Studying the AI & fully dynamic game sessions

[MarkShot]

I came across this interesting technique for understanding the AI that is more friendly and fun than reading the doctrine files.

(1) Put together whatever you want to test into a scenario.

(2) If you want purely AI, then create a controllable sub in a remote corner of the map.

(3) Run the scenario and turn on truth. You can watch how the AI performs TMA, launches weapons, evades, ...

I suppose all you scenario designers must know this already as show truth would seem a good means to debug scenarios.

One other thing I found out is that despite units being on different sides with ROE=WAR, you need TACTIC=TRANSIT SEARCH if you want hostilities. TACTIC=TRANSIT basically will result in opposed units just ignoring each other.

---

Right now, I am playing through other people's scenarios, but I have a dream of creating my own dynamic scenarios. What would make them special beyond just dynamic placement boxes, dynamic groups, probability of inclusion?

Well, I would like to do something similar to the TC (Take Command) model. This is an American Civil War game which was designed by a friend, Adam Bryant. A large scale battle goes on around the player and upon that back drop the player plays his role upon the grand stage.

So, unlike your typical SC/DW mission where you the player are tasked and are center stage, my vision would be is that there would be a larger action going on within the battle space. There would be a number of concurrent missions going on each having random elements such that the battlespace would be a dynamic world, since the units and actions going on would be too complex that events could not easily and precisely repeat. Within this dynamic world based on events and random luck, you might be called to intercept a convoy or assist support a commando team that has landed or be part of an alpha strike against a port and lead attrit enemy naval forces before the strike takes place. Part of this would be similar to WWII subsims, so your involvement and tasking would depend on bubbles of activity which you sail which instantiate objects and events and this could be done with goal oriented proximity triggers.

So, head towards a convoy forming, you'll be doing commerce raiding. Head towards a barren area of coast, you'll evacuate friendly nationals. Head towards an area where an airstrike took place, you might have some life guard duty.

This is the thing I miss in SC/DW the feeling that anything and everything is going on "out there". It all depends on where you sail to what you will encounter.

Just curious has anyone tried to create such a dynamic environment in SC/DW?

Other comments?

Thanks.

[MarkShot]

One issue would be the large number of platforms active in the scenario, since effectively there would be lots happening that was not taking place at the players location.

Another issue would be how to end the scenario and given the player some accounting and events on the larger stage.

Lastly, would the player be scored or is it enough to pass/fail missions from HQ along with staying alive as the clock runs.

A longer campaign could be built around this approach. It could be similiar to campaigns like Falcon 3 where if the players does well, generally the war progresses well for the player's side. If the player does poorly, the war goes poorly. Perhaps, the war lasts 10-20 days. A scenario then is 24 hours of dynamic events setting various flags for the next day.

---

Development would have to be done on a number of levels:

Level1 - plausibe 10-20 conventional war with the naval component.
Level2 - various flow paths through the war and feedback loops based on events at sea.
Level3 - individual day's objectives and the frag orders for that day's operation.
Level4 - translation of that day of the war into a scenario with many concurrent naval operations.
Level5 - tie in triggers based on map location and events to task the player and provide a role within a mission.
Level6 - daily player debrief global and personal.
Level7 - linkage to the next day of the war.

[MarkShot]

I have only created toy scenarios so far.

Which editor and engine would be better for a project like this SC 1.08/SCXIIC or DW 1.04/LWAMI 3.08?

For the moment, I am thinking of making this sub only and specific to one side or another.

---

I just thought of something else. Since SC/DW scenarios are simply text, you could pre-process and post-process scenarios similiar to what was done for making the static SH2 campaign into the dynamic Pacific Aces campaign (using the underlying Kriegstanz engine with a templating capability).

[MarkShot]

Does anyone know how to parse scenario files?

[Reaper51]

This all sounds very good, and I think it would make a great campaign, with a lot of replayability. So I hope you'll make it for DW, or if you decide to do it for SC, then I hope you'll port it over to DW.

