Ai Design And Implementation

[guyver6]

Ok, here's the topic to discuss on AI, as it's gonna be tougth to design and not easier to implement.

 

JakeDrake has started to develop AI design doc here: http://bugs.projectxenocide.com/view_bug.php?bug_id=21. As I read through it, it looks like what every programmer would need to start designing interfaces. So I propose that besides the Design Doc we could make a AI Technical Design (maybe UML?).

 

So, here's place to discuss about AI design and (later) implementation.

 

Greetings,

Guyver

[red knight]

AI Technical Design must be done after having this doc more or less ready. There are a couple of things to do already. I agree that UML like drawing could be needed for the technical design. The bad part is that they are a pain to mantain.

 

Greetings

Red Knight

[guyver6]

I've just got an idea: how about doing all documents using HTML (or XHTML)? This could give us possibility to store them on SVN. Yeah, I know that it's possible now, but Word documents are held there as binary files, and text files like html are easily modifiable. Not everyone has access to Word or does have Open Office installed.

 

Guyver

[red knight]

That is why we always keep an rtf file for everything, that can be put in the SVN repository ... Word's version control is useful for the writer cause it lets you accept changes on a per word basis, something that SVN do not let you. After something pass to a draft status we will provide the draft in the SVN repository as an RTF file.

 

HTML is not nice for people to write, it is even not nice for most programmers either (Personally I tend to hate it, I wouldnt write a document in HTML even if I get paid for it - I would convert it to it after finishing ).

 

Greetings

Red Knight

Edited December 15, 2004 by red knight

[JakeDrake]

I just noticed in the Design Document v1.5 that it refers to an "AI Design Outline" document. I tried searching for this in the forums but found nothing. Can someone point me to its location or forward it to me?

[DimmurWyrd]

I'm new here and I guess it was thought I might be able to contribute to this topic so here I am

 

IS there anything done yet? if not where to start what "Type" of AI are we talking about? decision tree or heuristic or what? to be honest the only code level AI I am familar with is the former albeit a ratehr complex variant.

 

Also what "level" of AI? tactical AI, strategic AI, etc?

 

I don't think there IS a strategic AI in a game like this of course so I will assume it's the tactical AI that needs work

 

I figure the AI will start with several "handicaps" that are lessened (or reversed) by game difficulty level so these are what I propose as a start hehe... Morale - The AI's global morale starts at a fairly low level of say 75% which means they lose morale fairly quickly if injured or if allies are lost in combat but not insanely so (a typical rookie might have a 50% or less meaning they are almost guarantee'd to break if anyone dies or if they are severely injured.) a LOWER value (around 25%) would indicate someone nearly gutless who would break if they were just shot AT... at 5-10% if they just heard a shot fired ... I shouldn't use that term "break" since I don't really mean it exactly...

 

Mostly every unit has perfect "morale" when they start a combat unless they are starting injured or maybe have fought too recently or were in too many harrowing fights etc. that is they have 100 morale points which are not the same as the units bravery (aha that's the term I meant before not morale) every time their bravery is tested they either lose or gain morale based on the situation... some situations NEVER gain but might lose less. (i.e. someones leg is blasted off they are not gonna get braver )

 

bravery controls when and how much morale is lost (just like in the originals right?) but morale loss can affect every action causing things to cost more TU, reducing accuracy and/or effectiveness of actions etc. I think for simplicity the early versions would be best using a 50 to -100 scale, and all actions are affected by that directly... for example say a shot requires 10 TU's and morale is at -50 we apply it like so... (100(default cost) -morale)% *TU's or in this instant 100- -50 = 150% = 1.5 * 10 = 15 TU's... morale scores above 0 are due to psych drugs, psi bolstering etc?, and whatnot.

 

how to apply bravery is difficult I think... essentially apply a severity and fright level to all actions that should have them (first sighting of alien, shot fired, shot fired AT someone, shot hits does light/medium/heavy/critical damage, friendly fatality, enemy fatality, etc.)

 

The fright level is the base bravery required to counteract... i.e. shot fired would probably only have a 10 or 15 rating here... anyone over that would probably be unaffected... I would do a random 25+fright > random 25+bravery = potential gain courage. < potential loss...

 

Severity then applies... if severity is very low 1-5 points then that is the potential loss to morale for the event. gain or loss would be equivalent to severity * (difference in scores)% so if an event ends up fright roll 50 bravery roll 75 and the severity was 10 then the person would GAIN 2.5 or 2 morale... if the rolls were reversed they would lose 3 (I round down for gain up for loss hehe)

 

and that's just concept 1

 

Other handicaps are "arrogance" which is a kind of override bonus/penalty to other things... i.e. they are "braver" at first until they start dying then they can't cope with dying to such primitive upstarts etc.

 

Also Organization which covers how well they stick together. i.e. will they move as a group and stay close or will they wander around alone... with arrogance will they form guard groups to cover passages or will they just wander as a mob etc.

 

Related is Teamwork or how well they WORK together (how individual they are despite their working together) i.e. will they all choose different targets or will they concentrate fire to try and take out individuals faster etc.

 

probably much I am forgetting but I am doing this off the top of my head atm

 

This is just the start of making the AI of course I am sure you know that though hehe... but it's a very important part as it sets up the specifics

[JakeDrake]

Ok well as you know I am in charge of writing the Non-Technical AI Design Document but I have not even started on the BattleScape AI yet (I only got this assignment a few days ago) so bascically yeah I could use some help on the subject. If you get an account at bugs.projectxenocide.com you can view the AI Design Document task and download what I have posted so far and get a feel for will be available when BattleScape missions start.

ex. In the current description of the PlanetScape AI, soldiers are not "remembered" and most likely will never perform more than one mission per flight so the soldier aboard a craft will not have just had a consecutive mission as you described in your last post.

Also someone will be writing the Technical Design Document sometime after I finish the Non-Technical one and then all there needs to be done is to get the gameplay working and we can implement it.

 

Welcome aboard!

[DimmurWyrd]

I suppose the first thing is to have the AI place it's units on the "board"

 

What I suggest there is something unique... Have the aliens placed inside the ship (if it's a downed alien mission) or at the entrance to a building/area for other mission types... then have them execute 1 turn every so often of geoscape/solarscape time until the player gets a landing craft there. i.e. maybe once per game hour.

 

The goals of these turns before x-com arrives is based on the mission... downed craft would be setting up a defensive perimeter or something (would depend on a table lookup of handicaps i.e. an arrogant/cowardly group might just hangout in the ship waiting... a low arrogance high bravery group might setup ambushes etc a very low arrogance very low bravery might just run and leave their craft ... very low arrogance, mediumish bravery would blow the ship first etc.

 

the goal of a terror mission is of course to do damage and kill/capture civilians making it a huge priority to get that craft moving ASAP hehe... also provides a reason to have multiple intercept transport group bases around the globe/system eh?

 

not a great deal of missions with old x-com game system of course so most of the best stuff would need to wait for V1++ hehe

 

once x-com lands the aliens would continue their initial goals UNTIL aliens spot x-com then goals would switch to "after x-com arrives" goal... maybe run away, maybe fight, etc... whatever handicaps suggest.

