Stealth gameplay

Non-technical talk about multiplayer and singleplayer gameplay and game design.
frag.machine
Posts: 2126
Joined: Sat Nov 25, 2006 1:49 pm

Post by frag.machine »

Sajt wrote:It's fun to theorize about more realistic and complex AI, but there's probably a reason many games don't include such things.

First of all, the game can become too difficult. Especially a game in which you are intended to fight dozens of enemies and come out alive. When each individual enemy becomes more competent and more clever, it becomes harder.

In the particular example I mentioned, that shouldn't be a problem, since you also gains lots of upgrades of strength, firepower, new attacks, etc. But what I really want to point out is the sloppy behavior: it's just common sense - so common sense in fact, that at first I was concerned about not being catch by a military carrying any weapons when disguised, until I realized that it was OK to the troopers if a civilian was carrying an assault rifle in New York streets. :D

Sajt wrote:Related to that, the game can become frustrating. When AI is sneaky, the player will often be hit without knowing where from, or attacked from behind.

That already happens a lot in Prototype, so no big deal.
I know FrikaC made a cgi-bin version of the quakec interpreter once and wrote part of his website in QuakeC :) (LordHavoc)
Urre
Posts: 1109
Joined: Fri Nov 05, 2004 2:36 am
Location: Moon
Contact:

Post by Urre »

All very valid points Sajt, but I've quite grown tired of these points as excuses to avoid the ideas altogether, instead of trying to solve them somehow.
Sajt wrote:It's fun to theorize about more realistic and complex AI, but there's probably a reason many games don't include such things.
This thread was mostly to theorize, it's fun :)
Sajt wrote:First of all, the game can become too difficult. Especially a game in which you are intended to fight dozens of enemies and come out alive. When each individual enemy becomes more competent and more clever, it becomes harder.

Related to that, the game can become frustrating. When AI is sneaky, the player will often be hit without knowing where from, or attacked from behind. Or an enemy marksman who picks a good hiding spot will easily get shots away before the player can locate him.

And finally, complicated AI can, to the player, be indistinguishable from random or even buggy behaviour. Such things as Arkage's example of minute changes in an enemy's "anger" level, possibly in conjunction with other dynamic attributes, will never be perceptible to the player. He will never be able to notice patterns, and will often become frustrated. The only remedies are 1) to dumb down the behaviours and make them obvious, for example to make the enemies say out loud "I'm really mad!"; or 2) explain it in the readme so the player has a starting point to understand what is random/buggy behaviour and what is the programmed AI behaviour (the latter being what he should pay attention to for learning patterns). It doesn't help that most game players are conditioned for bad AI, and will refuse to recognize advanced AI most times, passing it off as random or glitchy behaviour and complaining about it.

However, all these "compromises" don't sit well with me, and probably not you either. I still want to make sophisticated AI systems based on reality, and I would like nothing more than to include in the readme only the disclaimer "this game has advanced AI, forget what you have learned from other games and treat the AIs as you would real people". I always hate the heavy-handed approach of games like FEAR, which has somewhat "advanced" AI, but the enemies always tend to announce all their decisions out loud: "Flank him!" or "I see a flashlight!". But still, if you design a game from the start with an advanced level of AI in mind, and keep the areas roomy enough and the enemies not too numerous, you should be able to train the player to always look over his shoulder for sneaking enemies, and to forget the old method of "clear a room once, it's safe forever". And if the game is still too stressful to play, add subtle AI handicaps like louder footsteps and slower decision making (and of course a well-designed difficulty slider), but nothing so heavy-handed as fear.
If the game is marketed with this in mind, you should have no problem. Explain that the game really is about challenging, emergent experiences.

Some examples to improve the experience for the player, to reduce the amount of unfairness the player might experience:

At first the guards don't know the player has stolen a radio and is following their every movement (remember that guard actions are plotted out on the map automaticly as they announce their moves over the radio). The radio should even work as a GPS transmitter and reciever, so the player can see the guard positions on the map (the player shuts down his own transmitter obviously). This works as a great way to introduce the player to the level of intelligence these guards posess. Once they realise the player is following their every move, they'll shut off the GPS transmitters and they'll begin to avoid using the radio, perhaps only announcing where they've spotted the player, and only talking to guards in their vicinity. This will have the downside that they can't coordinate attacks as well, but the player will not have as much of an upper hand anymore, but now be much better prepared for the intelligence he is facing. At some point the guards could figure out a new way to communicate without the player being able to spy on them all the time, to further increase the difficulty as he progresses deeper into the complex.

