Page MenuHomeFeedback Tracker

Room Clearing with the AI
Reviewed, WishlistPublic

Description

There are many ways to skin a cat, but not too many ways work well with room clearing. There are very simple methods used today that can be replicated with the AI since they are very methodical and repetitive. AI in my opinion is mathematical, room clearing is taking numbers and calculating on the fly to produce the appropriate action for the given situation. The player controlled AI should be able to clear a building as good as any human team. Being the last one in the stack, the player won't interfere with his men and can be sure that they will not make the wrong choices on direction, targets, and be slow to get into the room.

Alexander Whatmore
United States Army Infantry
B Co, 2-8 Cav, 1 Bct, 1 CD
Fort Hood, Texas, 76544 {F19241} {F19242} {F19243} {F19244} {F19245} {F19246} {F19247} {F19248}

Details

Legacy ID
4195853366
Severity
Feature
Resolution
Open
Reproducibility
Have Not Tried
Category
Feature Request

Event Timeline

Stalker1 edited Steps To Reproduce. (Show Details)May 2 2013, 7:45 PM
Stalker1 edited Additional Information. (Show Details)
Stalker1 set Category to Feature Request.
Stalker1 set Reproducibility to Have Not Tried.
Stalker1 set Severity to Feature.
Stalker1 set Resolution to Open.
Stalker1 set Legacy ID to 4195853366.May 7 2016, 1:57 PM

At least the AI should be able to clear a single room, designated by the player. Point a mouse at the door or doorway, select the teammembers you want to make entry and use option "clear room" - similar to how you order them to get into a car, for example. Maybe there should be a limitation, that the "clear room" option is available only, when two or more AIs are selected? AI would clear the building room by room, as commanded by the player following his squad.

Adding clearing capability to AI-only squads would be even more interesting, but it would require to have the squad leader AI to be aware of the whole building and it's architecture, so it could effectively command the other AIs.

Related to:
http://feedback.arma3.com/view.php?id=6514 Command AI to secure house
http://feedback.arma3.com/view.php?id=4553 AI needs building usage routines

I typed in the search and they didn't come up. I actually know what I'm talking about and get paid to do it. You need at least one AI (assuming player is behind him). Clearing a room is very simple but it's a perishable skill

I like this...But....Real life Infantrymen, change their stance and tactical movement and tactics on the fly.

So in open terrain, Infantrymen naturally spread out and cover one another in bounding overwatch or just moving in formation, in Urban terrain they tend to stack up and mutually support very close together.

We would need a more natural AI that changes seamlessly and accurately between Urban terrain, and open terrain.

Would the AI flip to Urban in a Farm Complex? Would they flip back to an Open Stance in a Urban Park or open area? It's not insurmountable, but it would take considerable coding and resources.

OIBUA/ FIBUA is nightmare in real life, and requires a high level of professional skill. I think ARMA handles open battle very, very well. It falls down massively in Urban terrain, I'm not sure we can solve it due to the path finding and the way Teams of AI work together.

It is a very worth goal and I would love to see it, I am voting this up but I am unsure it can be achieved in a realistic way.

You have to understand that this is a simulator. If you take "statistics" and put them into a game you have a good foundation, but realism comes from experience in the field and how equipment actually handles under the stresses of combat and fatigue. Even in real life, humans have flaws. The job of the team leader is to continuously make sure his men are moving in the proper formation and not bunching up. Soldiers don't automatically change their distances and stances. Everything is based around the team leader. AI doesn't have to change, their relation to their team leader and strictly following orders is what makes everyone react appropriately

Yeah I totally understand that, but are you talking about Human lead AI Fireteams?

Because an AI fireteam Squad Leader will have to behave like an Human/Real Squad Leader, in order for this to look correct.

Otherwise we are in the realm of Human Led squads look ok/good (depending on the skill of the human SL) whereas the AI SL squads look appalling. We cannot even get natural sounding Communications "ONE! go here, TWO! do this...." etc etc

This is currently what we have..I would suggest watching SHACK TACTICAL VIDEO's or other ARMA Co-op Clans on Youtube to see more natural Human Playing, AI does not or cannot simulate this.

The problem for me is that we want ARMA to be natural as possible, I want your suggestion to be integrated 100% hence why I voted it up. But we need to be aware that in the Resource vs Benefit vs Size of Task is it likely to be implemented by BIS who have Limited Resources? I would think it unlikely.

I think the MOD community will do this as a Labour of love.

The mod community can only do it, if its doable with current scripting commands and FSM knowledge. Also performance impact by scripting, rather than beeing hard-coded and BI-Dev optimized, has to be taken into account.

IMHO: Don't rely on the community every time to fix lacking features and bugs.

Well if BIS have to re-do the AI don't expect ARMA 3 to be released this year. With new AI, Alpha testing would have to be redone and then Beta.

AI is pretty Integral, unless this games goes Multiplayer only (which it won't).

We are going to have to accept a degree of compromise here.

Actually, the AI of the games and simulators is the most forgotten feature. Graphics, physics, sound, all this has improved since the beginning of the games, but the AI ​​has stalled, not advanced almost nothing. If we start seeing a jump in the simulation, we have to understand that we must first put the AI ​​to date. We have a clumsy AI, based on preset movements, which is not able to assess the situation one bit.
Since I play as much as possible OFP/ArmA1/ArmA2 avoid having to use the AI ​​in missions, because it is very frustrating to see how something as simple as follow, are unable to do so. Or view as a single enemy soldier, if he sees you, attacks you, but you bring 4 tanks and 50 soldiers.
Not that can not be improved, I have seen other productes, with reactions more logical AI.
Therefore, if BIS wants the arma3 a step forward, you will work on a far more advanced AI, dynamic, using the ground, he react to their environment (fear of enemy fire, evaluation of the situation. ..)

