🔒 Locked

Globally aware enemies in open world shooters

Started by Chocolate Milk Jan 11, 2009 at 3:45 PM 24 replies 3.9k views

Original Post

Jan 11, 2009 at 3:45 PM

Enemies in open world shooters I've played have only basic AI and a hate level. They might be camped or they might be on patrol. But they do not respond to global events. In other words, if they can't hear you or the enemy you are attacking, they don't exist. I firmly believe that a great innovation in open world shooters will involve the an enemy and how he responds to the player. GTA simulated a globally aware enemy with the police and imo was a critical component in design and development process. Lots of open world shooters ignored this element of GTA (!!!). Besides adding depth in the openness to the game, it can also keep action high. Take STALKER; They were forced to utilize enemy respawns so that returning to a place like a train station is as action packed as the first time you were there. Most people complained about the task of going through "recycled content", but the devs knew that it sure beat a cleared-out field devoid of action. A dynamic enemy could have resolved this problem because engaging the static enemy would only be the first part of the battle... dealing with the reinforcements (point of entry, level of response) would have brought about diversity in the conflict simply by adding these dynamic variables. The symptom is so obvious. When I'm attacking an outpost, the rest of the world does not exist. All that exists are the enemies at the camp. Compared to a linear game with heavily scripted scenarios this open world results in a much devolved and degraded player-vs-enemy relationship. Open world shooters commonly improve the conflicts between the player and enemy by utilizing mission-specific scripts. But this does nothing for battles taking place outside of mission content. This is so critical because a dynamic enemy adds to the sandbox appeal and I think an open world shooter is only as good as its sandbox elements. For example, in GTA, city driving was fun, interacting with the police response was incredible, and city exploration itself was pleasant. Without sandbox elements like these, your open world shooter is just a wanna be that should have been linear because creating a sandbox IS the reason for creating that open world. I've got some ideas about creating an open world shooter, and you better believe it revolves around an innovated approach to creating globally aware enemies. I think I might make it into a board game first! If you have something to say about sandbox or globally aware enemies, please share!

Stroppy Katamari

Jan 11, 2009 at 4:54 PM

Scripting could accomplish a lot of the same, without any new technological advancement. It's just that devs haven't been interested in doing that.

In general, very simple scripting (guards attempt to get on the radio and call reinforcements; reinforcements arrive from a particular direction and sweep the area) would make a huge difference. And you wouldn't even need to do that all over in the game to make it extremely memorable.

I was disappointed in the lack of that in FEAR, of all places. After all, it told you that your opponents are a *battalion* of *super soldiers*. Meh... more like mall security making the rounds hung over on a Saturday morning. They were nicely aggressive and reasonably quiet, but they just didn't act even slightly like soldiers, especially on defense. It would be interesting if the most dangerous guy among a squad of enemies was the one who has a radio and can call in two other squads, if he isn't killed fast enough. One alternative tactic would be to get one squad to call for help, to dodge the counterattack and wreck face in the other squads' positions which would then temporarily be lightly manned.

Codeka

Jan 11, 2009 at 5:29 PM

I think the problem has less to do with developers being "lazy" or whatever, and more with technological constraints that such a system would impose.

It's one thing to simulate a dozen soldiers right in front of you, it's a totally other thing to simulate ten thousand people in a giant world, even if the AI for each of them is fairly basic.

As Stroppy Katamari says, I think much of what you're after could be acomplished with some cleverly scripted events. You might even be able to fake it a bit for non-scripted situations. But I don't think you can simulate the entire world all at the same time.

Also, if enemies keep calling for reinforcements, every fight you got into would be extremely dangerous - because now you've not only got to worry about the guys in front of you, but also the reinforcements they'll call, and the reinforcements they'll call, and so on. How does it end? In GTA, for example, you can end the encounter by going to a spray-shop, but that's not going to work in all games...

Imagine if, in Crysis, they used a radio to notify the rest of the island of your presense. Realisticly, as soon as you got into your first fight, the entire island would've been alerted to your presence. You wouldn't be able to go anywhere without being shot on sight. And it wouldn't take long for them to start bringing in the helicopters and search-parties to come looking for you.

Chocolate Milk

Jan 11, 2009 at 5:38 PM

Quote:
Original post by Stroppy Katamari
Scripting could accomplish a lot of the same, without any new technological advancement. It's just that devs haven't been interested in doing that.

