Patch 4,1, AI improvements for ship behavior and the problems

[Lord Crc]

Plus i think that ships having blind spots and asymmetrical fire coverage is a good thing. Adds gameplay by giving you something to aim for, and provides a visible reason for ships to behave (or, admittedly, sometimes, fail to behave) a particular way.

I agree to a certain extent. My main issue with XR in this regard is that the AI does not (effectively) consider its own vulnerable areas/angles. Its primary focus should be on protecting important structures like engines, maximizing attack potential should be secondary.

I also think some ships have a too asymmetrical/illogical weapon placement. Others are quite good.

[RAVEN.myst]

The decision to keep big ships upright is a design decision; which is proper, i think. Design should drive implementation and not the other way around.

As a creative professional myself, I'm afraid I can't agree with that at all - I have never compromised the quality of my content by making it subordinate to its presentation (both are important, though, of course) : to do so would have been catering to the lowest common denominator by distracting with flash over substance. *shudder* Of course, that being said, there are media in which the presentation IS the content...

Plus, it's not an insurmountable problem. Just, well, tricky in some cases. Failure to do so in all cases lies, of course, with implementation. Working on it. Can't promise that i'll wake up one morning with a eureka moment for this one, though. Although the solution that is in should solve most of the cases.

Well, that is most encouraging to "hear"
"'Eureka' is Greek for 'this bath is too hot'." -The Fourth Doctor (Who)

At the end of the day, it's a game, and games are defined by rules that determine how they look and feel and, hopefully, results in a more effective game (effective, in this case, meaning that it's better in what it's trying to do, whatever that may be).

True enough.

and a "craving" on my part

Uh oh. Hope that doesn't come with expectation sparked by my expressing interest!

Hahahahah! Fear not - I tend to keep my expectations (in any context) as low as possible from the start - it cuts down on disappointment Just to put me in context: I don't even do sneak previews, or keep tabs on "developer diaries" and the like. Firstly, I am VERY anti-spoilers, and secondly, I prefer not to second-guess the artist (though I may, and often do, disagree with him/her/them AFTER the fact, heheheh) - yes, I know that some may construe this as non-participation in an audience-participation role, but I also think that despite the fact that the audience is the target market, it often doesn't actually understand what it thinks it wants, and its whims can serve to distract the artist/s and corrupt the work ("too many cooks", for one thing.) I prefer to trust during development (and criticize later, heheheheheh.)

Ultimately, for X, i think weapons physically moving on ships' hulls are one of those major things requiring work on multiple levels (asset work, design, physics, engine-level logic, AI, possibly UI) that won't really ultimately add much in the larger scheme of things. (Yes, i know that i just said that design should drive implementation. And if that decision is taken, that work would have to be put in. Would be to the detriment of other systems which are possibly more important, though. Only so many man-hours in a design cycle.)

Oh, I do understand - my "craving" is merely one of those pipe-dream fantasies that I don't necessarily expect to come true, at least not any time soon (though if surprised in that regard, I certainly won't be disappointed - because I wasn't counting on it in the first place heheh)
Plus i think that ships having blind spots and asymmetrical fire coverage is a good thing. Adds gameplay by giving you something to aim for, and provides a visible reason for ships to behave (or, admittedly, sometimes, fail to behave) a particular way.[/quote]

[j.harshaw]

The decision to keep big ships upright is a design decision; which is proper, i think. Design should drive implementation and not the other way around.

As a creative professional myself, I'm afraid I can't agree with that at all - I have never compromised the quality of my content by making it subordinate to its presentation: to do so would have been catering to the lowest common denominator by distracting with flash over substance. *shudder* Of course, that being said, there are media in which the presentation IS the content...

Just want to correct what appears to be a misunderstanding. By design, i don't mean how a thing is made to look, but what a thing is made to be. Implementation is then how that being is brought about.

Sorry if that comes off as being pedantic (not even sure if that's correct usage of the terms!), but i didn't want to give the impression that i'm of the opinion that substance should in any way be secondary to appearance.