Bohemia added a subscriber: Bohemia.May 7 2016, 1:57 PM

Good ticket and feature with breaching and flashbangs etc.

@Devs: Look at Rainbow Six 3, SWAT 4, teammates have a perfect real life tactics in cqb and closed space, good example;
Read a room clearing tactical documentation, with step-by-step reproductions
Current AI looks very stupid as in DOS Games in '95-98 years :)

Lot of skilled players always say FUUUUU omg, when looking on current AI, and they never sayd WOW COOL! in cqb and buildings cleaning in team,
so consider it and fix it please.
Because the devil in the details! ;)

Games like Rainbow Six, SWAT, Full Spectrum Warrior, as well as sci-fi Republic Commando, take place on small maps with closed structure, where you can afford to place smart waypoints manually and design nice looking and fluid resolutions to a limited number of corridors, doorways, corners etc. Not a chance to do it for a multi square km Arma island. How many of these games have maps comparable to Zargabad or Elektrozavodsk alone, in terms of size and complexity? Add to that mostly preplanned enemy positions and compare with vast tactical possibilities and unpredictability of Arma.

I have to defend Arma's AI, but only partially. I've seen the moments of great glory and tactical excellence on AI's part, but also unspoken stupidity and epic fails. For me, the AI is a great work, that has been started and never finished, it's half baked now. It has potential, but it's hidden. I've managed to "beat" lots of Arma2 coop missions with the AI companions, but it required me to adjust to their "style" and simply not attempt certain moves, which would be the best course of action with human squad.

The greatest problem with any AI, is to make it aware of its surroundings. Having a mechanical "mice" to exit labirynth is easy, the same for autopilot following preprogrammed waypoints. But having an automatic shopping cart, that could go to a shop two blocks away and buy you a beer, is another story.

The problem lies in giving the AI the ability to process data coming from sensors and to make rational and well thought decisions based on this data. An example outside of Arma, our current AI can recognize a face on the picture and can point to another photos of the same person in the database, but can't decide on its own, if it is male or female.

The fundamental problem with Arma AI is based on its two characteristics:

  1. While AI has some pretty good individual skills, it can't work in a team and act as a team. In combat situation, an AI firetam instantly divides into a bunch of individuals. Sure, they loosely stick together, but that's not exactly a team work. A team deathmatch more likely.

Suggestion: Make the AI to pay more attention to what's going on with its teammates and to stick with them more, instead of free roaming in their general vincinity. Increase the role of Squad Leader and Team Leaders and introduce chain of command (SL -> TL -> Team).

  1. AI seems to have some kind of awareness regarding its surroundings. It can take cover, conceal itself, drive along a road, find a flat spot to land. It can avoid the tailrotor area while getting into a helicopter, it knows how to lean from behind a corner, it can flank. AI can even take shortcuts through a building or enter a building in which I'm hiding - it happened to me on numerous occasions already (I thought I was safe there, because of Alpha problems with building routines - a bullet to the back proved me wrong).

The problem starts with using of this knowledge, especially for some procedural tasks: proper road/street crossing with overwatch, room entry and clearing, stacking up against a wall with proper sectors division and coverage, all round defense while stationary, walking with tail-end-charlie, procedure to break contact with teams, the whole fundamental concept of fire element and move element. These items are especially important and beneficial with the proliferation of missions with dynamic and random enemy spawning.

Suggestion: For standard situations, add standardized solutions. In a room entry example, ordering such maneuver should put the AI into a preprogrammed subroutine, which tells it how to move into a room and how to divide sectors of responsibility. After the room has been entered and enemy neutralized or found to be not present, the AI would return to its normal movement routine and be ready to follow the leader or execute another order of any kind.

I know it its easier to talk, than to make it happen. But most of the building blocks for advanced AI routines are already there. It's just they either do not work flawlessly or work separately of themselves, instead of together.

Example 1: stationary AI seems to divide area into observation sectors, but doesn't take into account full 360 degrees coverage.

Example 2: AI does things like on the attached screenshot, but the problem is, it doesn't produce expected results; all of them died there, even though they moved tactically and knew about enemy's whereabouts.

Example 3: AI knows how to flank and retreat, yet it's still possible to kill them all in an extended firefight across a valley; they just lay down there and shoot back, regardless of mounting own casulaties due to accurate enemy marksmen fire.

Example 4: AI knows how to drive along a road by itself, to a point where I have trouble keeping up, but it's greatly incapable of driving in a column, along the same road.

Example 5: AI knows, that leaning out from behind a corner is a smart move, but sometimes it leans the wrong way and all the bullets end up in the wall.

Example 6: AI can shoot quickly and accurately, yet for some reason can't use their weapons their full potential. Arma2 KSVK sniper, set to max skill, needed 3 bullets to finally hit me at 700m. I was standing stationary, wasn't shooting back, it was day, the sun wasn't blinding him and there was no wind.

Example 7: It's possible to have fire and move elements, with the use of "colourful" teams, but because of AI's inability to work as a team, often it's more reasonable to just go there with them all and do it the guns blazing way.

Example 8: AI uses tactical movement and sometimes it works great. But on other occasions it's like this: "c'mon, run across that street, into enemy fire and then I'll run there too, to see if they can shot me as well".

You guys are talking about the AI like it should be an army of one.

The only problem with the AI in Arma 3 is that they have all of these working parts and all of these behavior scripts that they don't do exactly as they are told.

Hence, if you are moving long distances in a wedge, they fall behind or they get stuck on trees or rocks. If I get onto an objective and call them to get on line, they are slow, and act retarded. When I tell them t bound forward while my team gains fire superiority, they stop while standing up, to engage the target.