In general, very simple scripting (guards attempt to get on the radio and call reinforcements; reinforcements arrive from a particular direction and sweep the area) would make a huge difference. And you wouldn't even need to do that all over in the game to make it extremely memorable.

Script use varies in its importance, appeal, use, etc. For example, in the classic Goldeneye 64, they had alarms that once pressed would bring reinforcements. Memorable? yes, but not at all game defining.

And in COD4, enemies spawn endlessly until you advance to a certain position on the field. The entire game from start to finish is defined by these scripts, and this innovation garnered big time critic attention, both positive and negative.

Like I said in the first post, I bet scripting the police in GTA was a major part of development and was a crucial part of what made the game work. It must have also been a major design challenge. Why open world shooters haven't carried this design into their game is obvious, lack of creativity and experimentation, and the duplication is difficult.

btw I'm almost suggesting an enemy that operates like a computer opponent in an RTS. Aren't there games out there that use some RTS elements to create a FPS sandbox? Maybe I just haven't seen them yet, because it seems like an obvious approach to innovate in the genre of FPS sandbox.

Kaze

Jan 11, 2009 at 5:39 PM

Quote:
Original post by Codeka
Also, if enemies keep calling for reinforcements, every fight you got into would be extremely dangerous - because now you've not only got to worry about the guys in front of you, but also the reinforcements they'll call, and the reinforcements they'll call, and so on.

You could do a lot of hit, run and hide tactics but I agree it would be difficult to balance performance requirements with it being more believable than your standard infinitely or not infinite but still far too many enemies spawning nonsensically scenario.

Chocolate Milk

Jan 11, 2009 at 5:52 PM

ack... let me rephrase this post.
@Codeka You wouldn't have to have 10000 AI units on the map simultaneously and you wouldn't have to involve them all in every scenario. The script design would dictate the "who, how, and when" characteristics of a given reinforcement or response to a player action.

Stroppy Katamari

Jan 12, 2009 at 8:57 AM

What I'd really like would be that the enemy had an explicit system of command and communication, not an inexplicable "global awareness" like in GTA. Squad and platoon leaders you can kill, putting all under their command in chaos for a moment until someone takes charge again. Radios and phone lines you can destroy, so the enemy won't know in advance you are coming, or can't send timely reinforcements to a unit you are taking apart. Normal enemy grunts wouldn't need to be very dangerous, as the danger would come from the fact that the enemy would try to do all sorts of pincer movements and e.g. trap you between two whole infantry squads.

This would also allow distinguishing enemies not according to individual grunts' weaponry, accuracy, etc (they could even be the exact same), but how good their morale and leadership is.

When you faced a draftee unit, you'd crush them because they take so much time to come back online after a leader is killed, and a squad that lost communications would - at best - carry out their last mission, or even stay put or head back towards HQ.

In an elite unit, on the other hand, leadership would be transferred almost instantly after the leader is down. When appropriate, they would tolerate a horrible number of casualties without retreating, and squads that have lost comms and have no objectives would take their own initiative (e.g. hunting you down after they spot you).

Re: reinforcements. I'd actually prefer that instead of spawning, the (potential) reinforcements are already deployed in the area, and if someone calls for help on the radio, they head towards where the distress call came from. If this didn't happen, they'd just stay where they were. And even if they were triggered, you could just break contact and they wouldn't magically know exactly where you are.

Even if they are spawned, like arriving in a helicopter, the reinforcements hardly need to be infinite. On some level, the enemy might have one chopper with an infantry squad ready for immediate deployment (1 minute), and no more. On another, they'd have that much and additionally a platoon of mechanized infantry which comes from far enough that it takes ten minutes to reach the mission area. These would create natural, soft time limits.

Chocolate Milk

Jan 12, 2009 at 11:19 AM

For my vision it would be counterproductive to let the player interfere with their dispatch via communication blocking. Although it certainly sounds fun as a player to engage in communication blocking for self preservation, I'd require the player to experience the intensity of dealing with the adds directly, either through evasion or combat, rather than reward him for preemptively neutralizing it. I wouldn't even want it as an optional playstyle because I can't let one playstyle turn my main focus of the game into an experience of fault/shame/punishment. Even so, the player who wishes to avoid combat would still find evasion to be as important of a tactic as combat.

deathtojohnny

Jan 12, 2009 at 2:43 PM

There are so many faults in sandbox worlds,

1. One thing that aggravates me with these sandbox games is that the interior of buildings are usually so simplistic in design, that it feels like you're playing in a cardboard box-- as opposed to a game that would make the interior design it's priority.