When the player dies, there should be a killcam, which shows the actions and thoughts of the guard who killed him, perhaps as far as a minute back if the player is interested to know how he died. This is to get rid of the feeling of random unfairness, and this will also teach the player what to expect from the guards.
I was once a Quake modder
frag.machine
Posts: 2126
Joined: Sat Nov 25, 2006 1:49 pm

Post by frag.machine »

Urre wrote:At first the guards don't know the player has stolen a radio and is following their every movement (remember that guard actions are plotted out on the map automaticly as they announce their moves over the radio). The radio should even work as a GPS transmitter and reciever, so the player can see the guard positions on the map (the player shuts down his own transmitter obviously).
You can make the player being easier to track while using the radio, so the obvious advantage is partially compensated by the guards having a bigger chance to find him. In other words, the player is discouraged to keep the radio on all the time. Using it sparingly can provide enough information without revealing his position.

More random ideas in the theme:

- the player could set some distractions to lure the guards attention while trying to reach some objective. For example, dropping a coin in the guard path could (randomly) make him grab it and try to buy something in that vendor machine mentioned while back - creating a small time window to the player escape to another part of the map, or to use a computer terminal, etc. Another kind of distraction could be starting a small fire in a closet, or activating any nearby machinery - it would be cool to have more than one non-usual way to solve this kind of puzzle;

- the player could steal maintenance uniforms, badges and walk in disguise (at least for some time, or only in some areas), either pick pocking closets or simple knocking down some unaware npc;

- the player could be required to grab non-combat related gear from fallen enemies - watches, access cards, IR goggles, etc. and use it to solve puzzles like "there's a room monitored by infrared sensors and if you try to walk by that without a IR goggle you'll surely fire it and you're fried". OTOH, making the player's inventory limited prevents him to adopt the "grab everything you find" philosophy. In fact, enemies could drop totally useless stuff (at least in that context) among the required items, turning the solutions less obvious.
I know FrikaC made a cgi-bin version of the quakec interpreter once and wrote part of his website in QuakeC :) (LordHavoc)
Sajt
Posts: 1215
Joined: Sat Oct 16, 2004 3:39 am

Post by Sajt »

frag.machine wrote:the player could set some distractions to lure the guards attention while trying to reach some objective. For example, dropping a coin in the guard path could (randomly) make him grab it and try to buy something in that vendor machine mentioned while back - creating a small time window to the player escape to another part of the map, or to use a computer terminal, etc. Another kind of distraction could be starting a small fire in a closet, or activating any nearby machinery - it would be cool to have more than one non-usual way to solve this kind of puzzle
Of course, such incidents are extremely suspicious and once they happen two or three times (or even once), a general alert will go out, and the whole place will go on an increased level of lockdown.


Also, I'll explain why I mentioned earlier that my game idea was a "one-level" game. (By the way, it was also intended for coop.)

This game would not be a "campaign" of successive levels like most singleplayer/coop games. Instead, all levels (if there were more than one) would be totally standalone, as in a typical multiplayer game. The emphasis is on replayability, not disposable levels.

When I was a kid my brother would get the "PC Gamer" magazine and we'd play the demos that came on the disc. I could play the demo of, for example, Quake 2, or Jedi Knight, a thousand times over without getting bored, and without feeling an urge to see the rest of the game. When, years later, I finally got my hands on the full versions of these games, I didn't even like them. It was more interesting when the game could be played in one or two sessions, and then it ended, whereupon you could start over and try to play through it differently. In the full versions, there's no end in sight as you go through countless, increasingly uninteresting, run-of-the-mill disposable levels. (You could argue that the first levels in those games, especially Quake 2, were of higher quality than the later levels.)

Of course I had more patience when I was a kid, so such simple linear games as Quake 2 couldn't interest me through repeated play-overs now. But a game with highly emergent AI like this would extremely fun, especially in coop. If this were developed as a full game, you'd be able to pick from a number of standalone levels to play, just like in a typical round-based multiplayer game. The level would take an hour or two (or three) to beat, after which you can play it again or pick another level. Unlike campaign-based games, you don't have to commit to 10 or 20 hours of mostly mediocre material; you can quickly pick out your favourite level and only play that one.
F. A. Špork, an enlightened nobleman and a great patron of art, had a stately Baroque spa complex built on the banks of the River Labe.
scar3crow
InsideQC Staff
Posts: 1054
Joined: Tue Jan 18, 2005 8:54 pm
Location: Alabama

Post by scar3crow »