 

This means more global tactical variables are primary, secondary, tertiary goals based on mission type and handicaps... so that things work in a lookup table style index by mission type, index by arrogance, index by bravery... so we get an array that looks like... Array(mission type, arrogance, bravery, goals) we just plug in the values arrogance would be 1-6 (none, very low, low, medium, high, very high) , bravery similar 1-5 (very low, low, medium, high, very high) and then 1-3 we get our goals... primary, secondary, tertiary... largeish array eh? but it gives us a way to make our aliens smart yet less predictable I think at least early on... arrogance and bravery vary by alien group and success/failure of missions (i.e. if they succeed at a terror mission when x-com doesn't show up is a minor gain if they succeed without losses when x-com does show up a major gain etc. x-com kicks their butts easily a major loss and so on.)

 

now we record variables for organization, teamwork, etc. and our goals and that's where the real coding begins so I'll leave off here...

[DimmurWyrd]

Ok well as you know I am in charge of writing the Non-Technical AI Design Document but I have not even started on the BattleScape AI yet (I only got this assignment a few days ago) so bascically yeah I could use some help on the subject.  If you get an account at bugs.projectxenocide.com you can view the AI Design Document task and download what I have posted so far and get a feel for will be available when BattleScape missions start.

  ex. In the current description of the PlanetScape AI, soldiers are not "remembered" and most likely will never perform more than one mission per flight so the soldier aboard a craft will not have just had a consecutive mission as you described in your last post. 

  Also someone will be writing the Technical Design Document sometime after I finish the Non-Technical one and then all there needs to be done is to get the gameplay working and we can implement it.

 

Welcome aboard!

<{POST_SNAPBACK}>

 

ah! nother account needed will do that then hehe... also it would be very easy to add a few integers to a soldiers structure to record a "history" or rather just the results of last mission combined with previous results to get new results and so on... but it's not a big deal just a way to keep continuity

[DimmurWyrd]

eek! you'll never see me THAT organized... my logic is WAY too fuzzy for that

 

so I'll just dump ideas here and you can sort out what you like if you can figure out what I said through all the haze hehe.

 

hmmm, I am really bad at flowcharting I am really a hands-on coder :/ I work best by at LEAST doing pseudo-code mockups if nothing else... Most of that comes from trying to communicate with people who have little or no understanding of program coding however so maybe I can try again here eh?

 

Actually MOST of it comes from the fact that I am 100% self-taught without even web references and such to draw from and so I write programs in ways very foreign to the accepted norm Might explain why I don't accept many of the limitations and backwards methods many people use today because I evolved on my own instead of building on old methods and keeping such around despite the new capabilities of computers/languages? hehe. oh well.

 

let me know what you want if you want anything and I'll give it a go... otherwise best I can do is what I am now which si more or less designing the AI from scratch

 

I do not understand the term fuzzy logic, I've heard it, I know it's a form of AI etc... but I do not KNOW it... I might practice it I might not but since I again am not part of the overall community of programmers due to massive incompatibility in coding styles I don't know what many terms mean.

 

My primary skills are in data manipulation/storage/etc. but sadly I work in VB6 as stated elsewhere ... however if there is a translater they can TRY to translate my code into C but I'll give it a very low probability of working hehe.

Edited December 15, 2004 by DimmurWyrd

[red knight]

Coding style is a must for every multiperson development, this is no exception to the norm.

 

About AI we are not going to do a rules based AI (at least not for now, that is the very bottom of everything) probably it will be a Fuzzy State Machine or something that generate a better emergent behavior. However, RPG rules wont be incorporated in the AI, that is part of the games mechanics (AI is completly alien to the Simulation Engine). So if you plan to colaborate on the AI part I suggest to research Agents, and FuSM as pointed out by JakeDrake.

 

Greetings

Red Knight

[LockNLoad]

Ok well as you know I am in charge of writing the Non-Technical AI Design Document but I have not even started on the BattleScape AI yet (I only got this assignment a few days ago) so bascically yeah I could use some help on the subject.  If you get an account at bugs.projectxenocide.com you can view the AI Design Document task and download what I have posted so far and get a feel for will be available when BattleScape missions start.

 

Hi all. I've been watching this project for some time now, but this is like my second post. Bad Lurker, bad!

 

I read through the Non-Technical AI Design Doc and I have a couple of suggestions/comments. Awesome start! I've worked on design docs like this before (although never for a game) and they are hardest to get started. Once they get started, they seem to take on a life of their own, and the trick is to rein them in

 

First, you mentioned that a "cowardly" ship would run away from the player at first sight, making them impossible to shoot down at the beginning of the game. I think the way to fix this problem is by adding another attribute to the ship for "cockyness" (for lack of a better term). Basically it would represent how worried the aliens are about human intervention. At the beginning of the game, the aliens have been running around Earth for decades with absolutely no interference from the humans, so the first time their scout gets buzzed by an [interceptor] they are probably going to just shrug and keep doing what they are doing. I'm sure their attitude would change once their scouts are getting routinely shot down, even their battleships might get hesitant to engage 4 human planes after the first couple of battleships don't come home. By the end of the game, most ships (except those piloted by [mutons] of course) would probably run away if they have the option. I would use this "cockyness" attribute to either modify the Flee attribute of the ship directly (Flee - modifier), or else use it to promote/demote the ship to the next level (ie: at the beginning of the game scouts are set to Average Engagement (Flee 40)... after a couple are shot down they get a little worried, and get promoted to Cautious Engagement (Flee 75).

 

Second, I didn't see any mention of the probabilities or trigger events used to generically promote/demote states of the units. For instance, I would expect a scout to ignore an [interceptor] until they were shot at. At that point, I would expect they would switch to Cowardly Engagment and try to flee. Similarily, a Brave Engagement craft might have a 25% chance of switching to Reckless Engagement once shot at, a 50% chance of staying as Brave, and a 25% chance of turning Coward and running away, with these values being further modified by the extent of damage. By switching states rather then having a Flee value, we would only have to build the "run away" AI once, and then we could use it anytime someone should be running away. This theory is valid for just about any level of FuSM. For instance, an alien craft set on "Escort" duty might switch to "Search and Destroy" if the ship they are guarding gets shot down, or a Terror ship might decide to attack a base if it happens to fly overhead and spot it (or an assault craft might decide to go terrorize a city if their escorts get shot down enroute to the base and they chicken out)

 

Oh, and GameAI has some interesting stuff on people who are implementing fuzzy state machines, most of them you can go to the site and download the code to play with.

[JakeDrake]

Excellent, some feedback!

 

Yes, actually near the end I had thought about the ability to switch missions and personalities mid-way through a Mission (it seems silly to have them strictly adhere to what they launched with), but I decided to put if off until later because 1) Static CraftAI will be workable probably even for V1, 2) I wanted to at least get down the basics of the other AI aspects before polishing it all up and adding advanced features and 3) "Dynamic" CraftAI will rely on some of the concepts included in the OM AI.

Great ideas though! I haven't given it much thought yet because I have been trying to belt out this OM AI, which is the most important besides maybe the BS AI, so anything you can post now will help me immensely later on, thanks!

 

Right now I'm getting close to closing up the OM AI, and when I do it will still probably need lots of polising. So far I'm looking at maybe 20+ pages , if red knight wants me to cut that count down I can try to write more of just the basics, let me know.

[red knight]

It is ok keep going until you have something fleshed out... The detail is important, but dont forget, try not to write hardcoded rules by this time (it is still too early). BTW, I still have the v1.0.1 in the Issue Tracker (Try to release at least an snapshot a week or so). So we can give feedback while you work at it.