2. Another sandbox pet peeve is that there are never different points of entry. You always come in through the same entrance, to fight the same barrage of enemies every single time. There should always be different ways to enter a stage so that you can set up a different plans to break in. How can this really feel like a full world, if you don't have the option to break into every single available window/door?

3. In association to the previous point, I would like it if the enemies responded a little more realistically. If there was a "situation" in a building, security will all run to the commotion and investigate-- of course with the exception of the persons watching the monitors.

Now, mixing all three points together-- you will now have a game where you could cause a commotion on one side of the building, and enter another way.

Pretty much, I think any one that ever creates an espionage game, should work as a security guard first to see how it works.

Pseudokai

Jan 12, 2009 at 6:24 PM

I like the idea of having reinforcements respond to the situations specifically. If you've killed half of an enemy of squad in an average amount of time they could call in a second one to turn the tide. If you quickly devastate a squad they would report a major threat and send two or three times the response. An easy way to stop continuous reinforcements would be to set a number of soldiers that can die before the battle is considered a loss for them. I would have liked to see something like this in Far Cry 2. The enemies could be seriously wounded, but that only meant that they were going to die a little bit later. Having an enemy survive and comment appropriately when you run into them(or someone with the same model) again would be pretty cool.

Stroppy Katamari

Jan 13, 2009 at 6:08 AM

Quote:
Original post by Chocolate Milk
For my vision it would be counterproductive to let the player interfere with their dispatch via communication blocking. Although it certainly sounds fun as a player to engage in communication blocking for self preservation, I'd require the player to experience the intensity of dealing with the adds directly, either through evasion or combat, rather than reward him for preemptively neutralizing it. I wouldn't even want it as an optional playstyle because I can't let one playstyle turn my main focus of the game into an experience of fault/shame/punishment. Even so, the player who wishes to avoid combat would still find evasion to be as important of a tactic as combat.

I understand strategic thinking might not fit in your game, but I'm not sure why you think this kind of element would constitute fault/shame/punishment. Those are something I'd normally associate with excessive difficulty and/or fundamentally unfair things (like police in GTA spawning in physically impossible ways behind your back, but not in front of you).

When the enemy has weak spots like communication, their other difficulty could be made correspondingly higher. So depending on the difficulty tuning, you could have roughly the same amount of action, but more tension through another (orthogonal) type of challenge. If the player can just barely defeat five enemies straight up, it's hardly less intense when there are 20 enemies in total, and the player must make the right calls to avoid fighting 10 of them, to evade/delay/distract five, and beat the remaining five.

The strategic part need not always be calm pre-fight planning. Say you run into a squad of enemies you weren't expecting. Both sides open fire and take cover. Now do you take the risk of immediately rushing in and try to wipe them out before they can call for help / coordinate an attack / ... , or fight more carefully and whittle them down while accepting that more enemy might come, or do you try to break contact? And if you want to break contact, should you first try to take out their communications so they can't tell other enemies which direction you were last seen going?

Chocolate Milk

Jan 14, 2009 at 10:59 AM

Quote:
Original post by Stroppy Katamari
Quote:
Original post by Chocolate Milk
For my vision it would be counterproductive to let the player interfere with their dispatch via communication blocking. Although it certainly sounds fun as a player to engage in communication blocking for self preservation, I'd require the player to experience the intensity of dealing with the adds directly, either through evasion or combat, rather than reward him for preemptively neutralizing it. I wouldn't even want it as an optional playstyle because I can't let one playstyle turn my main focus of the game into an experience of fault/shame/punishment. Even so, the player who wishes to avoid combat would still find evasion to be as important of a tactic as combat.
I understand strategic thinking might not fit in your game, but I'm not sure why you think this kind of element would constitute fault/shame/punishment. Those are something I'd normally associate with excessive difficulty and/or fundamentally unfair things (like police in GTA spawning in physically impossible ways behind your back, but not in front of you).

When the enemy has weak spots like communication, their other difficulty could be made correspondingly higher. So depending on the difficulty tuning, you could have roughly the same amount of action, but more tension through another (orthogonal) type of challenge. If the player can just barely defeat five enemies straight up, it's hardly less intense when there are 20 enemies in total, and the player must make the right calls to avoid fighting 10 of them, to evade/delay/distract five, and beat the remaining five.

