AI Driving - Feedback topic

[escforreality 35]

I 100% support this. AI driven vehicles to which the player is commander should not be murdered by the driver when he gets out of the machine. 

(1) Friendly AI driving any kind of vehicle needs to prioritize avoiding their own side and civilians. 

(2) For this specific issue vehicles should not follow an on foot commander unless specifically ordered to. I.E. A commander exiting a vehicle should also trigger a "Stop" command to the AI. 

 

It is not typically useful the have a vehicle follow you in formation. 

[R3vo 2652]

I support this too. Getting out of a vehicle as commander should trigger "Stop" command for this specific vehicle.

[pierremgi 4739]

As I wrote, stop crushing any time any civilian/friendly units, just enemies on path!

[dragon01 902]

TBH, collision physics between people and cars need to be looked at. Right now, the vehicles just flatten anyone in their path. IRL, it's a much more complex process. A person hit by a car will either be thrown in the air (if the speed is high enough) or slide across the hood and hit the windshield if it's not (sadly, I'm speaking from experience, at least as far as the latter case goes). Also, even at low speed, civilian cars wouldn't come out unscathed, with impact possibly resulting in hood deformation, headlight and radiator damage, especially if you hit a large man clad in heavy combat gear. Either way, the car will be slowed down by the impact. Military vehicles are trucks are tougher, but they should also take damage, especially to the windshield and lights. At low speeds, it should also be possible to survive the impact (especially if wearing a helmet), although with heavy wounds. 

 

AI, on its part, should try to jump or dash out of the car's way, though I understand that if existing animations can't be used for that, it'd be unfeasible. On the other hand, make sure that if they see the vehicle coming, they don't actively walk or run into its path (don't worry if it doesn't work all the time, humans can't always manage that, either...). Both vehicles and humans need to have routines dedicated to avoiding colliding with each other.

[andersson 285]

I seen videos where one can see how the AI vehicles plan their route. If the foot soldier AI could be aware of that route 50m infront of the vehicle and move a few steps away. In RL people move out of their way from vehicles.

But I guess easy to dream up and hard to implement.

[mickeymen 322]

In this topic, long ago I don't see the messages from the developers. This ticket is forgotten?  I hope BIS is reading their own forums and this topic in particular!  In this thread a dozen important ideas, desires, definition errors.  

Somebody of the developers, please report that you are reading this ticket. People occasionally report about bugs or nice ideas. I would like this conversation to take place not only between users, but also, at least occasionally with developers

[Alwarren 2765]

56 minutes ago, mickeymen said:

In this topic, long ago I don't see the messages from the developers. This ticket is forgotten?  I hope BIS is reading their own forums and this topic in particular!  In this thread a dozen important ideas, desires, definition errors.  

Somebody of the developers, please report that you are reading this ticket. People occasionally report about bugs or nice ideas. I would like this conversation to take place not only between users, but also, at least occasionally with developers

 

I would imagine that work on the Jets DLC is in high gear at the moment, so this topic will be on the backburner. I do not believe it is forgotten.

[mickeymen 322]

13 hours ago, Alwarren said:

 

I would imagine that work on the Jets DLC is in high gear at the moment, so this topic will be on the backburner. I do not believe it is forgotten.

 

Today, there is work on Jets DLC, tomorrow the work will go on Orange DLC, the day after tomorrow will go to work on Malden DLC ,after all this work will go on - TacOps DLC, Tanks DLC, etc, etc, etc.  

 

You still do not believe that this ticket is forgotten?

Tomorrow it will be forgotten as today, because DLC in this game has no end.

 

 

 

[Alwarren 2765]

 
Tomorrow it will be forgotten as today, because DLC in this game has no end.

Oh well, believe what you want. But maybe concentrate on actual feedback instead of bitching about on of the most openminded and generous DLC strategies I've seen so far?

Sent from my SM-T580 using Tapatalk

[dragon01 902]

Also, jets fly in the air, so they need different AI logic. Tanks drive on the ground, so the devs will have to revisit the driving system then.

[R3vo 2652]

1 minute ago, dragon01 said:

Also, jets fly in the air, so they need different AI logic. Tanks drive on the ground, so the devs will have to revisit the driving system then.

 

Yeah, my guess would be that tracked vehicle driving will get improved when the work on the tank DLC starts.

[mickeymen 322]

1 hour ago, R3vo said:

 

Yeah, my guess would be that tracked vehicle driving will get improved when the work on the tank DLC starts.

 

Yes, and until Release the Tank DLC, 13756 soldiers, will be crushed by means allied land vehicles ) Lol!