 

Greetings

Red Knight

[JakeDrake]

apologies apologies!

 

okok sorry folks but I've barely had time to get online more than 3 times over this break (since it take like 10 minutes just to get to this forum). I did try to release the latest version onto the task tracker earlier but there was an error for some reason but I am trying again right now. I haven't had the chance to get much work on the Document recently so I am very sorry. As soon as I get back to school though I will be back up to 100% productivity so then you can expect much more progress (similar to rate I was producing before the break ). That happens at the end of this week.

 

Check out the task tracker hopefully the attachment will work this time. Post some comment on the direction I'm taking with the OM AI. Once again sorry, lookin forward to being online again

[LockNLoad]

Woo Hoo, now we're getting somewhere! This is looking great.

I have quite a few notes to share, so this post is pretty long. You will notice that I have more notes on the craft AI then on the OM AI. I've had more time to ponder it is all, and it is a simpler thing to ponder

 

The craft AI is starting to look very much like a simple FSM (Finite State Machine). This is a very common AI technique in games, mainly because it is so easy to build. For the most part, it has very clear lines defined as to how and when transitions from one state to another occur. We might need to "fuzzy" the transition lines some in order to reduce the predictability of the AI.

 

1.1.2.1 As much as I like the idea of having 270 personalities for the craft, I'm afraid that some combinations don't make much sense to me (Like a reckless, sniper in an unarmed, scout ship). I think we may want to at least map which ones normally go together. From a technical standpoint, I'd store this in a file somewhere so that it can be tweaked without recompiling, and if someone wanted to mod the game to force all enemy ships to reckless, they could do so by simply making changes to the file.

 

1.2.1.1 I think I'd have the OM provide the craft with a destination point rather then an area for these movement types. Once the craft reaches the destination, let the craft AI pick how to best move around randomly (see the scouting at bottom).

 

1.3.3.2 As a personal observation, I think that plotting an intercept point is a much better idea then just always heading straight for the enemy craft you want to attack. Always heading to where the enemy craft is right now leads to a tail chase every time. In the orignal X-Com the human craft used an intercept path.

 

1.4.0.2 I'd lean towards a fuzzy transition for fleeing rather then a hard one. I'd set up a table like this to control the fleeing, with a different table being used for each Engagement level.

 

Type: Cautious

damage % % chance to flee

0 - 10 5%

11 - 20 25%

20 - 25 50%

26 - 30 80%

30 - 50 95%

50 - 100 100%

 

Again, I'd store this in a config file of some sort so it could be tweaked, but basically it would make it harder for a human to predict how the craft is going to react. In this case, even though the craft was set to a cautious engagement, there is still a (very small) chance that it won't lose it's nerve and run away until it has been fairly heavily damanged.

 

Do we want to include a "critical hit" modifyier to the fleeing? For instance, if the craft is set to Brave (so should flee around 80% damage), is it possible for them to lose their nerve if they get hit and lose 50% of their hull in one shot? This may be a V1+ addition, but I don't think it would be all that hard to add in right away.

 

1.4.0.3.2 I'm assuming that the "emergency landing" maneuver that you are refering to is what causes a craft to crash-land instead of blow up and be destroyed. We are definately going to need something like this or there isn't going to be much for the soldiers to do except sit around and wait for a battleship to land close by so they can go to work. Perhaps adding another column on the table from 1.4.0.2 so it reads like this?

 

Type: Cautious

damage % % chance to flee % chance to crash land

0 - 10 5% 0%

11 - 20 25% 5%

20 - 25 50% 10%

26 - 30 80% 25%

30 - 50 95% 60%

50 - 100 100% 80%

 

Or should the % chance to crash land be based on the type of craft instead of the pilot's attitude? (I don't care how angry you are at the bogey, if you have no engines, you are coming down, RIGHT NOW!) Maybe just set it so that the % damaged = % chance of crash landing (50% damage = 50% chance of crash landing) What does everyone thing about this?

 

1.4.1.3 Actually, we could make it so that the player could get one shot off against a craft set to cowardly engagement by redefining what "engaged in combat" means. If it means that there is a craft on an intercept course, then the scout would have plenty of time to run away. However, if it was defined as "they shot at me!", then the Gryphon-Fighter would get a single shot off before the Recon-Scout ran away. Depending on what they shot with, they might even be able to bring it down

 

1.5.0.1 What range are we refering to, detection range, weapon range, or some arbitrary value? Speaking of weapon ranges, is the OM ever going to outfit a UFO with 2 different weapon types or are the weapons already set depending on the craft?

 

1.6 Where is the infiltration mission type? Isn't that what was required in the original X-Com before a country signed a pact with the aliens?

 

1.6.1.1 Making a destination of "north america" in the ship class is going to be a real pain. Can we instead have the OM AI provide a x,y coordinate as a starting place, have the scout use a Mission-Goal path to get to the intial point, and then have craft AI do the actual scouting by picking another location, going to it, and then moving on to the next, until it runs out of gas and has to return home? (see scouting for at bottom for an example)

 

1.6.2 Might the supply ship also be the one responsible for delivering the Xenium to existing bases? If so, it would be possible for the player to cut the supply lines for a base that they know is there by ambushing the supply ship that comes in once a month (if they can catch it) Since this would be a space-capable ship, it would have to be a big supply ship, and would probably make mince-meat out of early human craft, but the payoff of nailing it (say, 200 xenium) might make it worth trying... This sounds really cool, but it should probably be a V1+ enhancement...

 

1.6.6.4 I think the escorting ship MUST use the escorted ship's radar to determine when it should try to intercept a human craft. Either that or we have to hard-code how far they can get away from the escorted ship before they return to formation. If we rely on the escorting ship's radar, it would be far to easy to pull it out of possition. To do this properly we would need a messaging system between the escorted ship and the escorting ship, which is definately a step up in complexity (unless the OM wants the ability to change a ufo's mission or recall it to base, in which case we need the messenging anyway...)

 

 

Ok, now onto the OM AI. This one is sounding a lot like a FuSM (Fuzzy State Machine). Just in case not everyone knows the difference between a FuSM and a FSM, it's basically that a FSM can exist in only one state at a time, and a FuSM exists in one or more states simultaneously. (In a FSM you can be hungry or tired, but you cannot be both. In a FuSM, you can be a little tired and a lot hungry, or not hungry at all and very tired, or very hungry and very tired) FuSMs are much better at working with Needs based AI simply because they exist in all possible states (to varying degrees) simultaneously, and the state that has the biggest imbalance is usually (but not necessarily) the first one that the AI tries to balance out.

 

So, before I continue, this is looking really good. I wish the projects I get paid to work on were this well defined on the first go, it would make my life much easier

 

 

3.1.0 Should we make this max value a setting of the difficulty level? We could handicap the OM AI by simply not allowing it to get a need up past 600. (We might have to make the equilibrium point always halfway to make this work, but that shouldn't be to hard) Perhaps we should always start this out a little lower and then let it go up every time the AI makes another base (but not go down when they lose a base) This would mean that at the end of the game the OM would be able to stockpile much more material then it could at the beginning of the game. (Which would make it possible for the OM to outfit 3 battleships and send them to 3 different bases late in the game, but not even be able to afford a battleship early on) This is kind of the same thing that you are talking about in 3.1.5 Progression, but it might help to tie both the needs max and the progression speed to the dificulty level.