THE AI DOES NOT HAVE TO UNDERSTAND THE SITUATION. They should follow orders to the teeth because I know what I'm doing and I know what's going on.

In order for room clearing to be effective and I can rest safe assured that they will do their job, is if they follow their guidelines to the teeth. When they are in line in a stack ready to enter a room, the scripts should select the first man to go right, the second man to go left, third man to go right, fourth man to go left. Secondly, it should outline the edges of the room and generate paths to not stop until they have reached their final destination. Thirdly, they should identify targets and shoot on the move focusing their barrels in the proper direction. First man goes middle of the room to the right corner, second man middle to left corner, third man middle high, second man middle and then turn to the door.

AI DOES NOT HAVE TO BE SUPER SMART TO FOLLOW SIMPLE, PREDETERMINED, PREDICTABLE PATHS.

WHEN AI IS FIRING AT THE ENEMY, HE IS PUTTING ROUNDS DOWN RANGE LIKE A FUCKING PIG. WHEN HE IS MOVING OR REMAINING IN FORMATION (moving in a wedge long distance, getting on line under enemy contact or bounding) HE IS HAULING ASS

Great ticket, upvoted. Especially since A3 plans on having big enterable buildings!

I agree with Stalker 1: when assigned an order, all smaller rank unit should set that order as a priority over even their life preservation instincts. Would save a lot of fast-forwarding seeing your team twitch and krump while taking cover from one grain of sand to the other...through an area they've just cleared.

@Stalker 1

"You guys are talking about the AI like it should be an army of one"

No, we don't want an Army of One AI, we are trying to explain why something you want implemented to look "Like it does in the Real World" is much, much harder to implement.

You are Blue Skies thinking, and using justifications but you need to fundamentally understand how the AI works

"THE AI DOES NOT HAVE TO UNDERSTAND THE SITUATION"

Yes it does, the current system is that it understands Orders and carries them out even in stupid situations or even if it results in their certain death.

"AI DOES NOT HAVE TO BE SUPER SMART TO FOLLOW SIMPLE, PREDETERMINED, PREDICTABLE PATHS."

This is actually what ARMA does right now, and you want it changed and for good reason. We need to break away from "deterministic approach, where a Human Programs it all" to "The AI makes educated decisions based on rules and scripts which more closely mimic Human interactions"

Here is the piece of the puzzle you are missing - CONTEXT.

If you told a Soldier to walk 10 metres right and he did, is he Obeying Orders? Yes

If you told a Soldier to Walk 10 metres right and he went over a cliff to his Death, is he Obeying Orders? Yes

In both these examples rigid adherence to orders without regard or thought is not what the Modern Military wants. It want's Soldiers who can think for themselves and problem solve, so that Orders are more Directive. It gives the Squad Leader, Fireteam leader, freedom of action to make the best decision whilst still trying to, or achieving the aim of the Directive.

CONTEXT is everything, or our AI's will be on Rails and perform predictable actions and this will itself look un-natural.

CONTEXT is where the AI falls down, it obeys predetermined rules, like "If underfire, Take Cover" they don't consider "Is this Cover any good? Am I being Flanked, Do I need better Cover? Do I have decent Arcs of fire?". It's also no able to make Judgements like "My Firing arcs are limited, but the Cover is much more substantial, therefore I will stay here because I am protected even if it affords me little in the way of effective firing arcs."

Humans naturally contextualise everything based on our understanding of the World and our sensory imput, and in some cases prediction. e.g Thought processes

"I need to walk 10metres right, I will go over that cliff and fall. I predict I will die if I do that, so I will not do that. If I move 5 metres, I will stop, I could stop here, or maybe work my way down to the base of the cliff...I'm not sure what to do, should I ask the SL? No think.......Okay at the base of the cliff I cannot see very well. Okay I will stop at the top of the Cliff, and then speak to the SL and tell him I cannot go any further right".

CONTEXT is everything, you need to move from "This is how it is! Do this BIS" to more of a "How can we do this be done with the resources we've got? Then come up with a thoughout solution and present this to BIS"

A good example of a well thoughout post is this one...

http://feedback.arma3.com/view.php?id=3505

AI needs to be smarter or your really good suggestion at the top of this page, will never be implemented effectively.

but the AI might become self-aware!!

But, you are bringing up the individual soldier. You are saying that every soldier can assess a situation as if he was in charge.

Yes, of course you wouldn't walk off a cliff, and yes you would seek cover. The AI already do that. As a Team Leader, your job is to direct your men and their job is to follow to the teeth and execute with precision. They only need to think of one thing at a time unless it comes to something complicated such as room clearing where they need to fire while moving to their point of domination.

This is how we do it in real life. Barney style, because everyone who first joins the army has no idea what they are doing. It isn't to insult your inteligence, but if you aren't in the military with a job specified for engaging the enemy then you probably don't practice it, and therefore, do no know it.
Priorities Under Fire:

    Battle Drill 2 React to Contact:
               Tasks, Conditions, and Standards: Your team is moving in a wedge               
               and you are to stay alive.. too easy.

               Bang Bang Bang (enemy fire)
            1. First thing you do is return fire while at the same time 
               dropping to the prone
            2. Call out distance, direction, description (3 D's)

               At this time it's called the mad minute. You are trying to gain 
               fire superiority against the enemy by firing in cyclic rate of 
               fire (750-900 rounds per minute, no pauses, barrel change every 1 
               minute) for 10 seconds.
   