Sajt wrote:It's fun to theorize about more realistic and complex AI, but there's probably a reason many games don't include such things.
Deadlines, engineering budget, performance budget (consoles versus average PC versus high end PC), scope, lack of pre-implementation design covering these details, and/or focus on "story" so autonomous NPCs are a liability rather than a tool. Some are aspects of reality, some are reasonable choices, some are fashionable but bad choices, some are aspects of reality distinct to video games as a business rather than an art or passion. Also sometimes the coder just doesn't know how to do it, or the coder's he is working with setup a bad architecture for getting much of anything done.
Sajt wrote:I always hate the heavy-handed approach of games like FEAR, which has somewhat "advanced" AI, but the enemies always tend to announce all their decisions out loud: "Flank him!" or "I see a flashlight!".
The announcement of a flashlight can be explained as just surprise, but things like 'flank him' is more of excitement in the heat of battle. Granted, in FEAR most of the sound clips were used to usher the player into a scenario that was simpler for the AI. "I'm being flanked? Better get on the move" and the player might go into an easier area for the AI to work with. Another example of such a thing is in Call of Duty 4, which had hardcoded regions where if a player idled in them, the game would spawn a live grenade near them to get the player to run out of the region - and into line of sight for the AI.
Urre wrote:What I meant earlier though, was that the guards might not even have thought of you entering the ventilation system, or the floorwell in the above example, so they won't include those waypoints in their searches. Once they've spotted you to do such a thing, they will attempt to figure out where the vents can lead. I'm thinking that they'd have more trouble figuring out where you've gone if you enter a vent, but they'll atleast be on the lookout for vent entrances/exits, and guard those a bit more if possible.
Player modeling. The AI comes in with an expectation of player behavior, "player flexibility is 2" meaning that they will use hallways AND doors. But waypoints through vents, storm drains, etc are flagged with a 4. When the player is seen by the guards, they will evaluate the value of the nearest waypoints to the player, and the highest value will then be attached to the local guards model for the player. The guard if he survives the encounter will then use his radio to update the central commander's player model, which updates all guard's player model - to suspect storm drains and vents. This can then be used by the player in terms of spreading their focus thin, as they now have a wider range of waypoints/portals to cover, the player can count on classic hallways and doors being a little less observed.

For player expectation, each reported player encounter or dead guard found should raise expectation levels. If the player is reported at a location, they have a high expectation of the player being around those waypoints. There is a value tied to this which rots slowly, if they reach those waypoints before the value rots down, intelligent guards may then analyze nearby waypoints above the value of the player model - "Where could he have gone?" "There is a service vent access point in here..." which would result in a short term global issue to investigate waypoints that meet that value. If it leads to nothing, the global issue expires, and the player model remains the same other than elevating general expectation and estimated skill of the player.
frag.machine wrote:For example, dropping a coin in the guard path could (randomly) make him grab it and try to buy something in that vendor machine mentioned while back
This should only apply for guards with a very low expectation, or if they have a higher expectation, particularly low intelligence.
frag.machine wrote:You can make the player being easier to track while using the radio, so the obvious advantage is partially compensated by the guards having a bigger chance to find him. In other words, the player is discouraged to keep the radio on all the time. Using it sparingly can provide enough information without revealing his position.
Simple, radio static, clicking sounds, the sound of chatter. If its on, the player is transmitting these sounds to the local environment, which acts as a point of interest to any nearby guards.
...and all around me was the chaos of battle and the reek of running blood.... and for the first time in my life I knew true happiness.
Urre
Posts: 1109
Joined: Fri Nov 05, 2004 2:36 am
Location: Moon
Contact:

Post by Urre »

scar3crow wrote:
frag.machine wrote:You can make the player being easier to track while using the radio, so the obvious advantage is partially compensated by the guards having a bigger chance to find him. In other words, the player is discouraged to keep the radio on all the time. Using it sparingly can provide enough information without revealing his position.
Simple, radio static, clicking sounds, the sound of chatter. If its on, the player is transmitting these sounds to the local environment, which acts as a point of interest to any nearby guards.
That's certainly interesting!

Cool to see all the responses, love to read them :)
I was once a Quake modder
Team Xlink
Posts: 368
Joined: Thu Jun 25, 2009 4:45 am
Location: Michigan

Post by Team Xlink »

The player thinks he has advantage, until the player notice's there surprised assaults were totally expected, such the player starts to investigate, he opens the radio and notices it can record messages to so, he records a message and sets it to play and leaves it in a closest, while going to the main, communications center and hears his message being played, he then creates a new message of a totally false plan, and then makes his attempt at a goal, in an area where some of its guards were relocated to the new surprise and the other guards weren't paying as much attention to the guarding as they were there radios and monitors trying to see if it was a success. The player then plants a device in the main security system.

The guards having found that the player learned of there plan now think that the player doesn't have a high advantage, but that the player's intelligence is higher, so there alert raises.

All, the while the player has the real advantages, or so he thinks.

Untill next time....
Downsider
Posts: 621
Joined: Tue Sep 16, 2008 1:35 am

Post by Downsider »

Team Xlink wrote:The player thinks he has advantage, until the player notice's there surprised assaults were totally expected, such the player starts to investigate, he opens the radio and notices it can record messages to so, he records a message and sets it to play and leaves it in a closest, while going to the main, communications center and hears his message being played, he then creates a new message of a totally false plan, and then makes his attempt at a goal, in an area where some of its guards were relocated to the new surprise and the other guards weren't paying as much attention to the guarding as they were there radios and monitors trying to see if it was a success. The player then plants a device in the main security system.

The guards having found that the player learned of there plan now think that the player doesn't have a high advantage, but that the player's intelligence is higher, so there alert raises.

All, the while the player has the real advantages, or so he thinks.

Untill next time....
Pro game design.
Post Reply