 

3.1.1.3.2 What time scale are we talking about for the 2 per second thing? Is that real-time? We need to factor in the player's ability to speed up the game. We just need a frame of reference is all. Also, I would again put this setting into a config file that can be tweaked without recompiling. (Someone could mod the config file to make the OM virtually unbeatable, but hey, that's their problem

 

3.1.2.3 Since a base has a fairly large power requirement, doesn't it make sense that Base Building missions require a fairly sizable amount of Xenium available initially, not just as a maintenance thing? Or would that throw off the whole needs-balance thing?

 

3.1.4.2 From a technical standpoint this is sounding like an influence map. The globe will be broken down into areas (countries) and Alien influence (Fear) will go up when the aliens successfully complete a mission there, and it will go down when they fail the mission. Human bases would have a modifier that radiates out from their location, as would alien bases (maybe? Depends on if everyone knows they are there). If the alien influence is high enough, the OM may decide to send an infiltration team to try to subvert the government. If the alien influence is really low, the OM might send a scout out to see if there is a human base close by to hit. This sounds great, except I think I'd make the influence map a standard grid overlayed on the globe, and then add up all of the grid elements that are inside the country's borders to get the country's average influence. I would do this to balance out the size difference in some countries. I think that the original X-Com did influence based on the country, but the AI could gain a lot from using a more granular approach. The biggest advantage would be that a human base just off the US-Canada border would effect both countries almost the same amount, instead of the US getting all the benifits and Canada getting none. Does any of that make sense?

 

3.1.4.4.1 Should the retaliation mission cause high fear? Isn't X-Com a somewhat covert operation? How can blowing up a base that nobody knew was there cause fear in the general populace? I guess the government official would know, but... I'd say that Terror missions would be far more effective in causing fear then removing the human base (of course, we might need to leave it like this for game balance, otherwise the OM will always do a terror mission rather then a retaliation mission, which is clearly the wrong move in most cases) Again, if these settings were in a config file...

 

3.1.4.5.1 Should the amount of fear gained because of a successful mission taper off as the local fear rating gets higher? Once a region is basically terrified, adding another terror mission would probably get more of a "how many did they get this time?" reaction then the desired "we're all going to die!"... However, doing a terror mission in a country that has never had one should have a huge effect.

 

4.0 Is scouting required before any other mission? For instance, can a harvester go out and find a spot to harvest without a scout having found the spot already? Can a terror mission go without a scout having found a city? Can a retaliation happen if a scout hasn't already found the base? It looks to me like the scout is required for basically every mission that the OM could assign. This leads me to a couple of design issues.

 

1) When a scout goes out looking for something, is there an overall influence map that keeps track of what has been found and where, or is it random? For instance, can a harvester land again and again in Kansas, or will it take a while to get more cows in the area after they take them all the first time? I'd lean towards random, simply because it is easier to build, and certainly less prone to "cheating" by the AI.

 

2) How do we determine if the scout found something that it wants to investigate? Does it scan a radius around itself as it is flying, or does it only check when it is patrolling? I think the original X-Com only checked on patrolling, which is why you had to send out your interceptor to sit around for a couple of minutes to find the alien base that you knew was in africa. It is definately easier to build so it only checks while patrolling.

 

3) If we go the way of randomly finding the good spots on patrolling, what are the chances of finding something good with the scout stops to patrol? These settings could be stored in a config file, and should also be related to difficulty (the OM AI could be seriously handicapped if the scout only had a 1% chance of finding food. It would have to send out lots of scouts just to keep itself fed.)

 

4) Does the scout have to land in order to confirm that the site is good? In the original X-Com, scouts routinely landed on the ground to go look around before returning to base. If this is required for the various missions, how do we determine if they should land or not, and how long they should stay on the ground?

 

5) When does the OM find out about the scout's findings? Does the scout have to make it home again (meaning that if you can catch the scout that found your base and shoot it down, he can't report) or are we going to allow for these inter-galactic geniuses to figure out how to use a basic radio?

 

So, my take on the scouting goes something like this.

- OM needs food.

- The influence map says the North America is fairly neutral, so it should be fairly safe to harvest in (it will try to avoid gathering food in places the harvesters might be shot down if possible)

- Pick a medium sized scout craft, set it to a scout mission, looking for food, randomly assign the attributes to it, and give it a destination of somewhere in North America (randomly chosen)

- Scout takes off and proceeds at best speed directly to the location chosen by the OM. Once it gets there, it goes onto patrol to look at what is there. It looks for an enemy base (10% chance/second of spotting it if close enough). Not finding that, it looks for a city to terrorize (100% chance to find it if close enough). Not finding any cities, it looks for food (5% chance of finding some).

- Having struck out on all of the checks for this location, the scout picks another destination not to far away and flies to it along some path it generates.

- Reaching the next destination, it does the same 3 checks. This time if finds a city. However, it was not sent to find a city, so it simply records the location as an UNCONFIRMED city site, and picks another destination to go to.

- Reaching this new destination, it does the same 3 checks. This time it finds food.

- The UFO lands at the food site to confim it's findings.

- Assuming that the food source was confirmed (% chance), the ufo takes off and returns to base. (If it did not find food there, it would remove the unconfirmed food location from it's list and keep looking)

- Arriving at the base, the UFO reports to the OM the location of the confirmed food source. It also reports the location of an unconfirmed city.

- The OM immediately tasks a harvester ship to go to the confirmed food source and bring back the food.

- The OM checks and decides it needs more fear in the area. It sends another scout ship to confirm the city location (will have to land to do this). This ufo may return with a confirmed location of the city, and an unconfirmed location of food that it flew over and detected, but did not land at because that was not the mission.

 

Notice that during this entire exercise, the OM AI was only responsible for picking the overall mission and assigning it to an ufo. Once the ufo got the mission, it worked autonomously until it either completed the mission, was shot down, or ran out of gas and had to return to base.

 

So, there's my 2 cents... Comments?

[red knight]

I would add a couple of comments.

3.1.4.2 From a technical standpoint this is sounding like an influence map. The globe will be broken down into areas (countries) and Alien influence (Fear) will go up when the aliens successfully complete a mission there, and it will go down when they fail the mission. Human bases would have a modifier that radiates out from their location, as would alien bases (maybe? Depends on if everyone knows they are there). If the alien influence is high enough, the OM may decide to send an infiltration team to try to subvert the government. If the alien influence is really low, the OM might send a scout out to see if there is a human base close by to hit. This sounds great, except I think I'd make the influence map a standard grid overlayed on the globe, and then add up all of the grid elements that are inside the country's borders to get the country's average influence. I would do this to balance out the size difference in some countries. I think that the original X-Com did influence based on the country, but the AI could gain a lot from using a more granular approach. The biggest advantage would be that a human base just off the US-Canada border would effect both countries almost the same amount, instead of the US getting all the benifits and Canada getting none. Does any of that make sense?

Yes it makes a lot of sense, in fact that allows you to create an AI that simulates the decaying of information (thus promoting the scouting). You cannot trust outdated information, so the AI must have a forget mechanism to handle that. (that works very good with a discretized overlay.

 

4) Does the scout have to land in order to confirm that the site is good? In the original X-Com, scouts routinely landed on the ground to go look around before returning to base. If this is required for the various missions, how do we determine if they should land or not, and how long they should stay on the ground?