            3. Rate fire change called by the FIRE TEAM LEADER to rapid (45-
               60 rmp, 5-6 second breaks, barrel change every 2 minutes)
            4. TEAM LEADER CALLS everyone on line. Since your SAW/Light
               Machinegun is the highest casualty producing weapon system, he 
               will move last. Rifleman moves first, then grenadier, then the 
               AR. [AI should understand that they should still use cover]
            5. Bound to the objective. The TEAM LEADER orders the movements 
               of his team using 3-5 second rushes. [The AI must follow MOVE              
               without stopping in the open. They do not have to have a new 
               script telling them that their team leader and AR gunner are 
               suppressing.]
            6. Once the team is close to objective, the TEAM LEADER will call
               his team to get on line. Then will move through the objective
               and kill any enemy that they missed. 
            7. LOA (limit of advance). Once team is 35m from the enemy position,
               they will stop. The TEAM LEADER assigns sectors of fire and 
               submits a LACE report to higher up (Liquid, Ammo, Casualty, 
               Equipment)

For game purposes, I wouldn't go into any more depth than that. The AI is controlled by the team leader. Now, just because you don't want to take control of your team, doesn't mean they should be extremely smart because you don't know how to use them and you want them to make all of your decisions for you. In REAL LIFE (I can say that cause I do this for a living) soldiers do not think, they react. TEAM LEADERS think. If a soldier has a suggestion, they bring it up, but AI prbly won't be able to tell me something I don't already know.

Terms to consider for AI that aren't already in place. When initial contact is recieved, AI automatically switches to Cyclic rate of fire. When team moves into building if they make a new room clearing feature, they move to their point of domination (A pre determined path generated at the point of the order given to clear the room) and FIRE WHILE ON THE MOVE.

Other than that, I don't see how creating a smarter AI would benefit that specific situation. Now, the enemy AI should be smarter on the team leader level. But individually, no. The AI just needs to be changed in accordance with what is supposed to happen in REAL LIFE and what their goals should be (fire superiority and following movement orders/staying in formation).

I am not trying to insult you or tell you that you are wrong, I'm just telling you how we do it in real life, and I would like some of that to be revealed in this realistic military simulator so it can be used as a training aid by people in the army

I agree. I would also like to have it working in Arma that way. But there are certain differences between Arma and real world.

"The AI must follow MOVE without stopping in the open."

Fulfilling that requirement means having a smarter AI on the individual level. Smarter doesn't mean having a self-aware intelligence here. Smarter means "being able to act in the most effective and reasonable manner possible, without using cheats (like see through walls, for example)".

You don't have to tell your soldiers exactly, what each of them needs to do, you just issue "bound to the objective" or "move out" order and they act. But they know, HOW to act. And this is based on their training (for AI it is an equivalent of predetermined behaviour scripts with procedures), as well as their own knowledge of surroundings, the situation and their common sense (this is, where the AI is lacking and should be "smarter").

Soldiers won't stand up in a firefight, which is what the AI does. But they know, that sometimes you have to simply RUN, which is somehow hard to understand by the AI. Or a dilemma: to fire or not to fire? This depends on circumstances, which humans understand well, but with the AI, it's best to have them on "hold fire" for the most of the time.

Human has procedures, common sense and intuition - AI has only a set of strict procedures. If something isn't there, AI won't make it up by itself. And this requires lots of thought and writing of preplanned procedures, which will take into account multiple possible variables, that can be encountered on the battlefield.

Couldn't have been said better armapirx

Stalker1, as someone who has spent a Career in Uniform and now retired you don't need to tell me how it's done. I now work in Security Software Development and believe me I understand both sides of the coin.

Frankly....I've tried to explain why this is important.

Individuals matter in ARMA because the Team does not operate as One Cohesive whole (unlike a military unit which does and occassionally does not if a squad member sees something developing a Superior cannot have prior knowledge of). I have talked exclusively about Context. Individual Squad members are tied to a Squad Leader but their AI can act independently e.g some move, some fire, some take cover.

If you programme the AI to do as you say above, in Battle Drills they will look and react correctly 75% of the time, 25% of the they will react completely inappropriately. This is the Problem of Context, they aren't able to dynamically react realistically to the near infinite combinations of events and occurances in the Military Sandbox that is ARMA. You end up with situations where AI take decisions when shot at, like going prone in the open when Hard cover is 5 metres left or right of them. Equally sometimes bunching up causes no end of problems especially if someone drops a 40mm Grenade or a Bursh of MG fire.

You need to minimise Human interaction, in order for it to be an effective "Artificial Intelligence" it needs to act independently, the problem as I see it is that Complex Routines for Room Clearance will be exceptional Difficult especially with Multiple AI teams doing the same thing the Overhead Serverside will be considerable. Equally will the AI understand the concept of Defender and Attacker, or will both teams Room-Clear towards each other?

Ideally, you should be able to enter the Editor as Civilian and Watch BLUFOR and OPFOR engage one another with Real World Tactics.

I'm telling you what you've got to work with and the constraints, I'm not saying you are wrong (Real World is Real World) but in a Simulation compromises have to made because quite simply you cannot do everything we wish to do with the Technology and the resources available.

Other Examples would be:-

Digging Prepared Positions into the ground that look realistic - Cannot be done.
Better Squad Communication - Massive body of work.
Proper Artillery Calls - Massive body of work.
AI Overhaul - Enormous body of work.

With limited resources, much of this isn't going to be done, so you have to prioritize what can be done and what is on the "Nice to have.." list.

I've said this several times, I support your premise, but with a dose of realism this isn't going to done without some major Fudges and work arounds which may compromise and have unforeseen consequences in other area's of game play.

Ok mwnciboo, I apologize. There are a lot of people on here who say things and have no experience to back it up.

Right now I am trying to create a scenario, and yes it is taking a long time to perfect but I also have a 630-5 job I have to worry about. For the sake of single/cooperative player scenarios, you can make triggers which provide the proper radio etiquette and actions from the AI that best fit how you want the situation to go down.

