Jump to content
klamacz

AI Driving - Feedback topic

Recommended Posts

1 hour ago, road runner said:

Again, your preaching to the converted, but at the end of the day, what's the alternative?

 

Open source the engine ? If each one of us was correcting a small issue (no more black borders in the editor, alphabetical sort of the maps, "signaling flare"...etc), I think A3 would be better at a faster pace than what we currently have.

  • Like 4

Share this post


Link to post
Share on other sites
 1:06:48 Out of path-planning region for O Bravo 1-1:1 at 12419.1,7723.1, node type Road
 1:06:48 Out of path-planning region for O Bravo 1-1:1 at 12419.1,7723.1, node type Road
 1:06:48 Out of path-planning region for O Bravo 1-1:1 at 12419.1,7723.1, node type Road
 1:06:48 Out of path-planning region for O Bravo 1-1:1 at 12419.1,7723.1, node type Road
 1:06:49 Out of path-planning region for O Bravo 1-1:1 at 12419.1,7723.1, node type Road
 1:06:49 Out of path-planning region for O Bravo 1-1:1 at 12419.1,7723.1, node type Road
 1:06:49 Out of path-planning region for O Bravo 1-1:1 at 12419.1,7723.1, node type Road
 1:06:49 Out of path-planning region for O Bravo 1-1:1 at 12419.1,7723.1, node type Road
 1:06:49 Out of path-planning region for O Bravo 1-1:1 at 12419.1,7723.1, node type Road
 1:06:49 Out of path-planning region for O Bravo 1-1:1 at 12419.1,7723.1, node type Road
 1:06:49 Out of path-planning region for O Bravo 1-1:1 at 12419.1,7723.1, node type Road
 1:06:49 Out of path-planning region for O Bravo 1-1:1 at 12419.1,7723.1, node type Road
 1:06:49 Out of path-planning region for O Bravo 1-1:1 at 12419.1,7723.1, node type Road
 1:06:49 Out of path-planning region for O Bravo 1-1:1 at 12419.1,7723.1, node type Road
 1:06:49 Out of path-planning region for O Bravo 1-1:1 at 12419.1,7723.1, node type Road
 1:06:49 Out of path-planning region for O Bravo 1-1:1 at 12419.1,7723.1, node type Road
 1:06:49 Out of path-planning region for O Bravo 1-1:1 at 12419.1,7723.1, node type Road
 1:06:49 Out of path-planning region for O Bravo 1-1:1 at 12419.1,7723.1, node type Road
 1:06:49 Out of path-planning region for O Bravo 1-1:1 at 12419.1,7723.1, node type Road
 1:06:49 Out of path-planning region for O Bravo 1-1:1 at 12419.1,7723.1, node type Road
 1:06:49 Out of path-planning region for O Bravo 1-1:1 at 12419.1,7723.1, node type Road
 1:06:49 Out of path-planning region for O Bravo 1-1:1 at 12419.1,7723.1, node type Road
 1:06:49 Out of path-planning region for O Bravo 1-1:1 at 12419.1,7723.1, node type Road
 1:06:49 Out of path-planning region for O Bravo 1-1:1 at 12419.1,7723.1, node type Road
 1:06:50 Out of path-planning region for O Bravo 1-1:1 at 12419.1,7723.1, node type Road
 1:06:50 Out of path-planning region for O Bravo 1-1:1 at 12419.1,7723.1, node type Road
 1:06:50 Out of path-planning region for O Bravo 1-1:1 at 12419.1,7723.1, node type Road

 

A75FCE28EF3436B97AC8770D65375D1CF9B7EC06

 

641AABA64B507C3466763D779540E783A11A6AFD

 

 

 

  • Like 1

Share this post


Link to post
Share on other sites

After so much whining, I believe it's time for a proper repro again:

 

Demo Mission:

https://www.dropbox.com/s/o549o0nn5t7a2uq/AI_vehicle_pathfinding3.Tanoa.zip?dl=0

 

1st Video:

Just a recording of the AI driving with AIDriving diag enabled.

 

2nd Video: 

Shows the paths created in the first video and  highlights critical areas.

The arrows on the ground show the path of each vehicle and the direction of the arrow shows the direction of the vehicle which it had.

 

Blue Arrow: APC

Cyan Arrow: Prowler

Red Arrow: HEMMT

Green Arrow: Hunter

 

Things I noticed:

- Stopping at every waypoint

- Inconsistency between runs. Every run I did, made the vehicles use a slightly different path.

