapatalk

[JonathanRL]

They seem to be doing a lot of things right.

 

In your post, the important information seems to be missing.

 

What do you think they are doing right? Details help when offering feedback.

[Silver_Dragon]

What i wish more than anything, is that ED rebuilds the whole DCS core of units handling. No more "Groups" but "units". It starts from individual soldier. Then pair, team, squad, platoon, company etc etc.

 

 

 

You give a single soldier parameters as moral (calm, panic, furious, frightened etc), honor, obedience, health (injuries) etc. You give each Soldier as well vision, hearing and reaction time.

 

 

 

Now you can start to have more realistic simulations. A vehicle crew can be more or less handled as one as long they stay inside vehicle. But each crew member is unique "part" of vehicle, so damage modeling applies them too.

 

 

 

You handle each unit as group, until someone happens to individuals, like a squad has three teams, and one is under fire, so you change the effected team statics from default to something negative (panicking, stressed, fear, obedience lowers, reaction times gets longer etc). Meanwhile two other teams get slight boost (honor, moral) as friends are in trouble and they perform better as start.

 

 

 

These values change the basic unit elements (speed, scouting, observation, stance etc) that are taken counted in the most important factor of all, communication.

 

 

 

Units to maintain commands structure is important. Radio calls, visual signals, audible commands etc. The command structure can not work if one can not transmit command, and other can not receive it. So you have importance to maintain that.

 

 

 

Be it a submarine that goes silent once it goes to mission, as it operates by the commands given before going to mission. Same way Sniper team goes silent once they go on mission. You don't know where they are, what they do, until they return.

 

A recon team might be required to break silence and report situation, like call a airstrike or report something.

 

A patrol team follows known route, keeps up the schedule and makes reports by the schedule.

 

 

 

What is the idea of this?

 

 

 

In real battlefield, one doesn't know all the time the exact troops positions or status. You know their commands, and last reported situations.

 

 

 

So when you command a specific unit to move from X to B, you know their commands so their route, time of departure and expected arrival times. But you do not know where they exactly are, what is happening. So you are as commanding officer, blind to any minor details that happens between those two. You can not open map and see where troops move, only their commanded locations and tasks.

 

 

 

So if something happens, you might need to revise other troops commands, depending can you contact them. They report more accurately their location, and you can issue new orders.

 

 

 

So the battlefield strategic and tactical map should be all based to delayed/old information based commands and last reports, not to exact unit positions and status in realtime.

 

 

 

This as well change requirement for heavy processing, as lots of simulation can be delayed and processed on a simpler form. Because "if no one see tree falling, do trees fall?".

 

 

 

This automatically creates new kind combat, more realistic one. Where a pilot needs to report their flight plan, as otherwise none knows where they are, and IFF can't be trusted enough alone. As well pilots can't know who they need to engage or defend as they can't go just fly and try to find visually the objects. So pilots needs to know where to go, and to whom report over radio or other means to assist them.

 

 

 

Lots of reports from enemy locations are old, obsolete or very inaccurate. So one needs to spend far more time to recon enemies and find their current locations, types, amount etc. Then get it reported back to own troops.

 

Some people can see you have logged a patrol flight with flight plan, but they can't see anything you see unless you report something. And reporting something just place them to general area, instead exact location.

 

(Example: four gun artillery, one klik on east side of town A).

 

 

 

So you get general area, unit types and count, but no other info.

 

Now you send a recon team on area, that needs to perform their commands and spotting, and they can report back to you more accurate data, like map coordinates etc at given accuracy. But not perfect information.

 

 

 

So how does a dynamic campaign work? Lots of it is simulation, without simulation. So in it's original idea "do something without doing it". So if two units meet, they are calculated to perform given things. It becomes like a card game where values are thrown against each others in step by step order.

 

 

 

If a two platoons makes contact, the engagement is not over in 1 second. It is long and furious combat where lots of waiting is usually happening. And all that real time can be done in real time, no need to resolve end game in a second. The CPU can be performing a lot of things meanwhile as simple, but deeper simulation is performed between these two platoons. A one engagement can last 30-180 minutes or even days depending scene, commands and many other factors.

 

Stalemate situation is plausible and accepted, so reinforces can be requested to get the upper hand, like an air strike or a close air support, an artillery fire mission, more troops to be transported as enforcement etc etc.

 

 

 

Logistics and commands becomes important simulation of dynamic campaign. Where going around, ambushing etc are primary means than just direct frontal assault with "to the last man" attitude like right now.

 

 

 

In real world war tactics can be used to make the simulation rules. Like 3:1 ratio in forest etc to engage and expect a victory. 6-9:1 ratio in urban areas for same.

 

When each unit is modeled with basic expected logic, you can only issue orders and those basic doctrines fills the time between. Like if you order unit to move given place, depending the unit it takes time to prepare for move, and then process given phases as doctrine states. There is no need for scripting and micromanagement (unless wanted) as the AI follows training. Now add there the moral, obedience, fatigue etc factors, and you can have units performing better or worse than expected. So unit that has fought whole night is not so great shape as just fresh reinforcement is. But old unit can have information that is their advantage (enemy locations, types) as new doesn't, until old informs via communication the situation , and even then new unit has just general information.

 

 

 

In dynamic campaign we don't talk 2-15 hour playtime, more like a 6-30 days.

 

You could speed up simulation, but there comes heavier processing and inaccuracies as some factors are simplified more.

 

But let's say one wants to advance campaign between 1 hour flights by a 24 hours to get a new mission, it is possible but less expected by what could happen in slower pace simulation.

 

So instead "card X vs card Y" as in card games, it becomes more like a "flip a coin" kind experience.