Right now, I have a trigger and a script (voice) which reads off the real way to call up rotary-winged support. They come in at the perfect time to destroy enemy vehicles that are a Quick Reaction Force to the area which you just assaulted. Now, for artillery support, it's a little harder because you have to personally call that in. I haven't figured out a way to make a trigger which brings up a voice script and the rounds impact at the right time.

Room clearing will require some nasty scripting. A trigger system could possibly be made, but it would require more options, such as Triggered by BLUFOR Conditions: Halted. So once blufor moves to the trigger and stops, the script will cause them to enter the room on predetermined paths. But this only works for user-made/single player/coop missions. Not for when ever you order your team to do so, but the capability is possible.

I apologise too, I have been overly critical or unfair in my wording. (I owe you a beer as an Apology!!).

I've been thinking about what you've suggested....

I think there should be a "Stack up" Command aligned with a Point in Geographic Space outside a Building. A Fireteam if ordered into a Building they would stack up and execute a "Blitz Command".

"Blitz type" Command, this would be a High Aggression, High Dynamic Command which would cause your AI to move quickly, disregarding Cover, Firing on Full Auto and moving at Tactical Pace, and in the type of way you attributed in your premise.

I believe with these two elements we could implement a better Room Clearance.

Another element to think about, is garrisoning Buildings, which would make the Garrison Forces automatically the defenders and therefore couldn't room clear because the Attacking forces are doing this? Once they Un-garrison the buildings they can clear buildings like the Attacking force?

If we could get BIS to do this, we could be onto a Winner.

<a href='http://www.dakkadakka.com/gallery/492136-Storm.html'><img src='http://images.dakkadakka.com/gallery/2013/5/4/492136_md-Storm.jpg' alt='Storm' border='0'></a>

So in these Examples :-

The green Dot is the Orientation Stack up point.
The Blue Dots are a Fireteam.

The Second Example shows the Stack in the wrong orientation, therefore each orientation Stack up point also needs a Vector or stack up direction, to stop the Fireteam doing something silly like stacking up into the open.

I need some help with idea of "Blitz" this would have to be Fast moving, dynamic , no stalling and no stopping e.g Firing on the move!

If you put a Way point into a Building you should get the option to Clear, or Storm. Clear is a cautious Movement through (Default), Storm involves throwing in Grenades and "Blitz" Entry starting from an Orientation Stackup point.

We could even do a Trigger Area, so you place a Radius of say 300m around a Storm or Clear Buildings Trigger, and the AI will methodically clear each building (either by CLEAR, or if ENEMY PRESENT = TRUE then BLITZ) in sequence that is within the 300m Radius of the trigger (Great for huge urban clearance).

We would need to see how we could co-ordinate with Multiple Fireteams e.g an Entire Platoon Clearing a Town because each Squad (and Fireteam) would have to work to cover the area and not clear area's or buildings that had already been cleared...

Once Inside, the team just keeps rolling through without stopping, maximum speed, violence and firepower.

If we keep this a Script Command it means Troops will still act like normal in the open fields, but when clearing a Building will stack up realistically.

http://i602.photobucket.com/albums/tt104/vor033/USMC%20-%204/4f1cd274.jpg

Well, we are taught to never use full auto/burst when clearing rooms because it is dangerous. A really good movie that illustrates this is Act of Valor. It just gets me hard when I see them weapons at the high ready firing single, well placed shots into multiple enemies who are firing full auto cause thy can't hit shit like that. It's mostly a safety concern because you roll into a room, boom you fall with the weapon on fire, and burst half a clip into your buddy before you realize what's going on.

Stack up option is good. It can also be used against you, so you have to be fast. Stacking up makes you vulnerable. Stack is on the door, the AI gives the "ready" option once they are stopped. Making sure they point their muzzles in different directions, and then move into the room. Controlled pair, two shots to the chest every time they fire.

All they would have to do is move along the walls a certain distance and then stop. 1 and 2 man, barrels pointing center, corner, center. 3 and 4 man, center, high. If you would like me to illustrate a complex building clearing technique called leap frogging, I can do so. It utilizes 2 teams doing simple bounding techniques, but you will always have one team stationary at one time. Then you can encourporate more complicated clearing, such as stairs, limited danger areas (LDAs [roads, railroads, paths]), dead space, moving around obstacles such as couches, tables.

That wouldn't be too difficult. To create a garrison building option. AI should be able to interpret entry points, and windows in the building and occupy them properly. I would like a feature that allows the weapon system to be supported by the window sill.

We could easily solve that Blitz issue, to be the controlled style entry e.g Slow is Smooth, Smooth is fast type building clearance, with controlled shooting but still with an always rolling forward movement and not taking cover.

From an AI Stand point, as soon as the Fireteam was in position it would "initiate" automatically and play a sound like "go-go-go", so we wouldn't have to worry about being exposed, once in position they will move immediately.

So what I really need here, is the orientation points, I know ARMA3 has waypoints in buildings but I need a position displaced from entrances so that aren't exposed to those in a building when they stack up, equally windows or Glass walls could be a problem.

I think I might raise another ticket for Stack up Orientation points outside buildings, we could then as a Community exploit this for various scripts etc for many different projects.

http://www.google.com/url?sa=i&source=images&cd=&cad=rja&docid=a41oZ36Vt1rLJM&tbnid=Enc9w87rHvj-pM:&ved=0CAgQjRwwADifAQ&url=http%3A%2F%2Fscience.dodlive.mil%2Ftag%2Fsimulation%2F&ei=waiFUYubGrS24AObnYDgAw&psig=AFQjCNHT0fKpvgYAcEupVs7SLEVfyO5H5w&ust=1367800385472511

The only way to aviod windows, even in real life, is to crouch under them. I can't find any good pictures of stack because there is always two or three guys with their muzzles pointed lazily at the ground.