Nope, in fact the best way to go is to have 2 different scouting modifier (one for landing with a very high modifier) and one in patrol speed (with a lower modifier). Sometimes it is wise to not land because of high enemy ambush rates.

 

Keep working in this and we will have something very interesting, but try to make it as simple as posible. Dont overengineer it.

 

Greetings

Red Knight

[LockNLoad]

4) Does the scout have to land in order to confirm that the site is good? In the original X-Com, scouts routinely landed on the ground to go look around before returning to base. If this is required for the various missions, how do we determine if they should land or not, and how long they should stay on the ground?

Nope, in fact the best way to go is to have 2 different scouting modifier (one for landing with a very high modifier) and one in patrol speed (with a lower modifier). Sometimes it is wise to not land because of high enemy ambush rates.

 

Not sure I understand this. Do you mean that every scout that is sent out would have a chance of either doing an arial survey at low speeds, or landing, but that it would be set when the scout was launched? So in a high threat environment, the OM would set the scout to not land, and if the scout spotted a good site, it would return over it again at low altitude/low speed for an arial recon. In a low threat environment the OM would just set them to land. Did I get that right?

 

Wouldn't it be easier to just rely on the Engagement setting on the scout, so a reckless pilot is more likely to land then a cautious one? Roll a random check whenever the scout spots something interesting to see if they land or just circle a couple of times, with the chances being x% to land, y% to circle, z% to ignore and go to next waypoint? If we wanted to split it off into another attribute for the scouts (rather then use Engagement settings) we could do that to.

[red knight]

That is another modifier that we can have in mind, sometimes the OM wont allow landing if it is not in the OM interest to do so. However if the information is not present a Aggresive crew may land even if not required. ( But lets keep it simple )... Unless we are going to implement the crew style in the next release I would move into give agents less detail while still getting the OM interests fulfilled.

 

Greetings

Red Knight

[JakeDrake]

Excellent feedback!!!

 

 

I'm back now so expect updates and any ideas expressed here to be given lots of consideration. Check the bug tracker for the newest update with these corrections, which now will be highlighted in purple.

 

  1.2.1.1 I think I'd have the OM provide the craft with a destination point rather then an area for these movement types 

The reason why I chose to break with convention of the other missions goes like this: How to choose the point-based destination? The Scout will be covering basically an entire continent if left undisturbed anyway, having it go to a point somewhere in that area and only then initiating scouting behavior will look rather unnatural methinks. Also the decision-making process of determing the specific point will probably be heavily random and so won't really make any difference where it is anyway. WIth the goal being a continent or perhaps even a country, the OM can quickly and simply target a general area for possible actions. I think there is some understandable confusion (for you and I as well) over the specifics of the scouting behavior. THe reason is because I have not yet explained it in full detail. The Scout Mission is deceptively important. So far I have only described the Brain and appendages of the OM, but the Scout Mission is the primary perceptor, the eyes of it, and has been detailed at all as of yet. Once I finish OM I will flesh out the real Scout mechanisms I have in mind and hopefully things can be better critiqued.

 

1.3.3.2. For some reason I thought Xcom simply plotted a direct path (on the globe) to the enemy. Correction made, thanks!

 

1.4.0.2. I want to fuzzy up the Craft AI as well, but I have made it a lower priority than some other things because I don't think players are going to really notice it enough to make it worth a V1 addition. It definitely is worth doing for V1+ however, and I like what you've suggested. I was thinking less in tables and more in some slightly randomized functions that determine fleeing using bravery variable and possibly even a value that rates the chances of surviving based on similar previous battles. This method makes the Craft AI not only fuzzy but almost neural, which would be desirable but complicated to get right, test, etc. So for now I'm gonna put off the fuzzying for another day, but keep it in the back of your mind and I'll do the same

 

Do we want to include a "critical hit" modifyier to the fleeing?

Possibility added

 

1.4.0.3.2 I'm assuming that the "emergency landing" maneuver that you are refering to is what causes a craft to crash-land instead of blow up and be destroyed

Sorry that this was not explained in the Doc. Actually for "emergency landings" I had in mind an idea that I proposed in the labs somewhere that if a UFO retreats from a losing battle and determines that it cannot escape the enemy completely (or just gets "scared") it can force a landing even though its not in the mission plan. In other words the Intimidator craft that gets caught by an <interceptor> with <plas cannons> and manages to escape direct engagement might decide to land the craft asap, making the interceptor completely useless and forcing the player to bring out a transport with some troops and sit on the site with the <interceptor> which will probably run out of fuel before the troops come, giving the Intimidator a chance to take off before the transport reaches it and retreat or continue its mission. I'll add this description into the Doc. Thanks for bringing that to my attention.

 

In Xcom every craft crash lands if it is shot down over land.

 

1.4.1.3. I see what you are saying. Perhaps there can be an "Ignore" personality in the Craft-Engagement AI that does not take action until fired upon which then switches it to cowardly. Perfect for Scouts.

 

1.5.0.1. Fixed

 

1.6. I have never heard of this (I'm not a Superhuman pro ). Could someone PM me with a quick description of this (don't post it here)? I'll see whats up with that and add it in (Doesn't seem too hard).

 

1.6.2. This seems interesting, would require much more sophisticated Bases (each one with supplies and Needs, like a miniature OM). I think this would be best left for V1+.

 

2.6.6.4 I think the escorting ship MUST use the escorted ship's radar to determine when it should try to intercept a human craft

Agreed

 

3.1.0. I like how that sounds, but I have a feeling it will fudge up the behavior at the different difficulty levels and cause them to be inconsistent and very hard to test. It might end up being a completely different beast (all demanding of equal testing time to catch crazy behaviors) for each level. The possibility has been noted though.

 

3.1.1.3.2. Wow, I have no idea what I was thinking with that one. For now I'll simply describe it as a "Unit of Time" because the actual time frame will surely be tweaked during testing and I cannot even guess what would the correct rate.

 

3.1.2.3 Since a base has a fairly large power requirement, doesn't it make sense that Base Building missions require a fairly sizable amount of Xenium available initially

If I understand what you are suggesting correctly, perhaps a Base-Construction Mission could impose a greater requirement of Xenium to launch? If this isn't what you meant please clarify.

 

3.1.4.2 From a technical standpoint this is sounding like an influence map

Does any of that make sense?

Yes, and I love it!

This sounds so great, a very elegant solution

 

3.1.4.5.1 Should the amount of fear gained because of a successful mission taper off as the local fear rating gets higher

Sounds V1+ to me as it won't have much affect I predict, possibility noted in Doc.

 

4.0. As stated above the Scout behavior will be defined soon, the notes currently there are for my reference and don't really even say much. Everything you have suggested is really great. Some of it I had thought of some of it I had not. You will definitely see most of those ideas in the Doc.

 

 

Thank you thank you!

 

keep em coming!

Edited January 9, 2005 by JakeDrake

[LockNLoad]

In Xcom every craft crash lands if it is shot down over land.

Actually you could outright destroy ships. A very small scout ship was virtually impossible to shoot down, 2 avalanche missiles were enough to completely destroy it (no crash site). I routinely sent out my Firestorms loaded up with twin plasma cannons to intercept small scouts late in the game which proceeded to annihilate the poor ufos. Like my sig says, my way's not very sportsman-like

 

1.4.1.3.  I see what you are saying.  Perhaps there can be an "Ignore" personality in the Craft-Engagement AI that does not take action until fired upon which then switches it to cowardly.  Perfect for Scouts.