- Wooden bridges are an issue, see the 2nd Video. The HEMMT actually missed the bridge, turned around and used a completely wrong path.

 

 

 

  • Like 10

Share this post


Link to post
Share on other sites

I'd love to see someone put one of these vehicles inside the NE base on Sangin. Somewhere deep inside it. Then place a player in another vehicle and have the AI vehicle follow you out with the return to formation command. 

 

The last time I tried it was one of the most frustrating things I've ever tried to do in this game.

 

Thanks for the video, R3vo. That does highlight some interesting behavior.

  • Like 3

Share this post


Link to post
Share on other sites

Sorry to repeat but I guess my question was buried under the last general considerations.

Hi, is it possible to make something for more fluidity when a vehicle is completing an intermediate waypoint ? I seems the vehicle stops at each waypoint, then continue. Why?

  • Like 2

Share this post


Link to post
Share on other sites
5 hours ago, pierremgi said:

Sorry to repeat but I guess my question was buried under the last general considerations.

Hi, is it possible to make something for more fluidity when a vehicle is completing an intermediate waypoint ? I seems the vehicle stops at each waypoint, then continue. Why?

Because the vehicle commander is an officer and he can't read a map!! ? 

 

In all seriousness I cannot disagree with the issues concerning AI driving ability or lack of it to be more precise!

Share this post


Link to post
Share on other sites

Although I couldn't test the recent R3vo example due to not having the mods I did try the older version and it also had the HEMMT  missing the bridge and it then got stuck.

 

It seems something must have changes since I last tried it a week or so back as now they're not going the wrong way at junctions and doing U-turns.

 

One vehicle did try and do a U-turn for no obvious reason on an open road.

Share this post


Link to post
Share on other sites

I decided to step back from Dev Branch to try some things out in Stable, that worked for others on stable, but not for me.

I'm going to have to eat a little bit of humble pie when it comes to AI driving, I created a little experiment, and on Stratis I created 3 small 2 vehicle groups, and gave them all a SINGLE waypoint to their destination, straight from A-B

Group1 was 2 MRZR's with a single way point from Camp Tempest to Air Station MIKE I set the speed to limited, and the formation column

Group 2 was 2 M1232 MRAPs with a single way point from Air Station MIKE to a compound in Stratis Airbase the speed was normal, and again formation was column

Group 3 was 2 M1A2SEPv1 (TUSKII) with a single way point from a compound inside Stratis Airbase to Camp Rogain, the speed was full, and formation column

 

I then used Zeus to monitor all 3 groups one at a time from start to finish.

Group 1 did not stray off the roads once, and followed the roads taking the turns as and when needed, they did not go direct from point A to point B, they actually did their job.

Group 2 was the exact same as group 1, and followed the roads, no veering off into the Ulu

Group 3 The MBT's 95% of the time also used the roads, because they were traveling full speed, taking corners close to obstacles was a small issue, they knocked them down and continued, they went off road once, on the single track that leads to Rogain going from West to East, just before the wooded area, once they got there, they took the road all the way into Rogain.

 

I have to be honest, I was pleasantly surprised that the AI driving actually wasn't that bad, what I believe maybe the issue is when creating too many way points, and the AI has to stop and think what does it need to do at each way point, and if the way points are close together, it throws a spanner into their collective brains. I kept it as simple as I could, a point A to point B.

Now, for the serious planners this creates various issues, like trying to be true to Mil doctrine in using way points and then FUP's and then having the AI crews, switch to battle mode etc. Perhaps it's better to use triggers along with the way points to create a more fluid battle plan?

I used Stratis as a test, as it's small, and it's not so compact as other maps out there.

I'm off to eat a plate of humble pie and then try this with groups of 4+ vehicles now on other maps.

  • Like 3

Share this post


Link to post
Share on other sites
2 hours ago, road runner said:

 I believe maybe the issue is when creating too many way points, and the AI has to stop and think what does it need to do at each way point, I kept it as simple as I could, a point A to point B.

 

Without saying that the behavior/combat mode of the units can override the path.

  • Like 2

Share this post


Link to post
Share on other sites
2 minutes ago, pierremgi said:

Without saying that the behavior/combat mode of the units can override the path.

I should have also mentioned that groups 1 and 2 were in Hold Fire, and safe modes

Group 3 was in aware mode.

Very good point though.