 

 

 

In realtime or almost realtime mode, combined arms and aircrafts becomes very important modules. But in fast speed mode, it is more and more like a let's see what happens if...".

To make all calculus, scripts, AI patterns, logic situations and more, and make a "human realistic", can be require a server only.

 

Enviado desde mi RNE-L21 mediante Tapatalk

[Mars Exulte]

What I don't understand is why ED wants to reinvent the wheel, and didn't aqqire the Falcon 4.0 license when it was up for grabs. I guess I'm "stoopid"!

 

Sent from my ANE-LX1 using Tapatalk

 

If you want to play 90s games you can find them down the hall.

[MAXsenna]

If you want to play 90s games you can find them down the hall.

I was referring to the part of the code that has the dynamic campaign.

Cheers!

 

Sent from my ANE-LX1 using Tapatalk

[Fri13]

I was referring to the part of the code that has the dynamic campaign.

Cheers!

 

Sent from my ANE-LX1 using Tapatalk

 

I believe that thing could be done better, a far better today than what it is in falcon 4.

[MAXsenna]

I believe that thing could be done better, a far better today than what it is in falcon 4.

On that we totally agree!

Cheers!

 

Sent from my ANE-LX1 using Tapatalk

[Fri13]

To make all calculus, scripts, AI patterns, logic situations and more, and make a "human realistic", can be require a server only.

 

Enviado desde mi RNE-L21 mediante Tapatalk

 

Not really, as almost all is never win all the time. That is the problem in DCS that all units are calculated every second, all the time.

 

All those things are related and dependent from situations.

Example, an soldier that is on march doesn't need simulating of anything almost. He is on the move, nothing else. When there is none to see, there isn't even models or anything than information from starting and ending that is used if required to position it on map.

.

In combat lots of things happens, but combat is not 1 second action but multiple minutes if not hours. Many checks are not done unless something else requires it.

 

One of the key relents is that there is no one AI my l but multiple AI's that are unaware as well aware of others. Like if you are going to cheat, you cheat so that AI to be cheated doesn't know that. You separate AI's from each others that they can not know everything. There is no one AI to run all by scripts or by triggers etc.

 

Example, we put a single soldier on watch tower. Any soldier in a such passion is required to keep a watch (duh!) but as evidence shows, humans in watch tasks changes their behavior and they become worse in the guarding fairly quickly.

Now we have that guardpost middle of nowhere, an empty space. We do not need to give any processing for that unit unless two things happens:

 

1) someone comes to look at them (human player)

2) an enemy comes close

 

But as we don't run the guarding profile, how do we detect the enemy? We cheat. We have a standalone, independent AI for cheating, that performs checking when two idling units are to become in contact.

And when that happens, that Cheat AI will wake-up the simulation AI by poking "please, start the simulation on these units".

The same thing is with human coming to look at something (moves camera in area), simulation is started and run until camera is gone for given time (doesn't return on area by distance and time) and then when there is no human, simulation is ended.

 

Like if we do not know is enemy somewhere but it is possible, we need to keep guarding to find IF the enemy comes. It is boring to wait something possible, that likely doesn't happen. And if history is that in last 2 months nothing has come, you let your guard down differently than if 15 minutes has gone.

 

But if you would know 100% certainty that enemy doesn't come for next 3 days, 15 hours and 2 minutes, why would you guard something for next 3 days, 10 hours?

 

So, as a AI making the guarding shouldn't cheat and know things it doesn't know, we need an cheating AI that does exactly that, it turns on or off simulations of AI that is not knowing that.

 

And DCS is very slow paced game in dynamic campaign. Tens of thousands of units on the map, we can not run all for performing stuff that no one is seeing or is useless. But as we know that the campaign can be 3-4 months long as in reality, there are different schedule demands than in a game that lasts 15 minutes like in some RTS games.

The challenge is in DCS when a player wants to speed up the time, jump to days or weeks forward. As then in quick process many simulations are required to be done. But again, if we know that given guard is still guarding a given area and there is no enemy, no simulation is done and simulation for days forward is simple as changing a date, as we know the unit will be in location.

 

The most heaviest processing comes to units that are in combat mode and especially engaged and player is looking at them.

 

But as we play in realtime, lots of time to process all kind things.

Challenge is the accelerated time as more things needs to happen in shorter time period. Why we can pause game, run simulations and give a result of a dice roll.

 

Eventually we have randomness in game. Just like a normal guard game. You pull two cards and you put them against each other and you run by numbers which one will win. If a somewhere at distance where no one is seeing, two infantry squads engage each others, random numbers are used to check who wins who. It doesn't matter to large scale much if one side wins or not and how much, as there are checkings that at some point someone likely wants to retreat because too much casualties or bad numbers on their side etc.

 

Because DCS is a Real-Time-Strategy game after all. Simulation for aircrafts, but it is RTS game by it's core.

 

And if someone tries to do it as realtime simulator for everything at real time, they can't do it. Because every single unit can't be simulated all the time performing all codes and checks real time. That is what DCS does now as it has very simple design for simple simulation (no infantry, no other than basic rules for units etc) and when you collide it all with big numbers, it becomes challenging to calculate everything all the time.

 

Instead we need to cheat. Schedule things and properly simulate (do things, without really doing them) all. So there ain't enemies near by? No need to run checking do they spot an enemy...

No spotting of enemy? No need to panic or be alarmed.

As we know that whole company is not waked up from a sleep if one of night rotation guards hears a dropping rock at distance...

But we are not simulating a FPS game of special forces either.

 

The key is cheating that minimize a proper simulation from units that doesn't require it.