So you're thinking the normal Craft-Engagement AI classifies anything that flies close enough to be a threat, even if it is not on an intercept path? Or are you thinking of plotting courses of the human planes to try to figure out if they are trying to intercept the ufo? Or are you thinking of letting the AI cheat by getting told when the player sends a plane to intercept a ufo?

 

1.6.  I have never heard of this (I'm not a Superhuman pro ).  Could someone PM me with a quick description of this (don't post it here)?  I'll see whats up with that and add it in (Doesn't seem too hard).

I'm almost positive I used to see infiltration missions fly past once I got my hyper-wave decoder up and running. If anyone can confirm this Otherwise I'm going to have to start up a new game and prove this...

 

3.1.2.3 Since a base has a fairly large power requirement, doesn't it make sense that Base Building missions require a fairly sizable amount of Xenium available initially

If I understand what you are suggesting correctly, perhaps a Base-Construction Mission could impose a greater requirement of Xenium to launch? If this isn't what you meant please clarify.

Yup, I figure it might take some power to carve a base out of the mountains in Peru

 

I guess I'll just sit here and wait for the scouting update. Oh wait, maybe I can go do some of the work I'm being paid for...

Edited January 10, 2005 by LockNLoad

[guyver6]

I've finally found it! The Fear!

 

That's pretty nice AI library we can take a look at when we'll be deciding on implementation.

 

Another option is NeL::AI, which is part of Nevrax Library (MMORPG engine, thou AI lib can be disconnected and utilized pretty nicely, and I think it's pretty good library to use cuz it was utilized in game that acctually ships and it's RPG, which I think is pretty demanding gerne for AI).

 

Guyver

[red knight]

FEAR hasnt been updated since 9 months, that is not a good sign.

 

Greetings

Red Knight

[Beetle]

FEAR hasnt been updated since 9 months, that is not a good sign.

 

 

Yes, but we alweys can look into the source and maybe reuse some part or take some ideas (opensource )

 

In other hand if we will be only implementing another peoples code, we will miss the "fan" part od making the game:

Finally... remember that we are doing this project for the fun of it, so have fun!

 

Micah

 

[JakeDrake]

3.1.4.4.1 Should the retaliation mission cause high fear? Isn't X-Com a somewhat covert operation? How can blowing up a base that nobody knew was there cause fear in the general populace? I guess the government official would know, but... I'd say that Terror missions would be far more effective in causing fear then removing the human base (of course, we might need to leave it like this for game balance, otherwise the OM will always do a terror mission rather then a retaliation mission, which is clearly the wrong move in most cases) Again, if these settings were in a config file...

 

Sorry I didn't notice this one LockNLoad when I replied earlier to your post. Acutally you are right, obviously successfully driving back invading foot soldiers will be much more important to Fear than destroying a hidden base. What I had in mind at the time was unless XCorps covers everything up about battles, it will be slightly more impressive to see a downed battleship than a Scout, so it should simply slightly higher fear loss I suppose. Nice catch, thanks.

 

Also for future reference I should express my hesitation to simply throw things into a config file for the AI to reference. Personally I would much rather create a more intelligent system that can make the correct choices on its own (and will always come to the correct conclusion because the choice is so obvious) than telling the AI what works and what doesn't right out. That is an implementation issue however so I won't bother with it here. Let the poor guy who has to write the technical AI design doc worry bout that .

[LockNLoad]

Also for future reference I should express my hesitation to simply throw things into a config file for the AI to reference.  Personally I would much rather create a more intelligent system that can make the correct choices on its own (and will always come to the correct conclusion because the choice is so obvious) than telling the AI what works and what doesn't right out.  That is an implementation issue however so I won't bother with it here.  Let the poor guy who has to write the technical AI design doc worry bout that .

I feel the need to clarify this. (Just once, because like you said, this is a technical design issue). I'm not advocating a file that has if-then determinations in it. That being said, there are going to be a bunch of places in the AI code that compare the current calculated value against a constant in order to figure out what the "correct" choice is. All I'm advocating is to put all of these constants into a file rather then directly into the code, allowing non-code people to tweak these constants, and thereby subtly effecting the AI behavior. That way we can debate what the "correct" behaviour should be, and implement it easily to see the effects of changes...

 

This seems to be a pretty common implementation these days, especially with the current popularity of modding. I know that games like Civ3 & Galactic Civilizations stored the stats for both the units and the the different races, which allowed a person to actually change how a unit or race reacted simply by changing the values in the file, not by modding the code.

 

Anyway, something to think about. I'll leave this alone now until the technical document.

[red knight]

Dont worry, we advocate the idea of data driven game implementation, and that is exactly that.

 

Greetings

Red Knight

[JakeDrake]

Status = ONCE AGAIN IN PROGRESS! W00T! I'm so sorry for the long delay but due to some RL issues that are still ongoing I have not had much time to work on the doc the past couple months. Part of this is my fault however and on friday I kicked my butt into gear and finished up the Scouting overview, this plus the previous small amendments I had made equals something worth posting. Thanks for waiting guys I hope interest hasn't waned. Please tear the sucker apart with criticism if you can.

 

You'll notice a new feature of the doc: purple. Anything that has been changed or added since the previous posted version will be highlighted in purple. SO basically if you remeber everything else you can simply scroll to the purple parts and read them. Hopefully this will make it much easier to keep up with new versions if they take forever to make (like this last one). Enjoy!

 

Check the bugs handler for the d/l in RTF format.

Edited February 13, 2005 by JakeDrake

[red knight]

JakeDrake, can you continue working on the document at the docs.projectxenocide.com site?

 

Greetings

Red Knight

[JakeDrake]

I tried it out and I hope i did it right. Check the site and tell me if what I did is what you wanted.

[red knight]

I was looking to have the whole document reflected in an entry. But do the following when you have the document almost done (not more big changes) we will schedule a move to the documentation site.

 

Greetings

Red Knight

[Guest Azrael]

Missions

1.6.1. Scout: Search for Interesting locations.

1.6.2. Supply: Transfer supplies to a base that is being built

1.6.3. Terror: Cause havoc, Terror Site.

1.6.4. Seek & Destroy: Self-explanatory

1.6.5. Assault: Otherwise know as Retaliation. Land and assault X-Corps base

1.6.6. Escort: Protect assigned craft until mission achieved

1.6.7. Harvest: Land at “interesting” harvest sight to procure resources

1.6.8. Infiltration: Attempt to persuade

 

I'll post here to keep a record

First, missing UFO missions: Abductions, Alien Base and Alien Research.

I was wondering about the Seek & Destroy, what is its real purpose?, I mean, s & d what?, civillian? X-Corps?, a ship a base what?

I wasn't aware that we were going to introduce an Escort mission... should we write a text for it

Also, how about an Interdiction mission like in X-Com 2?, the Aliens patrol an area they are interested in with some combat craft and police it, maybe Interdiction near Alien Bases?, the Aliens attempt to intercept any craft trying to land in the base?