[Hawk66]

Originally Posted by MarkShot

Does anyone know how to parse scenario files?

I think, it's not that diffcult, it's just a text file. Just create a very sample mission, look at its content in a text editor and the add more features to the mission etc.

Your idea of a fully dynamic session is good. I would focus on a limited time scale. You can't build a SH4/Falcon4 capable campain with the tools provided by DW (from my understanding). I see the problem in debugging (time compression is not fast enough) and the complexity.

Also, I would have a look at the doctrine scripts (not only in the mission editor, but those in the doctrine folder also). You can do additional things like an 'moving' AEW patrol etc with them.

[FERdeBOER]

You have named all the desires of [almost] all of us who have messed up with the editor. Maybe this should be discussed on the editor forum

I have also thought on a mission with objetives based half random/half player movements.

The problems are various:
- time scale/acceleration: is very limited, so transiting to one point to another will take long.

- editor: it's long time since las time I used SC's editor, but the DW ones have some problems, like: it's not possible to create a dynamic group with dynamic locations: the game crashes.
Solution: create varios dynamic groups on different places and random boxes. Problem: the maps starts to be plenty of things...

- IA behaviour: as You said, very lacy, but very VERY improved with LwAmi mod; despite that, still has bugs that will force you to "simulate" somethings (like most airstrikes). problem: simulating things adds more stuff to the editor map...

- game stability: very long missions, specially the ones with a lot of triggers and plattforms makes the game a bit unstable. Solution: simulate the actions and then make the report to the player (specially usefull if the action takes place far).
i.e.: the result of a distant battle will make possible or not air escort. Better to "make" the battle that will require to use a CV, variuos escorts, planes, bombers... and the kill or attack triggers... create a fishboat with probability of inclusion of 50%: if appears: your side wins; if not, your side loses.

- and the most difficult thing: order and memory. If you go away and don't touch the mission for a while... when you return to it you will not remember what you were doing, why did you put that there... what things you did and what are not already done... solution: have a great memory, being very organized (I'm not neither of those) and not leave without, at least, finishing that speciffic part/goal/mission/whatever...

- antoher problem is testing the mission: the more dinamicly, the more difficult to test.
Moreover, if you test one thing after included a lot of changes, triggers, scrpits... maybe you will never know what is what worked bad...

But, if you are very patient, organized, and have a lot of spare time, go ahead, please, if will love to play it

[MarkShot]

NOTE: Admin feel free to move this thread ... originally, I put it here, since I was talking more about a way of playing SC/DW, then actual techniques of mission writing. However, now, the conversation will probably be getting more technical.

I have noticed what you described with regards to long missions resulting in the CTDs.

You are proposing the "bubble" as opposed to the "full stage" method of simulation:

Bubble: Only instantiate world events/objects within a range that the player would be able to perceive them. Distant events are abstracted and the outcomes are purely summarized as opposed to being decided by simulation.

Full Stage: The entire stage is set and put in motion. Although it is not necessarily likely that the player may bring his troops a few miles (by foot) across the battle field in 20 minutes, if somehow the player achieved a break in the lines and exploits it, there will be something more than emptiness awaiting him and his brigade beyond the next ridge line.

I see that the advantage of the "bubble" in the reduction of complexity especially since doctrine language is interpreted and probably results in two orders of magnitude of simulated code lines being realized for simple scenario constructs.

However, I was hoping that the "full stage" method would result in the creation of random situations with a depth and richness not realizable by simple permutations or decision trees. The thinking was: if you apply some random objects and locations into a complex battlefield situation, then you will realize further complexity based on how things proceed from initial conditions. Now, the player has X number of hours to live in that world; maybe with an extention clause such that he needs to break contact in order to close the current session after 24 hours have passed.