I believe that, in order to have a whole platoon clearing a village together, you would need triggers on each house saying whether opfor not present for the whole building to be rendered "no more need for clearing" so they don't re clear the same building. Also, each platoon has roughly 4 squads. one squad per building, and one fireteam per room. I'm going o upload a picture of the leap frog method.

http://feedback.arma3.com/file_download.php?file_id=2855&type=bug

http://feedback.arma3.com/file_download.php?file_id=2862&type=bug&show_inline=1&file_show_inline_token=20130505a8c390b457d13040073c3966d613843fb86d989f

One important thing the AI should do when placing their stack is to be far enough back off the door to not expose their rifles past the door. We call that FLAGGING. Pointing your weapon at your battle buddy is also FLAGGING.

I would suggest using the fastest speed with their muzzles up to ensure maximum number of barrels through the door as soon as possible. With us, slow is smooth, smooth is fast, but we can train to have our muzzles over our team mate's shoulder. The main thing for the AI is to MOVE WITH THEIR MUZZLE POINTED IN A SLIGHTLY OFF POSITION FROM THE DIRECTION THEY ARE GOING. So if I'm moving straight towards the door behind someone so they look official, and once they have cleared their corner in the room, they point the muzzle back to the middle of the room

Stalker 1, pretty much all the things you're suggesting are what I'd like to see in ARMA. Hope it happens at some point. +1

It would be great to have a feature which can do all of the above, but the more I think about the solution, the more I feel it becomes impossible to do...

Scripting the AI won't be a big issue, as long as the buildings are "breach ready".

The way I would build this would be the following:

  1. all enterable buildings will get 2 arrays (in config) with unit positions: a) 'stack locations' = left and/or right of entry point and b) 'clear locations' = locations (eg corners of rooms) where AI needs to go and clear before calling out "all clear".
  1. all enterable buildings with windows or other openings will need an array with 'danger zones', so the units won't stand in front of a window with the possibility to get shot before entering.
  1. the buildings will need a prefered stack side (left or right of the entry point), depending on how the door opens.
  1. the buildings will need a prefered stack formation (eg. 2 left, 2 right or 4 left/right), depending on the size and space outside the structure.
  1. if the building has multiple rooms and/or floors, it also needs the above information for each entry point in-door.

With all these variables added to the configs, it will be relatively easy to script the AI behavior since all they have to do is "start at point X, open door, move to point Y, clear, start at point Z, etc..."

The problem is that object makers will need to add all these variables too, and without testing it's almost impossible to see if everything works flawless.

Ofcourse, if the BIS devs or anyone else has a better/easier solution, it would be even better :D

Fri13 added a subscriber: Fri13.May 7 2016, 1:57 PM
Fri13 added a comment.Jun 10 2013, 1:12 PM

Every building would require a accurate waypointing to get these. It isn't after all so big job if there isn't too many different buildings. Like place 1-3 points on both side of doors and every point gives 2 different directions, one for preparation direction and other for entry direction. One on front of door, two on either side. Then get AI somehow to move in tight formation quickly trough door covering different directions, firing on the move without stopping.

Then check which side the AI is and select correct direction rule, causing AI to move first next to door to preparation point with direction of middle of door, open door and move trough it to correct entry point firing on move and when gets to entry point on other side taking correct direction for a second.

Yes, it sounds easy and simple but it isn't.

Do we require dozens of different tactics? I would say no, just 3-4 is enough. Few different clearing tactics and then different manner based the tactical move order (careless, aware, combat etc). If AI is ordered to clear room "combat" AI will throw a grenade first or go almost guns blazing inside, was there a person in or not. If AI is ordered to clear room aware, they move little slower and lowered position. If "careless" mode is selected then AI will stand and walk in normal manner.

Of course I would love to see a recording scripting for this kind things. Where I could just press a shortcut (like Ctrl+F12) in editor preview to start macro recording and stop it. And while recording I could press a shortcut (like Ctrl+left button) what generates waypoint to that position and stores the direction where I was watching.

Then after macro is stopped, it would ask a path name and it is stored for that building as list. I could run few different routes and positions, headings etc to get a 4-7 man squad full route to building. The stored paths could be grouped so all recorded macros (every one presenting different route etc) is in one element.

In editor then I could click the waypoint attached to building and select a the element I want to assign the squad to perform. And of course edited the macro elements to sync the positions.

This would allow to make very pretty entrances to buildings and if players can import/export them, they could be shared among community and even developers could copy them to official game based community voting etc.

I tested once a 5 building clearing on same street and then passing one street between block with 4 soldiers. I came only to conclusion it was better to do with 4 ungroupped soldiers and design every single one a route with speed, stance, timer and sync them together. Result was not so great but it was way better than ordering group to enter building.

I was amazed by the animations and stance positions in ARMA III the feeling how character is possible to be controlled when compared to ARMA 2 and when I entered to first building I got dreams of possibility to get good working building clearing done. Not to do else than learning in ARMA 2/3 editing that it is pain to get even character to choose any pre-made positions in buildings with waypoints.

I don't even remember accurately anymore about Rainbow Six 3: Raven Shield editing possibilities, but in its editor player could rotate the waypoint direction in small 3D view where I was suppose to look. It gave lots of possibilities to avoid AI turning back at enemy or blindspot in room.

In ARMA 2 there was a macro recording capability for helicopters to do a infantry rope landing etc, but recording worked (as far I know) only for helicopters and not well with ground units.

@Fri13 that Rotate Waypoint direction kind of reminds me of the Gun orientation in "Frozen Synapse" which is an excellent Isometric type Tron-esque Tactical Game.