[JakeDrake]

-I am thinking maybe I should: Remove penalty for low food (like guyver noted, there is no point in penalizing the AI when it is getting beat), and instead of always commissioning a harvest mission when food is needed it could choose an abdution or research mission in order to gain more "food" back. Does that sound better? Also a supply mission to an area without a base builds a base is how i had it imagined.

 

-What I was imagining for Seek and Destroy is when the aliens gets PO'ed at you because you keep downing their scouts and harvesters or something, yet they don't know the location of your base; then there is a chance that they will just shoot a battleship out there where there is low Fear and attack any crafts in sight just to annoy you. Now that I think about it though this might be better replaced with just a "find enemy base" scout mission assigned to a battleship and with an attack on sight personality.

 

-I don't know if you really need a text for Escort haha. I just thought it would be an interesting AI mechanic for harder levels that makes more senes than sending out one craft out at a time disjointedly, maybe this should be V1+ though since it wasn't in Xcom

 

-Since "interdiction" wasn't in xcom1 and I know nothing about it, I'll list it as a possible mission for V1+.

 

Thanks for the great feedback!

[Guest Azrael]

-I don't know if you really need a text for Escort haha.

<{POST_SNAPBACK}>

 

Don't be silly, we always need more texts

[JakeDrake]

Azrael: I implemented your corrections into the next version (hasn't been posted).

 

Can I get some feedback on the structure or writing style of the doc? Is it easy to read? Clear and concise or is it ambiguous?

 

This will help me out not only because it will be easier for the programmers to understand but also because I hope to put the finished product into my portfolio so employers can be assured that I am able to write well, so one of you Proofreaders take a look and tell me if the writing style is ok! Thanks!

[Guest Azrael]

I'll take a look, besides, it may give me ideas for yet more texts

[Guest Azrael]

This in unsorted could be useful I think, take a look, for superhuman battlescape at least.

 

http://www.xcomufo.com/forums/index.php?showtopic=3702

[LockNLoad]

Hey JakeDrake, everyone. Sorry it took so long to respond this time. RL issues. (Since those issues involve a new house and a new baby, if I'm not changing diapers, I'm unpacking boxes... hopefully things will settle down in time... say 18 years or so...)

 

Anyway, on to the good stuff. As always I have comments, but you already knew that. The last point I put in here is just a general thing that's been perking in the back of my mind for a while, ever since we came up with the influence map for Fear. Let me know if it's completely incomprehensible.

 

4.0.2 - Who calculates where there are good resources? Is it calculated by the OM when the game is started (like the placement of resources in CIV III, even if you can't see them, they are on the map right from the start) or is it just a random chance that the scout calculates as it flies around? Calculating it at the beginning of the game provides more control, but it's harder to work with and will take more time to tune. (How many food spots are going to be needed by the OM over the course of the game? How fast do they regenerate, if at all? If you can always go back to the same one, the player is going to exploit the OM's habit of returning for food by building a base close by and ambushing the harvester ships as they come in.) Letting the scout randomly check to see what, if anything, it finds means we can't really predict when or where the scout will find something, so it is a little hard to balance, but at least the scout will return with a random spot every time, preventing the player from "camping"

 

Is there a limit to how many times a given "area" can be harvested for food? Should the probability of finding food decrease after each successful harvest mission, with it going back up gradually over time? This is sounding like a second influence map, do we want to get this fancy? I fear that the AI may always send out it's scouts looking for food in the immediate vicinity of an alien base because it is safest. If it can always find food there, why would it go anywhere else? This will firstly wreck the purpose of food scouting, and secondly make the alien base very easy for the player to stumble across. Where harvesters are found, a base is not far away. We would at least need to keep track of the last 5 or so harvest missions to target scouts in some other location.

 

4.1.1 What determines when a terror mission is a good idea? Is it when the fear is highest, and a terror mission tip the balance in favor of the aliens, allowing for a successful infiltration mission? Or is it when fear is lowest, showing that nobody is safe, and "resistance is futile"? Both maybe? Maybe a better question is when would the OM not want to do a terror mission?

 

4.2.3.1 If I read this right, the scout always has to land at a location in order to confirm that the resources it came looking for are actually in the area. Is this changed from previously? I thought that we were not going to require every scout from landing, instead there would be a chance that the resource could be confirmed from an aerial survey. In the original x-com, the scouts did not always land, most just flew around real slow and then left again. I'd think the chance to confirm that you have found what you are looking for should be a factor of the speed you are traveling. The faster you go, the less likely you are to confirm your findings. That's why in the original xcom the scout would sometimes return for another pass, but at very low speeds. Maybe have the scout always try to confirm from an aerial survey, and then land only if aerial confirmation cannot be achieved? If the scout always lands at some point (assuming it finds what it is looking for before it runs out of fuel that is), the player is going to start intercepting them with troops instead of interceptors, just trail around behind the scout until it lands, then land nearby and collect a fully operational scout ship.

 

General

Let's say the OM has a base in Mexico and X-Com has a base in Montana somewhere. To continue the story, say the OM sends out a scout looking for food with a target of the US, let's call this event "T1". Since the player has good detection capabilities, they detect the UFO shortly after take-off, and the player immediately scrambles an interceptor (T2). The UFO is still too far away from the interceptor to detect the launch, so it continues on its way looking for food. Once the interceptor and UFO close for a bit, the UFO detects the player's plane, and figures out that it is being intercepted (T3). The scout immediately makes a run for it back to base, with the interceptor in tow. Somewhere around the Mexico border, the alien base detects the interceptor (T4). The interceptor catches up with the UFO just outside the alien base and shoots it down (T5). The interceptor then flies away back home, dropping off the alien's radar at the Mexico border again (T6).

 

At which of the points (T1-6) should we adjust the Fear influence map?

 

T1 - The player has detected a UFO. Presumably the UFO came from a base that is not to far away, especially if the UFO is first detected well within the detection range of the base in Montana. The player is going to revise "their" influence map to say that there is "something down there", although they won't have many specifics at this point.

 

T2 - While the player has scrambled an interceptor, the OM has no way of knowing this at this point in time. Neither the alien base nor the UFO are close enough to see the interceptor. It is unrealistic to modify the OM's influence map in any way at this point.

 

T3 - Ok, this is where is starts getting interesting. The UFO now sees the player's interceptor, and knows that it is the target. It could, at this point, update the influence map for the OM to decrease fear in the area around where the interceptor was first spotted. (This is similar to what a human might do). Of course, we have specified that the UFO has to make it home in order to report back to the OM, so if the player can shoot down the scout fast enough, the OM would never know what happened to the UFO, and would not update the fear map at all. (This doesn’t seem to be all that fair to the OM, the player knows whenever any of their planes spot a UFO, and they know exactly where and when one of their planes goes down. Maybe just apply the “must return to base” idea on the actual survey results, not interceptions?)

 

T4 - At this point it doesn't matter if the UFO gets home or not, the OM can see the interceptor on it's own. However, where the UFO first detected the interceptor is a much better place to change the fear then where the base first detected it since the UFO can place the interceptor's original location much closer to the player's base. We could update the fear map at this point, and then re-update it if the UFO actually makes it safely back to the base. (This point is moot if we tell the OM about the interception at T3)

 

T5 - Ok, the OM is now seriously upset with the player, having just had it's scout shot down practically on it's doorstep. We could update the fear map at this point if we wanted to, but it doesn't seem quite right. The fear is going to drop right outside the alien base, not because something important happened there, but because it took the player's interceptor that long to catch the UFO. Reducing fear in the sector that the UFO actually got shot down in will tend to scatter the fear map badly I'd think, with the possibility of a sector with really high fear sitting right next to one with really low fear.