The strategic part need not always be calm pre-fight planning. Say you run into a squad of enemies you weren't expecting. Both sides open fire and take cover. Now do you take the risk of immediately rushing in and try to wipe them out before they can call for help / coordinate an attack / ... , or fight more carefully and whittle them down while accepting that more enemy might come, or do you try to break contact? And if you want to break contact, should you first try to take out their communications so they can't tell other enemies which direction you were last seen going?

There'd also be security cameras to shoot, right? If I put stuff like this into the picture I'd probably have a simple 1-click item that took care of it. Like, an EMP pulse with a 30 meter radius that lasts 20 seconds. Reinforcements would still come, just not as many. That would be the shooter's approach. Another approach would be to have technology that locates all coms devices in view so the player can neutralize them or their user. That would be the silent-tactical approach and would focus only on sequential elimination of the prime targets rather than spending time gathering intel on their location. I say silent because it would require silencers so the dude indoors with a headset doesn't immediately get on the radio... he'd just come outside to investigate. If I really wanted to get into this diversion of combat I'd throw in a spy approach where you rely on stealth and daggers. This would lead to the creation of class mechanics, and an enemy that is fun and balanced for all class types. A bit of a detraction from my main focus, but sometihng I wouldn't turn down if it could be done right.

deathtojohnny

Jan 15, 2009 at 9:42 AM

Pseudokai: Reinforcements also give you a chance to act quickly, hide, and find a strategy to wipe them out. Hopefully, of course, reinforcements will eventually stop coming in from the same door and opt for their own strategy, which adds more of a thrill to the game.

thk123

Jan 16, 2009 at 5:08 PM

Quote:
Original post by Stroppy Katamari
What I'd really like would be that the enemy had an explicit system of command and communication, not an inexplicable "global awareness" like in GTA. Squad and platoon leaders you can kill, putting all under their command in chaos for a moment until someone takes charge again. Radios and phone lines you can destroy, so the enemy won't know in advance you are coming, or can't send timely reinforcements to a unit you are taking apart. Normal enemy grunts wouldn't need to be very dangerous, as the danger would come from the fact that the enemy would try to do all sorts of pincer movements and e.g. trap you between two whole infantry squads...

As cool as that does all sound, actually balancing would be very hard. How many games have you played where the designer has put something cool in, but it is much simpler just to shoot the enemy than try and do what you are meant to. On the other hand, if it was impossible to do it without knocking out the phone lines etc. it would become a necessary task that you just had to do first. Even the perfect balance has its complications - there is no fun if the game explicitly says, "Why not take out their phones" but it probably won't occur to most players to do it without such prompt.

JasRonq

Jan 16, 2009 at 6:04 PM

Keep in mind that it is entirely reasonable that the backup would know where to go if you shoot the radio or the cameras. You can monitor the signal from both such that you know when the signal stops and something is up. You can also have tracking tags on your men, so if the radio gets shot, they send investigatory backup to the local of the team that had the radio, via the tracking tags on the team. Heck you could even have life sign monitors on the team such that the backup group knows when men start dropping.

EMP could take out all of that at once, such that the tracking tags can't relay the location, but they will still know something is very ip because every one's signal just cut out.

(BTW, EMP lasts far longer than 20 seconds and is often permanent, repair/replace worthy damage.)

Stroppy Katamari

Jan 16, 2009 at 10:48 PM

Quote:
Original post by thk123
Quote:
Original post by Stroppy Katamari
What I'd really like would be that the enemy had an explicit system of command and communication, not an inexplicable "global awareness" like in GTA. Squad and platoon leaders you can kill, putting all under their command in chaos for a moment until someone takes charge again. Radios and phone lines you can destroy, so the enemy won't know in advance you are coming, or can't send timely reinforcements to a unit you are taking apart. Normal enemy grunts wouldn't need to be very dangerous, as the danger would come from the fact that the enemy would try to do all sorts of pincer movements and e.g. trap you between two whole infantry squads...

As cool as that does all sound, actually balancing would be very hard. How many games have you played where the designer has put something cool in, but it is much simpler just to shoot the enemy than try and do what you are meant to. On the other hand, if it was impossible to do it without knocking out the phone lines etc. it would become a necessary task that you just had to do first.

Only when there was only one right way. In which case, why have enemy communication and command mechanics at all? Level design has to set up situations where the "best choice" is not obvious. See my earlier example (after running unexpectedly into enemy) of such a tactical choice.