Now, in terms of carry out an initial mission, he does not know what other assets might roam into his initial orders area. After completing the first mission, perhaps he now transits to pick up a convoy to escort for his second mission. What he will run across in the transit or what shape the convoy is in no one can say based on things that had happened hours earlier. Even the scenario designer after a few hours of running cannot predict the state of the space as permutations of outcomes arise rapidly and each juncture further complicate the possibilities. The world continues to become an increasingly chaotic situation despite the firing off mission orders at various intervals. With the "bubble" approach, the player can expect one of a few possible variants (especially if it is the designer or someone who has inspected the scenario). With the "full stage" approach, it is quite possible that other units with other missions and previous orders have wondered into the clean little defined problem which has just been delegated to you. Not even the designer after many hours can know what is lurking. The world is not a set of little disjoint action areas where you walk from one ride to the next in the theme park. The world is alive and while having dropped off commandos and moving to closing off a shipping lane, you may run into sub who has survived another mission and sees you as a target of opportunity along the way. Or perhaps you find yourself close to a friendly battle group that is unexpectedly being decimated by a wolfpack. A distress signal is sent to any close assets to join the fight and even the odds. Effectively, there are hostilities taking place along a broad front. At any given point in time, one cannot know where you will be along that broad front and what role you make be invited to play.

Your kills and their quality for the current session is tallied and sets force levels for the next cycle. Good performance and enemy assets being instantiated for the second day is reduced. Poor performance and the flood gates of enemy units are open. Each day (if you live), the balance of forces may shift. Perhaps, the player may trigger various difficulty options such that reinforcements will play a role in later cycle independent of performance related results. As the war drags and political solutions come closer, the player may bring an early victory to his side by helping his side to acheive sea dominance before a political peace can be imposed. Or the players has lived, but fails to be aggressive, the enemy are dominating the seas. Either the enemy wins before a political solution or the player fails to survive a sea ever saturated with hostile units. So, besides individual risk to the player, an aggressive player (despite randomness) will shift the course of the war with each cycle as the enemy suffers from attrition. A cautious captain will find the odds each cycle mounting against him. So, you have a complex dynamic where excessive focusing on realistically preserving ownship will make the war harder to survive.

[MarkShot]

You know the SC/DW community has benefitted by tremendous modding work, but at the very technical level; not at the strategic total mod level.

I would love to see this community realize a high-level strategic mod more along the lines of a Gray Wolves Expansion for SH3 or a Pacific Aces for SH2. As I have often said both of these games (SC/DW) have engines with fantastic potential. However that potential is not realized, since the world around the player is made of up of puzzles to play one after another. The world powered by the engine is not alive and fails to immerse the player in the potential that anything might happen.

[Molon Labe]

Originally Posted by MarkShot

Just curious has anyone tried to create such a dynamic environment in SC/DW?

Other comments?

Thanks.

I've had ideas to try to accomplish this from time to time.
The GDT concept was to use a human being (me) as a dynamic campaign engine to set up DW matches within a strategic context
http://www.commanders-academy.com/fo...splay.php?f=70

Of course, using a human being for that sort of thing is incredibly cumbersome and complicated, so its better to have a computer be able to do those things
http://www.subsim.com/radioroom/showthread.php?t=124314

But we may never get any help from SCS, so it's back to the drawing board...
http://www.commanders-academy.com/fo...ad.php?t=16518

I wouldn't try to pull off anything using the campaign feature as it exists now, both because I'm only really interested in MP campaigns and also because the amount of effort to get it to work would be absolutely herculean.

[kage]

It is technically doable to create 'mission of the day' for a longer campaign programmatically, by generating a mission file and reading the debrief.txt generated by DW.

----

SuBB and I have been working on a particular style of missions, that may be worth mentioning here.

We have done a lot to maintain "dynamics". I know SuBB can explain it better than I do, though it's conceivable he might dedicate a forum page or so to that.

We've been trying to make these maps MP usable, which means that the 'bubble' approach is unavoidable. (In particular since DirectPlay is involved) - Then again, there is little to no chance of the player going somewhere he would (realistically) meet another ship.

[MarkShot]

At the moment, I have put my grand schemes on back burner, since I have yet to play all the SC and DW missions hosted at SubGuru. Right now, I am still playing through the user missions for SC 1.08. Then, I will move on to SCX, SCU, and finally, LWAMI.