 

T6 - Similar to T4, the OM could update it's fear map at this point as well. The big advantage of doing it here instead of at T6 is that the player's interceptor is probably flying in a straight line back home at this point, making it easier to extrapolate where the player's base is. (This is something else that a human player would probably do) This kind of extrapolation, while simple for a human, is difficult to build into an AI routine, so we might want to leave it alone for now.

 

This example makes it look like the fear map should be updated when the human plane is first detected, not when a UFO is intercepted or shot down. This is a bit of a departure from what we have right now. (Note this does not address how to adjust fear based on the success/failure of other OM missions, just what happens when a UFO gets intercepted and shot down)

 

So, my 2 cents. Comments?

[JakeDrake]

Thanks for the reply (finally ) !!!

 

I'm leaving for spring break now so I'll look over your notes and reply soon, thanks for comin back!

[arrendek]

I've been thinking that the scouts and such shouldn't flee for home every time, because that would drag human ships straight to their bases. Maybe some sort of courage look up is in order, and the more courageous ones will go somewhere else. Both should call for help and the OM should possibly scramble an larger alien ship (if it's feasible) to take down the player.

[kafros]

There is something which either I miss, or isn't clear yet:

UFOs aren't able to communicate with the alien bases?!

1) No? Then I think T4 is the best step.

2) Yes? Then of course T3 is our choice!

 

I think that at T5, fear should slightly increase (A). Then, when T6 is complete:

1) If the Player patrols the area around the crash site, fear should increase further more.

2) If the Player returns to base just after T5, we have 2 possibilities:

i) The OM feels safe again, and doesn't change fear (but, the change from (A) should remain).

ii) The OM gets really pi*sed off, so he lowers fear and prepares for another mission (probably another scouting/harvesting mission, a slight chance for terror mission, and a TINY chance for retaliation mission).

 

Retaliation missions should be lauched when fear is ">95%" or "

What determines when a terror mission is a good idea? Is it when the fear is highest, and a terror mission tip the balance in favor of the aliens, allowing for a successful infiltration mission? Or is it when fear is lowest, showing that nobody is safe, and "resistance is futile"? Both maybe?

 

Another 2 cents, that's 4 cents in total! Add another 96 cents, and go buy a soft drink :wink1:

[LockNLoad]

There is something which either I miss, or isn't clear yet:

UFOs aren't able to communicate with the alien bases?!

From JakeDrake’s AI document

4.2.4. NOTE: The information gathered by the scout is NOT added to the OM’s perception of the globe until that scout returns safely home, meaning a downed scout = lost information.  This makes it possible for a scout to be a very important priority target for the player to keep track of even if it poses no direct threat.

So, at this point, we are forcing the scout to return home before it can update information with the OM. Now, this was originally centered around the scout finding a site containing some sort of raw materials (probably food), or maybe a player’s base. In the interest of playability, I think it is a good idea to leave this the way it is, gives the player a reason to chase down those pesky scouts (or at least another reason other then points and alien tech) . However, I think that the fear influence map should be updated independent of the scout’s primary mission. Remember, the OM is going to use the fear map to try to zero in on player’s bases, and also figure out what areas of the world are ripe for conquest since the player is virtually ignoring them. A human player can access the graphs of alien activity broke down by country, even if their nearest base is on the other side of the world. By updating the OM’s fear map whenever anything happens on the map that the OM can in some manner detect, it helps put the OM back up to the player’s level of knowledge, without cheating

 

Just a couple other notes ...

Fear goes up where the aliens are running rampant, and down where the player’s human forces have control. Think of it as zero where nobody cares about the fight in the first place, -1 being the player’s front door of their main base, and 1 being the farmer living next to the alien base that has lost all his cows, his children, and even his dog, to alien experiments. If this was already clear to everyone, sorry to re-iterate it. It sounded like there was a little confusion around this whole Fear thing. (JakeDrake, correct me if I’m the one out on planet Mars here)

 

Retaliation missions should be lauched when fear is ">95%" or "<5%"

I think you mean terror missions here, don’t you? Terror missions target cities; Retaliation missions target player’s bases. (right?)

 

Another 2 cents… we’ll get that soft drink in no time at this rate.

[kafros]

Actually no, I mean RETALIATION missions!

Terror missions are piece of cake to the aliens. The only problem they face is X-Corps! Thus, they HAVE to "destroy" X-Corps in order to take over/destroy/rule/whateva the world!

 

And, they probably don't want to help X-Corps win, thus they:

1) Either feel too powerful (>95%)

2) or feel too weak and do "suicide missions" (

 

They just don't want to let X-Corps (company) that it has the advantage over the war, and that the aliens can still cause havoc

[red knight]

Anyone to take over this task? As JakeDrake hasnt reported since a long time. I consider the task stalled in the meanwhile so we are waiting for a new assignee for this.

 

Greetings

Red Knight

Edited April 28, 2005 by red knight

[twhickey]

I'd be willing to take over ... I've been looking over it a lot and have some ideas.

 

Travis

 

Anyone to take over this task? As JakeDrake hasnt reported since a long time. I consider the task stalled in the meanwhile so we are waiting for a new assignee for this.

 

Greetings

Red Knight

[red knight]

Ok, go for it.

 

Greetings

Red Knight

[twhickey]

I've created a section for AI in the Wiki, and added a link to it from the main programming page. It's still under construction, but I'll work on getting as much info as possible moved over there.

 

Travis

 

AI Wiki Doc

 

Ok, go for it.

 

Greetings

Red Knight

[red knight]

Remember that the doc is already there in the attachment section of the issue tracker.

 

Greeting

Red Knight

[JakeDrake]

Once again, incredibly sorry about that everyone. this semester has turned out to be a real biotch in more than a few ways. I want you to know that I'm still checking up now and again (when i have time) and thinking about the project, I'm trying not to just mysteriously disappear like so many others. Thwickey take good care of the doc , I'll be back to help out as soon as I can! Until then...

 

Just to add my two cents, the things that I think need to be worked on most are the scout AI and consequently the resource/influence map. There should be some way for the AI to create an educated decision on where and what type of missions to commit. Right now in the doc I have suggested the influence map that was suggested by locknload. I think it is a very good idea but I couldn't take it much further than what lock said because i think it very much depends on how it can be most efficiently done in the coding. Luckily I've noticed that you are an active member of the programming team and even wrote a patch for the code recently so hopefully you can figure out how that type of thing could be implemented most easily for the programmers.

 

After that is fleshed out then the importance and behavior of scouts probably could be finished up. In my description of my ideas for the AI i tried to avoid lots of "simply random" behaviors that I'm guessing the original xcom used a lot. A fuzzy rhyme and reason will give the player a feeling that they are playing against a semi intelligent enemy. Good luck! Please post here when new versions of the doc are submitted so i'll get an alert.

[red knight]

Yes, twhickey will be working along with his mentor Rincewind on the more implementation aspects of this one... and complete only when required (and trimming of course ).

 

@twhickey: Rincewind will contact you soon, he isnt in Paris yet. Meanwhile continue working on the Design part of the document.

 

Greetings

Red Knight