AI sees through walls

[Adamiks]

"First this is NOT the suggestions forum."
I know this started as saying that sappers are bad but ended on suggestions.

"Yes he want his rooms to be unkown to the AI... sorry that is not how the AI works. And knowing what it takes to program an AI, it never will."
So what the point in creating new type of enemy that is annoying because have bad AI?

"AI should see everything you can, and that is the way the AI is programmed."
Yes, but at least this should look like AI isn't seeing everything. With sappers game is almost throwing you a message "AI knows everything!" in your face.

"Yes AI on sappers escorts is very stupid, and flanking will always kill them."
Like i said - sappers add literally nothing to the game, due to bad AI. Solution - AI need to be polished. Or removed. Keeping them isn't really smart, they're easy to kill and players just are upset due to their onlogicial behaviour (even when they are easy to kill they still exist in the game).

"You are confusing the issue with rooms you build, with the ruins that are hidden. Two totally different actions and events. Yes the AI will know about the ruins, they were here before you, and have been scouting the area for decades."
For decades? Who said that? Also planets are big, which raider would have time to travel around the world and search for hidden doors using some devices that don't exist in the game? And if someone can "scout" room hidden in the mountain then he is fucking Chuck Norris.

[TLHeart]

All enemies are annoying, that is why they are part of the game. The sapper AI is no more stupid than the tribals, pirates, mechs AI who run into the kill box every time. But you are not complaining about that, because there is a long established way to easily defeat them, with no danger to your colonists.

The sappers just add another event.

Just like the wargs added another event.

The sappers just have to be adapted to, and that adaptation is very easy, flank and kill.

[Adamiks]

Sappers are easy to defeat, but you know why i'm not complaing about another enemies? Because Tynan created sappers to make killboxes useles (or worse), but he just created a another type of enemy for another type of killbox. And the point is that i want sappers to be a real danger (even if sappers would defend themselves then their behaviour is really simple and you don't need much to create "weak" points).

[Mikhail Reign]

Because Tynan created sappers to make killboxes useles (or worse)

Not really. Kill boxes aren't a new thing. They aren't even a new thing for games. Kill-boxes have existed probably as long as warfare. Getting the enemy to be where all your weapons are pointed ISN'T a bad thing. It isn't something that you can, or should even try to stop the player from doing. What it shouldn't be is the ONLY thing that you do, which is what sappers (and sieges) are trying to do - giving the player other experiences other then just forming a gun line.

The 'problem' comes down to what AI is capable of. A good example is a FPS - in it the enemy AI just has to know how to take cover, and return fire in a convincing way, as the player only interacts with each entity for as long as it take to spot and shoot them. This mirage soon evaporates the longer you interact with individual entities - for example in most fast paced shooters the AI has no understanding of advancing. Upon spawn or activation the AI will find the nearest cover and then lay down fire - if you then backtrack in the level the AI will have no idea how to react to this and simply stay put. This is why in most games when you bump up the difficulty, it either, instead of making the enemies smarter: simply adds more enemies - they are already as smart as the developers could create them, so the only way to increase the challenge is by making more of them; or gives the AI superhuman abilities - 100% accuracy, higher health, more ammunition etc . The developers haven't wasted the time pulling the blanket over the parts of the AI that rarely seen.

Another great example is RTS games. In nearly every RTS game that you come across the AI 'cheats'. Pick your favorite RTS - the AI cheats in it. They either A: can see thought the FoW B: get more/infinite resources then you C: have a higher pop cap then you or D: all of the above. Everyone that plays RTS games acknowledges and accepts this because they are aware that the AI that they are fighting against is limited - it doesnt have a full understanding of a flanking maneuver, doesn't know how to properly make use of a choke point, doesnt know how to create a defense-in-depth, it doesnt know how to bluff convincingly and it doesnt know when an attack is breaking though or being drawn into a pincer. It gets around these limitations by having access for information and resources that a player in the same position wouldn't. It gives the AI a fighting chance.

You want the sappers to be smarter? Sure - you code the AI. Until then adding either A: more of then or B: let them cheat - its the standard practice to get around AI limitations.

[NemesisN]

All enemies are annoying, that is why they are part of the game. The sapper AI is no more stupid than the tribals, pirates, mechs AI who run into the kill box every time. But you are not complaining about that, because there is a long established way to easily defeat them, with no danger to your colonists.

The sappers just add another event.

Just like the wargs added another event.

The sappers just have to be adapted to, and that adaptation is very easy, flank and kill.

not true

Sappers are the only annoying ones and the ones that feel like a troll compare to other enemies...I don't find any other enemy annoying at all

[NemesisN]

Your rooms are NOT hidden, what part of that do you not understand? just because you dug them out of a mountain, does not make them hidden. You can see them, and so can the AI. You WANT them to be hidden, but that is NOT how the game is programmed, and probably never will be. If one pawn can see the room, then all pawns can see the room, which is the way the AI is programmed in RIMWORLD.  And your colonist are pawns.

can you see hidden rooms when you never dig them out, entered or did not claim them as your own ?

No you can't because they are not yours and you never discovered them or what is inside

that is how enemy AI should work as well they should not know about rooms that they did not enter, discovered or claimed as their own

I know that every pawn can see my room including enemy...I know that is how game is currently programmed that is the point of this topic...that is the problem that I am talking about

[Adamiks]

There is so much arguments about how AI is hard to code, yep it's hard but i think that Tynan should make sappers longer, but better. In Project Zomboid devs are working more than year on NPC in zombie world that will be like real humans (they send some videos, these NPC are really cool). AI is stupid in most of the games, because:
A - computers limitations.
B - devs think that stupid AI is better than no AI.