First, other people's scenarios although usually not very dynamic are dynamic, so to speak, the very first time you play them, and I have lots ahead of me.

Second, I am learning a lot about scenario design and the game engine by playing what others have done.

However, I still think a dynamic campaign set of scenarios or pre/post-processor would be a rebirth for these two sims. For example, SH2, was a relatively closed game system. The PA team through work done with Kriegstanz (a scenario templating engine) and pre-processors created SH/PA (Pacific Aces) and turned it into a true dynamic campaign. Sadly, like most great mods it was still hampered by original games bugs (like save games not working) and finally the arrival of SH3 killed off large scale community interest in the mod. At least, those of us who play SC/DW don't really need to worry about anything usurping these games. Yes, UBI could bring out a Cold War subsim, but it will be a hack on the SH3/4 engine focusing on graphics first and simulation second.

[FERdeBOER]

OK... I've been doing a small test.

It's about how to make acceptable random civilian "shiplanes" without finishing with the mission editor map plenty of ships and route lines.

I used "suicide attack". With this, you can place the ships wherever you want and will be sure they will head to port... sounds great, really? No! We face one of the game bugs: the AI is stupid and if an island is between the ship an its objetive...


This problem is not only with suicide attacks; the same happens whit other options.
I've tried the "pathfinding" and "avoid collision" scripts, but nothing...

But, when the area is clear between the ship and port, the "solution" works... well, not allways, as some ships instead of heading to port just turn arround 180º and others (generally those with speeds bellow 8 knots) reduce their speed to 3-4 knots. But I think we could live with that.


I've made this example with 15 minutes before the trigger fires so it can be visible the turning of the ships, but can be done just at the beginning of the scenario.

This is the editor map when this is done. As you can see, one script for all ships is enought. For making the map even more clear it can be done creating a dynamic group and all ships in the center with a random box, but I think this could make that sometimes some ships will appear together and collide.


Remains the final step, that would be to create a script to stop the ships before they collide against port. I fear this will force us to use one script for each ship, but...

kage, can You (or SuBB) tell us some of how you create your "dynamics"?

Thanks

[MarkShot]

What I would like to see for realistic behavior of platforms:

AI subs:

- Sprint and drift even when ordered to make high speed runs.

AI surface ships:

- Maybe also sprint and drift with zig-zags when ordered to make high speed runs.
- Handle emcon more intelligently. Meaning going active when there is reason to believe a contact is in close.

---

I have recently played a batch of user designed missions suffering from the following:

(1) High speed surface groups forcing escorting subs run at such a high speed that they effectively deaf and dumb.

(2) Escort subs placed in close proximity to the surface group making it impossible for them to take out an attacker before the attack can threaten the group.

(3) Surface groups running at high speed pinging away like crazy. Think about it. A surface group running at 30kts (should have been randomly zig-zaging) would pose a tracking targetting problem for a sub (especially using torpedos). If you (sub) run with it, then you can can keep pace, but you become deaf. If you slow to listen, then will not be able keep up with it make repeated attacks. However, if the group stupidly pings non-stop, then you can plot perfect solutions and wire guide wake homers while running along blind and deaf.

---

Maybe if folks are serious about producing more dynamic and realistic scenarios, we should start by listing the behaviors we want both at a high and low level and then develop solutions for each requirement. Finally, begin to build some prototypes scenarios and test them.

[GrayOwl]

DW it is not strategic sim. [Khe-Khe - if this game is possible in general is to named SIMULATOR ].

By it is designed only for individual actions. The reasonable Actions formations are simply impossible in the mission - on it you do not hope. You never will make 100 % the reliable mission - game uses casual factors for account of distances of detection - in the same mission, it will give always different distances, and you will never have the same distance at repetition of mission.

You have offered it to make with the help scripts - in individual scenario mission. But these scripts too same stupid as well as engine of game or stock doctrines... It is very a pity...

This game Live only duel events In Multiplay. It is Rock…