[mickeymen 322]

I just want to say that some problems of AI-Drivers, need to be solved right now, not postponing to the last point in the roadmap 2017 

[Alwarren 2765]

2 hours ago, mickeymen said:

I just want to say that some problems of AI-Drivers, need to be solved right now, not postponing to the last point in the roadmap 2017 

Provide repros and feeback then. Better than just a diffuse "they must fix their AI driving".

[pierremgi 4739]

24 minutes ago, Alwarren said:

Provide repros and feeback then. Better than just a diffuse "they must fix their AI driving".

Yes, we know, criticism is easy, art is difficult. But, it seems to me there are plenty of feedback on this subject, and, after all, I can't imagine BI developers need more videos of rotten convoys, dumb drivers stuck on a street corner, drivers crushing their group fellows, unseen rocks, trees,..

[mikey74 168]

Hi, Just curious. Is it possible to attach sensors on the vehicle maybe 5 to 10 meters in front of his direction and run an object finder routine without significant frame loss? I did it once a while back. It worked out ok, but if you put several vehicles on the map. There was frame rate loss. Thus why I never released it. I guess why I ask. I have limited knowledge in coding, but you guys are very good and way more knowledgeable than me. If I can pull it off with some trouble and frame drop. I'm thinking you guys could do it and know how to keep frame rate unhindered, and smoothly. Just a suggestion in case you guys may have overlooked that possibility. Or lol its not possible. :) Thanks for the wonderful work you guys have done.

[The Man Without Qualities 110]

On 11.4.2017 at 7:10 PM, Alwarren said:

Provide repros and feeback then. Better than just a diffuse "they must fix their AI driving".

With all respect, the entire A3 as is is a repo. If an AI driving is so much bugged, it requires a real overhaul, individual tickets etc are non-sense in this context. It is not a good design with few bugs, it is a wrong design or a maybe good design with a totally wrecked implementation. Noone needs to be an expert for this conclusion, just use A3.

Whenever I call a vehicle with 1-1 to my position, in almost 100% of a tracked, it is just overrun my Avatar, for wheeled it is a little bit better.

So it is a bug? Or f*ed up by design?

[Alwarren 2765]

12 hours ago, The Man Without Qualities said:

With all respect, the entire A3 as is is a repo. If an AI driving is so much bugged, it requires a real overhaul, individual tickets etc are non-sense in this context. It is not a good design with few bugs, it is a wrong design or a maybe good design with a totally wrecked implementation. Noone needs to be an expert for this conclusion, just use A3.

Whenever I call a vehicle with 1-1 to my position, in almost 100% of a tracked, it is just overrun my Avatar, for wheeled it is a little bit better.

So it is a bug? Or f*ed up by design?

 

Neither that statement nor that attitude is helpful. You can guess quite accurately that a "complete overhaul" will not happen in the near or mid-term future. Yes there are certain shortcomings. A lot of them. I know that if I disembark a vehicle from the commander's seat the driver will almost certainly run me over. I can either just jump out and curse at the driver every time, or I can order him to stop first and THEN get out.

 

Now, I have a problem in one mission of mine where the players have to protect a convoy, and every time the convoy gets attacked it tries to turn around and drive back to its origin. That is an issue that needs fixing. Calling a vehicle to your position, order it to stay back. Again, you can either actively try to seek out and provoke the issues, or work around them; most can be worked around. Is it an ideal situation? No it isn't. But as I said, your "complete overhaul" will not happen, so you can either work with what you have, or just walk away from it.

 

That is why I said "provide repos". It will get specific issues fixed, or at least increase the chance that it will get fixed. You can dwell on the general "shitty driving AI" all you want, but it will not get you anywhere.

[fn_Quiksilver 1633]

play PvP ;)

 

AI too expensive to program atm, with all the car/transportation companies paying top dollar for good AI programmers, raised the salary so its simply cheaper for game studios to work on PvP/pretty-looking artsy stuff and push 'emergent player-driven gameplay', instead of paying huge dollars for talented designers to work on huge logic machines to make good sandbox AI.

 

of course i am not serious, just spitballing!!!

 

For some context, 5 mins ago I ordered a truck to go forward 100m on a road and it failed.

[f2k sel 164]

How is the driving over bridges these days, I tried a week ago on Stratis and the vehicle wouldn't cross the bridge at  Agia Marina  but then I tried on Tanoa and it crossed the wooden bridges easily.

 

Previously it was the other way around.