Share this post


Link to post
Share on other sites

I need a new type of way point... " vechicle_01 Hit target_01; " 

 

I need this because I am trying to make a car bomb that will literally move and Hit the target spot without AI trying to prevent fatal collisions. If I want this to be a command to kamikaze pilot, driver of a car bomb, he will do what must be done.

Share this post


Link to post
Share on other sites

This is a request, and in DEV Branch, this should be asked in the Editing thread, it has no real bearing on what DEV Branch is about.

  • Like 1

Share this post


Link to post
Share on other sites

Sorry if this was mentioned already, but following feedback:

 

I can only assume this is because of turrent vehicles, but for vehicles that dont have attached turrents, is it possible to make drives eject vehicles during combat?  For whatever reasons the driver stays in vehicle driving around aimlessly without any impact and eventually gets killed.  Be great if they exit their vehicles with the rest of the team to fight.  Only way to force them to exit is shooting out the tires....

Share this post


Link to post
Share on other sites

unfortunately we haven't heard from klamacz (BI Dev who was tweaking the driving logic) in 3-4 months. what the implications of that are, i dont know.

Share this post


Link to post
Share on other sites
2 hours ago, fn_Quiksilver said:

unfortunately we haven't heard from klamacz (BI Dev who was tweaking the driving logic) in 3-4 months. what the implications of that are, i dont know.

He's still active - at least he logged onto the forums on Monday. So let's hope his absence from this thread means he's hard at work bashing out code.

Share this post


Link to post
Share on other sites

I was testing some AI driving and waypoints in Virtual map and noticed that the waypoints complete at about 150 meters from the WP.

If you copy the the mission and place it on a real map they drive to the waypoints.

 

I wonder why the terrain would effect the the waypoint in such a way.

Share this post


Link to post
Share on other sites

srsly, devs take a look at vcom driving AI. hes gonna publish an update hopefully soon.

 

compared to vcom driving the vanilla AI(yes the one on dev branch) is just unusable. specially on custom maps with high density like Taunus or other x-cam maps they cant drive without crashing into something for longer than 30secs.

  • Like 1

Share this post


Link to post
Share on other sites
3 hours ago, HeroesandvillainsOS said:

Wow this video. Genesis is doing really amazing stuff:

 

https://m.youtube.com/watch?v=93vAKOJ3I9c&time_continue=289

 

should tweet that to BI CEO, ask him his opinion on the ArmA 3 AI driving

 

and to be fair to the vanilla driving, Genesis attaches cans of dog food to everything, and also appears in that clip to be using some hit or damage events on the AI to move them aside. basically performance-wise its not cheap what he is doing.

 

What BI needs to deliver is vehicles that can drive for an hour without breaking themselves, breaking every road sign, breaking other vehicles and simply breaking the AI itself and freezing on the road.

  • Like 1

Share this post


Link to post
Share on other sites
7 minutes ago, fn_Quiksilver said:

 

should tweet that to BI CEO, ask him his opinion on the ArmA 3 AI driving

 

and to be fair to the vanilla driving, Genesis attaches cans of dog food to everything, and also appears in that clip to be using some hit or damage events on the AI to move them aside. basically performance-wise its not cheap what he is doing.

I don't doubt it. But using VCOM AI Driving in the past was simply amazing and I never noticed any performance issues.

 

What's astounding is how in his vanilla test, over 40% of the foot infantry were killed.

 

By one single truck on a relatively straight road.

 

That's abysmal.

  • Like 1

Share this post


Link to post
Share on other sites
35 minutes ago, HeroesandvillainsOS said:

I don't doubt it. But using VCOM AI Driving in the past was simply amazing and I never noticed any performance issues.

 

What's astounding is how in his vanilla test, over 40% of the foot infantry were killed.

 

By one single truck on a relatively straight road.

 

That's abysmal.

 

sometimes we are a bit harsh with the devs :) they are probably juggling multiple internal projects, and as always time is the finite resource. it is probably a big job to work on the driving.

 

what would be good is a bit of communication, like "yes there are some broken things in 1.66, some are scheduled for 1.68-1.70 fix and others are not feasible"

  • Like 2

Share this post


Link to post
Share on other sites

Seems like stealing mission idea's like their genius revolutionary new idea "combat patrol" seems to be more important than this.

 

I will never understand why priorities in this company are so fucked up and im sick of just everyone goes along with it. Just check how long this game has been released and still there is a ton of bugs in the game that even existed in ARMA 2. I rather keep speaking up about this major issue and get me banned again and and again. i just dont give a fuck anymore about your stupid feelings BIS.