If they could incorporate elements like that into the Editor it would be powerful indeed.

gotmikl added a subscriber: gotmikl.May 7 2016, 1:57 PM

mwnciboo - You never cease to amaze me with how you can over complicate the simplest of issues. Feel free to fast forward to about 5:20 in the below video, and you will see, once again, that this has already been done to a satisfactory level. The command is "assault building"

https://www.youtube.com/watch?v=jHKEMvQU-CY

PS - Working at Chuck E. Cheeses does not qualify as a life in uniform.

@gotmikl did you not read what Fri13 wrote....I was reinforcing Fri13's point that waypoint orientation is a good thing, and it would be powerful to be able to do that with AI Waypointing. Something that we cannot do at the moment.

I'm glad I amaze you. Have you got some axe to grind? Because it seems like you do?

P.S I wouldn't use OF:DR as an example, it was a less than amicable split between BIS and Codemasters. The Engine is also different. Oh and what is this Chuck-E-Cheeses?

You amaze me because you're such a DB. You muddy the water on simple issues. It takes you ten comments and 700 words to up vote? OF:DR modders made that game what it is and we all play both games and want to make the best game possible in the long run. Don't bother responding.

Look son, I'm not sure what to make of you. Your only contribution was a demand for an Xbox 360 controller presets - Something you could have done for yourself if you'd bother looking online.

Your posts, much like several other of "the usual suspects on here" are endless demands for things for which you have little if any understanding. Then when we point out the limits of what can and cannot be done, and you don't listen and make irrational counter points.

It's like trying to convince a child who doesn't want to hear it, that they cannot get to the moon no matter how hard they flap their arms.

Then you have the audacity to say to pragmatic people who work in Development (like me) that we are douchebag's, it's like having a child with no education telling someone who works in that profession they know nothing.

@mwnciboo "that Rotate Waypoint direction kind of reminds me of the Gun orientation in "Frozen Synapse" which is an excellent Isometric type Tron-esque Tactical Game."

It could be "enough" (http://www.youtube.com/watch?v=66OFxFcvP1Q#t=345s) to have a small display on corner where to rotate the direction or on map a arch to rotate it just on horizontal direction. It just would require that map (or at least building) is loaded so player gets the 3D model of the space.
If I now remember correctly, player could with press and hold mouse button rotate that view to get AI then spot/keep an eye vertical position as well.

I find it way better that player is allowed to record routes even with just pressing button to store "position + heading" in game world and AI would move to follow those spots.

As it would work on open environment and in very tight locations like urban environment or just simple "farm" where is few buildings.

My military training was expertise to the urban combat and I really like the urban combat tactics and all other close combats but not so much about open environment and forest combat so I understand the complexity and danger of clearing even a house with few rooms, not to mention of few buildings or whole street in city.

My dream would come from these macro recording what game then stores per building and when player clicks building (object) there is a list of all stored macros, what player can rename as wanted.
As you mentioned the AI behavior of 25/75 percentage error rate etc. I would like to try minimize it by my planning as you do as well know that AI can not be as flexible and tactically smart what human is. But we (mission designers) could help AI to achieve much better tactical behavior.

For open spaces I would take something like this: http://feedback.arma3.com/view.php?id=9421 (My idea, sorry for promoting it as I feel shamed doing so but...) as AI could do just fine in most situations with that feature.

"Equally will the AI understand the concept of Defender and Attacker, or will both teams Room-Clear towards each other?"

Like you said, how does AI know which one is attacking and which one defending a building?

I wouldn't even give that for AI, even if giving them now a waypoint "Hold" and other "Move". I would love to make in each building in my mission a good spots what I would then use and follow the macro.

And to even mention my dream, it would be possible by player to play the macro with a "ghost(s)" like in car racing games when racing against your own record to record more paths, locations and directions to get a multiple.

Think about situation where you plan entry to building with a seven soldier squad, first at outside you position soldiers to starting locations. Then you start planning by choosing one soldier and recording a macro for that building by moving itself. You chose positions, your stance, direction where you look (not just horizontally but vertically as well because game allows to tell all that in 3D space).

Once you have gone trough your entry tactics with first soldier, you stop the recording and you take control of the other soldier and you select macro again and you start recording again in "ghosting" mode, where you can see previously used soldier being controlled by AI and playing your orders and you can now record your second soldier positions simultaneously.

Think about recording all 7 soldiers in ghosting mode?
You would see their timing and actions so you get very complex tactics done where example one covers the window while every one moves at it and croach below it and then get up after passing it.
Macro would as well include all hand signals so you could get soldiers to line up next to wall and give signals when point can continue and start slowly sweeping corner or prepare to entry trough door etc.

It could even be possible to make a soldier do quick spotting around corner from different heights just showing enemy his head second or two and avoid easy headshot from enemy around corner and that just by recording different stances and positions.

Eventually, AI would became terrifying enemy if you could teach your urban tactics to use against you :D

Maybe that is one thing what many didn't get when some players used days planning their missions polishing routes and timings in Rainbow Six games (1-3).

Doing MOUT against an AI like that would be a nightmare!

I like your Macro suggestion, I believe (I maybe wrong) someone has raised that the ability to record Macro's is missing in ARMA 3 editor.

The one thing I was thinking about last night, was the Player / AI models. A Human being can move from A to B whilst their torso faces another direction.

e.g "Their Orientation can be independent of their direction of travel" (within obvious limits so you don't twist 100degrees +. You have always been able to move your head independently of you direction of travel, but specifically I'm interested in the torso (e.g the Orientation of the Weapon).

I'm not sure if that is modelled, as this would be an easy win for Urban Combat, it looks like Torso movement independent of direction of travel is modelled in Tactical Stance?