 

 

 

[The Man Without Qualities 110]

6 hours ago, Alwarren said:

 

Neither that statement nor that attitude is helpful. You can guess quite accurately that a "complete overhaul" will not happen in the near or mid-term future. Yes there are certain shortcomings. A lot of them. I know that if I disembark a vehicle from the commander's seat the driver will almost certainly run me over. I can either just jump out and curse at the driver every time, or I can order him to stop first and THEN get out.

 

Now, I have a problem in one mission of mine where the players have to protect a convoy, and every time the convoy gets attacked it tries to turn around and drive back to its origin. That is an issue that needs fixing. Calling a vehicle to your position, order it to stay back. Again, you can either actively try to seek out and provoke the issues, or work around them; most can be worked around. Is it an ideal situation? No it isn't. But as I said, your "complete overhaul" will not happen, so you can either work with what you have, or just walk away from it.

 

That is why I said "provide repos". It will get specific issues fixed, or at least increase the chance that it will get fixed. You can dwell on the general "shitty driving AI" all you want, but it will not get you anywhere.

 

Some generic thoughts upfront:

- There is not a single AI driving problem in the A3 world that was not mentioned in this forum or in the bugtracker.

- even more: most issues are inherit from OFP/A1/A2 and reported in BIS forum and community BTs where BIS had access to

- a generic attitude like yours to restrict yourself in expectations and from defining the ideal solution before trying to find the optimal solution is....wrong

- BIS should be pretty much aware which goods and bads exist in their engine

 

Leads me to the question: Why is BIS inviting to such a communication in a forum? Are they clueless about the capabilities of AI driving? Do they know and want to pretend taking care? Are they really interested to iron out the real issues existing since OFP?

 

I can't tell you. But I can tell you that I am fed up reporting again and again the same issues to BIS, even to representatives that were not even born when OFP was released.

I am also fed up by semi-senior community members giving me advises how devote I have to accept serious bugs and that I have to seek continieously for workarounds and that I must remember 10000000000000000  WAs for all those issues.

 

What would I be willing to do/contribute?

 

1.) Reporting bugs/issues and provide information and support for reproducing the bug scenario - > DONE!!!

2.) Buying all OFP sequels to keep the BIS boat swimming - > DONE!!!, sometimes several copies per game

3.) purchasing almost every shitty BIS DLC to keep BIS boat swimming -> DONE!!!

4.) Doing crazy crap and micro-management to make missions and gameplay work - despite all bugs -> DONE!!! I deserve a medal!

So you can see, all that was already done, what did it help? To little and far to late!

 

 

What do I expect?

 

1.) Real and transparent treatment of bugs - > just partially done -> the BIS bug tracker is IMHO just a huge pile of information not properly treated by BIS

2.) Serious attempts to overcome severe design issues in engine and function design - at the moment I rather see tweaking around to hide the worst symptoms

3.) More insight into the solution design BEFORE it is reaching a developer for coding - I have seen so many anouncements of solutions for bugs or other issues - where the real result was rather...disappointing.

 

 

What would that mean for AI driving capabilities?

 

It would mean to me that BIS should communicate more in detail HOW it works, WHICH parts of AI path finding and driving works OK and in which part they see potential for improvements.

 

The questions that MUST be asked are:

 

1.) Are the methods how the environment is made transparent to the AI routines the best way? What other methods are feasible?

2.) That each AI driver is an individual entity fighting for its own way without having higher level routines like "crowd manager"; "cross road manager"; "right of way manager"; "damn we queing up at a bridge - does it make sense to wait manager"....

3.) Review of waypoint handling: the end waypoint....should the vehicle stop there or just close to or should it just start slowing down from max-speed when reaching waypoint...

4.) Obstacle detection: review the methods and frequency the routines checking the clearance of the path - depending on speed and on known other individual AI drivers present in the closer environment.

5.) Formation handling: It is so obvious that AI react to late, drive to close and to fast, that they do not adapt distance depending on speed...all that needs a real overhaul. The tendency to follow the previous vehicle instead of just keeing order and distance and following the own path would make driving more fluent. When I order vehicles individually I get usually better results compared to group handling in formation - it would be perfect if individuals would maintain a minimum distance to the vehicle in front.

 

 

Just a few cent...

 

[fn_Quiksilver 1633]

7 hours ago, f2k sel said:

How is the driving over bridges these days, I tried a week ago on Stratis and the vehicle wouldn't cross the bridge at  Agia Marina  but then I tried on Tanoa and it crossed the wooden bridges easily.

 

Previously it was the other way around.

 

 

 

 

its fine on the wooden bridges

 

its the low stone ones they have lots of troubles with on Tanoa, and also sometimes on the long bridge too.