Ideally, different players would be effective through different tactics at least some of the time. After all, the players are not equally good at sniping, assault, disengaging, predicting enemy movements, sneaking, quick thinking, and so on.

If the game allows you to attempt to disengage and escape from an enemy that is too strong for you to beat, for whatever reason (maybe you made a bad tactical decision) then it's possible to challenge the player a bit more than otherwise.

Quote:
Even the perfect balance has its complications - there is no fun if the game explicitly says, "Why not take out their phones" but it probably won't occur to most players to do it without such prompt.

After the game has presented the mechanics to the player in a tutorial or (equivalent) integral part of the game, why would they not think of it? That's like asserting Crysis players would not remember to use the cloak without prompting.

Iron Chef Carnage

Jan 17, 2009 at 3:05 AM

It would be far from feasible to try to accurately model the entire enemy presence in a given region. As has been said, even ignoring technical limitations you'd have a nightmare of a time balancing the responses to that they seem reasonable and strategically sound. I spent three drunken hours one night camping the bottom of a rappelling line in Rainbow Six: Vegas 2, shooting a gun in the air then picking off the terrorists as their "mount rope" animation cycled.

It would be better to have a sort of spawn policy, dependent on the locality. Response time and magnitude would depend on what happened and where. If an enemy fires a flare, a chopper will arrive shortly with a few riflemen, a medic, a mechanic and a communications officer on it, in order to evaluate the situation and either deal with it or radio for a more appropriate response. If an enemy gets to a radio and reports the situation accurately, then a more focussed and effective response will be forthcoming. If you blow up a checkpoint that can be seen with binoculars from a nearby command post, you can expect a heavy response, whereas a similar attack three miles past the far ridge won't be detected until the plume of smoke gets noticed.

So have enemies spawn and add as appropriate, and model only the part of the world that the player can actively see. If you set up a timed demolition charge, then sneak over to the place where the "reinforcements" would likely come from, then you can watch some of your enemies depart to investigate, then deal with the rest. If you are at ground zero when it blows, then you get an add that presumably came from that location.

Unless you're going with the "finite enemy presence" angle, then there's no need to model both sides of the operation. It can simply be assumed that the population of the area has been reinforced by the time you get there.

Chocolate Milk

Jan 17, 2009 at 12:24 PM

Right. Elegance in the simulation is critical.

The best way to think of it is not in the scope of unit AI, but rather, the army AI. It's a virtual entity that defines the behavior and simulation, its not troops.

Cpt Mothballs

Jan 18, 2009 at 4:04 PM

Awareness isn't really a good thing.

I remember Oblivion's Guards.

They were this omnipotent force that made playing the game as a bad guy ZERO FUN WHATSOEVER.

I'd also prefer wiping out a bunch of people's underground lair and then being able to use it as my own hideout instead of having to fight them off every single time I go there.
That'd be lame.

thk123

Jan 20, 2009 at 4:45 PM

Quote:
Original post by Stroppy Katamari
Quote:
Original post by thk123
Quote:
Original post by Stroppy Katamari
What I'd really like would be that the enemy had an explicit system of command and communication, not an inexplicable "global awareness" like in GTA. Squad and platoon leaders you can kill, putting all under their command in chaos for a moment until someone takes charge again. Radios and phone lines you can destroy, so the enemy won't know in advance you are coming, or can't send timely reinforcements to a unit you are taking apart. Normal enemy grunts wouldn't need to be very dangerous, as the danger would come from the fact that the enemy would try to do all sorts of pincer movements and e.g. trap you between two whole infantry squads...

As cool as that does all sound, actually balancing would be very hard. How many games have you played where the designer has put something cool in, but it is much simpler just to shoot the enemy than try and do what you are meant to. On the other hand, if it was impossible to do it without knocking out the phone lines etc. it would become a necessary task that you just had to do first.
Only when there was only one right way. In which case, why have enemy communication and command mechanics at all? Level design has to set up situations where the "best choice" is not obvious. See my earlier example (after running unexpectedly into enemy) of such a tactical choice.

Ideally, different players would be effective through different tactics at least some of the time. After all, the players are not equally good at sniping, assault, disengaging, predicting enemy movements, sneaking, quick thinking, and so on.

However, for that to work, you must first create a game in which the player expects to be able do to anything that they can think of. If a game does not explicitly state that X is possible, most people won't run up to the telephone box and hitting it on the off chance that this game allows you to do that.

Topic Locked

This topic has been locked by a moderator. New replies are not allowed.