@mwnciboo "The one thing I was thinking about last night, was the Player / AI models. A Human being can move from A to B whilst their torso faces another direction."

The my idea what I linked would give that possibility. Now human player can hold look button (Alt) when aiming and gets about 20 degree free aiming zone independent from moving direction but is limited to walking. But gets about 270 degree free when not aiming and just turning head while even running.

In that my idea the waypoint speed would allow player to edit a group moving from waypoint A to B in "Limited" speed and give custom direction to each soldier in group and they would aim, walk sideways or at least run while keeping eye of the ordered direction, like right now with the script "Look at" parameter and playing around with objects what to look, just to make it simpler.

I need search a ticket for that and if not to add one, about widening the degree what is possible to aim while walking from 20 degree, lets say to 190 degree as it is not so easy in urban areas to do watch to flanks when walking.

We have the animation when player does corner watch to left side (Ctrl+A) and character moves weapon stock from right shoulder to left shoulder to give a better aim. As now everyone are right handed soldiers and no one use weapon on left hand than in that special stance position so aiming to flanks when walking in high stance is impossible.
There should be animation where soldier swap weapon hand when walking forward or kneeling (and swapping kneel) to allow aiming to right flank from heading 340 degree to 90-100 degree. When weapon would be kept on current position on left hand, aiming heading would be from 20 degree to 260 degree covering the left flank. It would force player to knowledge the demand to swap weapon sides with stance and allow a AI group move in tight formation where each soldier is covering correct flanks while still moving forward slowly.

I know for many people it is difficult task to fire swapping weapon side but for personal experience I know it can be learned to fire as well on either side.
In urban combat it would be huge benefit to have possibility swap weapon side from right to left depending corner and direction where aiming when walking. For game it shouldn't be more than just mirroring the animations of model skeleton and disallowing some special tasks like aiming a attached grenade launcher as reticle is on left side of weapon.
It would give some other problems for animations like clearing jam or reloading etc but I think those could be done by swapping weapon for that moment to right hand and then back to left after completed.

I have binded Alt+E and Alt+Q to infantry turn functions so I can turn character without mouse. And when I hold Alt, I have that 20 degree free look zone. So I can sidestep, turn and move back/forward, but it isn't perfect as turning character moves the whole body not just legs, even in the 20 degree zone. I need to add a ticket for that so when turning the character with keys, aim doesn't move unless you hit that 20 degree free aiming limit.

Aside from all the essays about AI and how clearing a room works, this is very simple. Assigning a group of AI to clear a room is a very easy thing to implement. People, they were doing this in games back when you were still collecting Pokemon cards. Really, if the AI in ArmA 3, a game from 2013, can't handle something that a game from 2002 can do with ease, then that's just embarrassing. Yes the maps are huge but that has nothing to do with it, they already have positions assigned to buildings and the AI know how to navigate buildings already, so I'm sure "stacking up" to breach and clear a room wouldn't be so extremely and impossibly difficult that they wouldn't implement it.

Short Version: Upvoted. Get on it, BIS.

Here that everyone!? BIS? This is a 'very easy thing to implement'. Case is now closed, the authority has spoken.

Hear*

I'm not the authority, the community is, the people who buy the game are. 100% of people who voted for this voted for it to happen, and "clearing rooms" has been happening in video games for over a decade. Heck, even Call of Duty does it, it would be a shame if ArmA 3 failed to implement it.

Perhaps a more intelligent comment and an opinion on the issue would help more than a smartass remark, mepwaygame.

Just because other games do it doesn't mean it's 'very easy' to do. I'm not saying it's the hardest AI problem ever, but it'd be nice to show a little respect for the people working on this thing. I want it in the game much as anybody, and I would agree it's worth the EFFORT to implement. You HERE what I am saying HEAR today? You're the one who started these smartass remarks anyway :P

This, easy or not, must be implemented... We still have the same product (with better graphics) from 2001 (OFP)... That's something the ppl who came from DayZ dont understand...

Mitor added a subscriber: Mitor.May 7 2016, 1:57 PM
Mitor added a comment.Jul 25 2013, 1:35 PM

Absolutely upvoted.

Please BI, could you try to implement it please?

Maybe it won't do it, ok then. Maybe it WILL!

Upvoted with both hands!

Please teach the AI to assault the building! May be if this AI uses waypoint "SEEK AND DESTROY". In this case, after all, the AI must find enemy and destroy it!
If AI-squad sees that a player located inside a building, then this squad should penetrate into this building.
Each building usually has several doors.
I think, during the assault, the AI should go into each door simultaneously and try clean this building!

BIS can do a special action -"ASSAULT" for every squad leader of squad in the actions menu.
For example, if the player is a commander moves the cursor to the building where enemy detected, then let this commander will be able use to assault/cleaning command only for this building. This will cool!
In this case the player or the AI, under special conditions can give the order to enter the building and to clean it.

This detail can be a great addition to tactics of the player-commander and for realistic AI behavior (if he uses waypoint "SEEK AND DESTROY")

Arkhir added a subscriber: Arkhir.May 7 2016, 1:57 PM

"assault" waypoint is a good idea, would be sweet both in Zeus and editor. Currently ARES mod provides Zeus with extended funcionality, so you can order AI to search the building, but I'd love to see BIS take on that, since we've seen some new waypoint types added recently.

@trooper226
https://www.bistudio.com/company/careers

I have to argue, the Bohemia developers still cannot even get close to proper human simulated activity. Of which, would usually merrily be resolved by integrated random number calls at the proper times, within the computerized reactions. But the calls tend to be resource accumulative.

What makes you think they can then get this right? Expect bugs. Expect lots of bugs! (Expect the unexpected. ;-)

ie. Bug #17772, "AI vision not being affected by enviromental changes / obstacles / vegetation"