[fn_Quiksilver 1633]

4 hours ago, The Man Without Qualities said:

 

Some generic thoughts upfront:

- There is not a single AI driving problem in the A3 world that was not mentioned in this forum or in the bugtracker.

- even more: most issues are inherit from OFP/A1/A2 and reported in BIS forum and community BTs where BIS had access to

- a generic attitude like yours to restrict yourself in expectations and from defining the ideal solution before trying to find the optimal solution is....wrong

- BIS should be pretty much aware which goods and bads exist in their engine

 

Leads me to the question: Why is BIS inviting to such a communication in a forum? Are they clueless about the capabilities of AI driving? Do they know and want to pretend taking care? Are they really interested to iron out the real issues existing since OFP?

 

I can't tell you. But I can tell you that I am fed up reporting again and again the same issues to BIS, even to representatives that were not even born when OFP was released.

I am also fed up by semi-senior community members giving me advises how devote I have to accept serious bugs and that I have to seek continieously for workarounds and that I must remember 10000000000000000  WAs for all those issues.

 

What would I be willing to do/contribute?

 

1.) Reporting bugs/issues and provide information and support for reproducing the bug scenario - > DONE!!!

2.) Buying all OFP sequels to keep the BIS boat swimming - > DONE!!!, sometimes several copies per game

3.) purchasing almost every shitty BIS DLC to keep BIS boat swimming -> DONE!!!

4.) Doing crazy crap and micro-management to make missions and gameplay work - despite all bugs -> DONE!!! I deserve a medal!

So you can see, all that was already done, what did it help? To little and far to late!

 

 

What do I expect?

 

1.) Real and transparent treatment of bugs - > just partially done -> the BIS bug tracker is IMHO just a huge pile of information not properly treated by BIS

2.) Serious attempts to overcome severe design issues in engine and function design - at the moment I rather see tweaking around to hide the worst symptoms

3.) More insight into the solution design BEFORE it is reaching a developer for coding - I have seen so many anouncements of solutions for bugs or other issues - where the real result was rather...disappointing.

 

 

What would that mean for AI driving capabilities?

 

It would mean to me that BIS should communicate more in detail HOW it works, WHICH parts of AI path finding and driving works OK and in which part they see potential for improvements.

 

The questions that MUST be asked are:

 

1.) Are the methods how the environment is made transparent to the AI routines the best way? What other methods are feasible?

2.) That each AI driver is an individual entity fighting for its own way without having higher level routines like "crowd manager"; "cross road manager"; "right of way manager"; "damn we queing up at a bridge - does it make sense to wait manager"....

3.) Review of waypoint handling: the end waypoint....should the vehicle stop there or just close to or should it just start slowing down from max-speed when reaching waypoint...

4.) Obstacle detection: review the methods and frequency the routines checking the clearance of the path - depending on speed and on known other individual AI drivers present in the closer environment.

5.) Formation handling: It is so obvious that AI react to late, drive to close and to fast, that they do not adapt distance depending on speed...all that needs a real overhaul. The tendency to follow the previous vehicle instead of just keeing order and distance and following the own path would make driving more fluent. When I order vehicles individually I get usually better results compared to group handling in formation - it would be perfect if individuals would maintain a minimum distance to the vehicle in front.

 

 

Just a few cent...

 

 

interesting thoughts

 

bottom line is Bohemia is reluctant to commit significant company resources $$$ to ArmA AI, at least not for the past many years. 

[Alwarren 2765]

4 hours ago, The Man Without Qualities said:

- a generic attitude like yours to restrict yourself in expectations and from defining the ideal solution before trying to find the optimal solution is....wrong

 

My attitude is not "generic". It's practical. I don't believe that there is going to be a general, all-out rewrite of the AI driving. So I can either try to get along with what I get, or cry myself to sleep. I chose the former. I do bitch a lot at BI minions (oukej is probably hating me already :D ) but as I said, you either accept it won't happen and try to get them to fix the issues that are practical, or you make it your job to write lengthy posts complaining and reach nothing at all, because, let's face it, it's not like BIS would read your post and say, "gosh he's right, we need to rewrite the entire AI code now. Let's get started".

 

So, what I do is point out individual problems that I think could be addressed (like convoys trying to get back to their origin when they are attacked), instead of whining how generally bad this is and how "in 15 years nothing happened".

 

And that's all I am going to say about it, I don't want to derail this thread any more.

[f2k sel 164]

One problem is with reporting issues with AI driving is that it's just so inconsistent, you can run the exact same mission multiple times and it's fine and then for no reason it falls apart with vehicles getting stuck and crashing for no reason.