In old games - A. In new games - B.

Example of old game AI - Counter strike with "mod" that adds bots.
In CS bots was stupid for players, but they wasn't so stupid at all. They was "talking" with each other:
"I'm going to A"
"Going with you"
"Ok, so i'm going to B"

Or:
"Terrorists in bla bla bla"
*NPC optionaly move on to help*

Or:
*Player said "follow me"
"OK"
"OK"
"NOPE! I'm going on B motherfucker!"
"I'm going with that guy on B"

In Rimworld this looks like this:
"There are turrets. There is danger zone. We should go around danger zone. We can go to Room A 56 squares in left or to Room B 45 squares right. Room A have less walls to dig. We should dig there."
"Ok, we're covering you"
*colonists are shooting to the group*
"O! Jon died! How?! We will stand in the middle of plain and watch for them!"
"Oh, no!! Sapper died! Come on, guys, we're moving to the killbox!" (reminder - colonists are still there)
"Look! There are guys over here!"

Like you think, A or B?

[Necronomocoins]

A solution could be a fog of war and line of sight with raiders/sappers/sieges wandering almost aimlessly until one of them spots an item/pawn or developments like walls, furniture or sandbags. Then beginning unified assaults or siege based on what they have seen, possibly causing a miss-raid, backfire or even surprise raids spawning possible strategic use of spotters for accurate operation of mortar etc.
I can dream too.

[Mikhail Reign]

<snip>

Ugh.... see thats.. just not how it works. Like... The AI escorts dont 'cover' the sapper, they just move with them and attack things in range, another situation in which 'dumb' AI appears smart because the role is a simple one, and the options are limited, which means that more often then not everything goes in a logical way. The illusion is broken when you use you human brain to think around the problem and put the AI in a situation in which its AI isn't designed to react to and everything goes pear shaped and you can see that the pawns aren't aware of each other in any sense other then if they are alive or dead. It is much more time effective, to create broad simple AI designed to react to at their basic level. You could spend the time making an AI which would evaluate mountain bases based solely on their external features and then try and guess where to drill in (realistically making it a crap shoot, somewhere in the area), or simply just make it so they know where the base is, and then maybe randomising X squares either way (realistically making it a crap shoot, somewhere in the area) for a time amount of the time with basically the same end result.

[RemingtonRyder]

I actually think it's better that there are stories about raids which are defeated because they're trying to accomplish a particular goal and failed because the AI could only do so much. The challenge is to defeat them before they do even more damage, not simply add some notches to your rifle.

[b0rsuk]

I treat Rimworld as a strategy/survival game. The lack of fog of war, total awareness granted by the satellite view removes all mystery. Why should enemies be surprised by my base layout if I can see everything ? It just doesn't fit. Few top-down games even bother having any sort of line of sight.

[Adamiks]

<snip>

Ugh.... see thats.. just not how it works. Like... The AI escorts dont 'cover' the sapper, they just move with them and attack things in range, another situation in which 'dumb' AI appears smart because the role is a simple one, and the options are limited, which means that more often then not everything goes in a logical way. The illusion is broken when you use you human brain to think around the problem and put the AI in a situation in which its AI isn't designed to react to and everything goes pear shaped and you can see that the pawns aren't aware of each other in any sense other then if they are alive or dead. It is much more time effective, to create broad simple AI designed to react to at their basic level. You could spend the time making an AI which would evaluate mountain bases based solely on their external features and then try and guess where to drill in (realistically making it a crap shoot, somewhere in the area), or simply just make it so they know where the base is, and then maybe randomising X squares either way (realistically making it a crap shoot, somewhere in the area) for a time amount of the time with basically the same end result.

You think that i'm stupid or from the future? Every AI in games works like this. If player will can do something to AI but AI don't have programmed reaction, well.... AI will probably do nothing. The point is that in my example AI had programmed reactions and if not, then that wasn't so visible, in Rimworld sappers are just dumb (read: small amount of programmed reactions, and/or these reactions are not logicial) and you see it so hard that some players just don't play in game.

[Tynan]

I appreciate lively discussion but please keep it friendly everyone.

[Mikhail Reign]

You think that i'm stupid or from the future? Every AI in games works like this.

Mate I didn't make any assumptions - I just replied to what you wrote. Instead of putting the emphasis on me to interpret your comments in the way that you intended, maybe just make your comments less open to interpretation?

If player will can do something to AI but AI don't have programmed reaction, well.... AI will probably do nothing. The point is that in my example AI had programmed reactions and if not, then that wasn't so visible, in Rimworld sappers are just dumb (read: small amount of programmed reactions, and/or these reactions are not logicial) and you see it so hard that some players just don't play in game.

I don't get ya - the sapper pawns are no 'dumber' then the rest of them though. Every pawn will happily walk into death, sometimes, I swear, even seeking it out. Sappers will stand around while people around them die - so do siege pawns, so do preparing pawns. Whats the difference? I don't feel like the sapper AI is a stand out in any way - its on par with the rest of the AI - they spawn, move to their target and attack. Sure they can 'see' your base, but so can siege crews when they mortar it from the other side of the map.

I'm just not seeing the problem with the current sappers - they act and respond in ways and to situations the same as the rest of the pawns. They suck when your at max range, they B-line to a target they can see anywhere, they singly mindedly complete their objective. How is that any different to a raids AI?

[b0rsuk]

I'd like to see sappers more consistently avoid points where previous sapper raids failed, was killed. I'm not sure they do that. They attacked the same spot 3 times in a row. The next time they attacked through another wall, towards my hospital, and were more successful. They avoid turrets, maybe they should also avoid areas which used to have IED traps ? At least for a while.