 

I demanded that the guy who is responsible to the long term strategy to be fired like couple years ago already.(fastest forum ban in my life) 

Im pretty sure many Devs would love to fix arma but there is this guy or this group of people who is pushing to milk the most money out of everyone asap instead of providing a solid engine that returns steady profits for decades.

 

The only thing that makes people stop playing arma and stop modding/scripting is basically the sheer amount of bugs. How many good modders have we already lost to other engines like UE4? Imho the best are long gone... They are already publishing standalones or working hard on them. And its because of this retarded long term strategy.

 

If BIS would have invested all their effort in fixing the game, instead of making unwanted, unneeded, overpriced and low quality DLC like marksmen, kart, tanoa, etc... we would have alot bigger player base, alot more modders and scripters etc.

 

Just look at Skyrim. These guys still focus on engine issues, years after release and dont waste their effort with releasing DLC that get boring after 2 hours. And it's still the most played RPG of all times and will probably be for the next decade.

 

And ARMA is dieing a slow and painful death. RIP. Mark my words: In 5-10 years, BIS, EA and Ubisoft will be on the same level of consumer exploitation.

Share this post


Link to post
Share on other sites
1 hour ago, Johnny Drama said:

I demanded that the guy who is responsible to the long term strategy to be fired like couple years ago already.(fastest forum ban in my life) 

You can not possibly think that you alone could have any impact on BIS hiring matters? I mean, sure. Buy the majority of stocks in BIS and sack the guy if you want. But demanding it on a forum is a far stretch.

 

1 hour ago, Johnny Drama said:

How many good modders have we already lost to other engines like UE4? Imho the best are long gone... They are already publishing standalones or working hard on them.

Many modders realize that they can use their skills, developed through arma modding, to create their own games to profit from them. It's a huge risk and calls for a lot of personal investment. However though, many so-called "arma 3 destroying competitors" still can't offer the same fidelity and experience as Arma 3 can. Wouldn't you say? Arma holds its ground because there isn't anything like it yet, and probably won't be for a while.

 

1 hour ago, Johnny Drama said:

instead of making unwanted, unneeded, overpriced and low quality DLC like marksmen, kart, tanoa, etc...

Whoa whoa... Slow down. So you mean the FREE engine upgrades that came with the marksmen, tanoa, helicopter expansions are wasted? The extra addons are all high quality and add to the game, but the core mechanics such as weapon resting, air-lifting, graphics update, audio update, 3DEN editor are all HUGE free expansions that enhance the ARMA experience.

 

1 hour ago, Johnny Drama said:

Just look at Skyrim. These guys still focus on engine issues, years after release and dont waste their effort with releasing DLC that get boring after 2 hours.

Are you saying that BIS are not focusing on engine issues? From my point of view they are still working on fixing and improving. With each patch, there's always a little new content, and a little redone code for better performance and or core mechanics.

 

1 hour ago, Johnny Drama said:

And ARMA is dieing a slow and painful death. RIP. Mark my words: In 5-10 years

This is a matter of personal opinion. I personally look differently at it. BIS somehow have the balls to keep releasing free content, and allowing compatible play with owners of new DLC and non-DLC owners, unlike many other gaming companies. I think all of the ArmA community know that Arma 3 has an expected life time of a few more years, but from all the experience BIS have using the Real Virtuality engine, they will create something new. I believe somewhere in the future there will be a new ARMA engine that is much better in that one field we all know arma has challenges, which is performance. Once BIS makes the perfect compromise between simulation and smooth performance, we will have our "ARMA 4". The question remains though: if good performance sacrifices the hard-core realism, such as bullet penetration and large scale AI combat, will more people play it? Or will the target group of players just shift from "hardcore pre arma 4" to "arcade-ish arma 4+"?

 

Either way, the future is bright.

 

I personally can't wait for Jets DLC :) I also have an eye out for Tanks, which I find rather bland and uninteresting in current arma (RHS mods armor is really good though).

 

To some people, arma 3 bugs is just another word for "splendid franchise charm". 

 

I hope you find some comfort in these words, rather than pure criticism.

  • Like 5

Share this post


Link to post
Share on other sites

Yes it's dying that much it's in the top 50 grossing games on steam.:eh:

  • Like 7

Share this post


Link to post
Share on other sites

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now

×