My main issue with XR in this regard is that the AI does not (effectively) consider its own vulnerable areas/angles.

True. Man, would love to have more time for this.

Its primary focus should be on protecting important structures like engines, maximizing attack potential should be secondary.

That, i'm not so sure of anymore. Yes, self-preservation first and foremost makes sense. BUT it appears to be heavily criticized at points, and can translate to bad gameplay. (See complaints regarding AI ships fleeing pre-4.10). To draw on past games, i think the extremes are the old Close Combat games where a squad of infantry would outright refuse a command to charge an entrenched tank unless morale were high enough vs Command and Conquer where a unit told to go there just goes there even if it means getting blown to bits. Still trying to find that sunny middle which we'd be happy with, or at least, something that would result in something we'd be happy with.

Anyway, @all, always looking forward to your feedback, critical or otherwise. XR, and, as it's shaping up, X4, are huge games and, as someone who played X3 and XR before joining the team, i'm entirely conscious that there are LOTS of scenarios that simply aren't apparent in development (and are subsequently hard to see playing the game during/after development since we're looking at it having gone through development), and seeing the thing from other eyes is always helpful.

@dhomstr, we, er, appear to have hijacked your thread. Sorry about that. If you want, i could transfer all of this following discussion to a different thread. (At least i think i could. Not sure if i have moderator privileges in this channel.)

[RAVEN.myst]

The decision to keep big ships upright is a design decision; which is proper, i think. Design should drive implementation and not the other way around.

As a creative professional myself, I'm afraid I can't agree with that at all - I have never compromised the quality of my content by making it subordinate to its presentation: to do so would have been catering to the lowest common denominator by distracting with flash over substance. *shudder* Of course, that being said, there are media in which the presentation IS the content...

Just want to correct what appears to be a misunderstanding. By design, i don't mean how a thing is made to look, but what a thing is made to be. Implementation is then how that being is brought about.

Sorry if that comes off as being pedantic (not even sure if that's correct usage of the terms!), but i didn't want to give the impression that i'm of the opinion that substance should in any way be secondary to appearance.

Ah OK, I DID misunderstand, then. (And also, as per my belated edit, although it turns it's not the point of discussion, I don't mean that presentation ought to be dismissed either - it IS also important, of course.)

My main issue with XR in this regard is that the AI does not (effectively) consider its own vulnerable areas/angles.

True. Man, would love to have more time for this.

Not a trivial task, for sure!

Its primary focus should be on protecting important structures like engines, maximizing attack potential should be secondary.

That, i'm not so sure of anymore. Yes, self-preservation first and foremost makes sense.

Then again, there's the aspect of "offense is the best defense" - if a unit is sufficiently effective and fast about how it dispatches its enemies, that forestalls its own vulnerability (this applies, of course, to "glass cannons" and snipers more than to brawlers and heavy assault types.) I think that whether a unit prioritizes its own defense or its attack potential has to be determined by the unit's role - "tanking" units would opt for optimizing their resilience, but damage-dealers... well

Anyway, @all, always looking forward to your feedback, critical or otherwise. XR, and, as it's shaping up, X4, are huge games and, as someone who played X3 and XR before joining the team, i'm entirely conscious that there are LOTS of scenarios that simply aren't apparent in development (and are subsequently hard to see playing the game during/after development since we're looking at it having gone through development), and seeing the thing from other eyes is always helpful.

Yes, tunnel vision and blinkers obviously happen from various perspectives - a developer's vision will gradually and unavoidably narrow based on problem-solving challenges being concentrated on, a player's vision will narrow based on his/her own style of play, and an outside observer might see things altogether differently to either.

Good discussion, and highly responsive - thank you!

[dholmstr]

I don't mind the highjacking. I actually had hoped for a discussion. While yes it went abit off it is a good one. But could we get abit back to the problem about the AI at this point at this XR

[Lord Crc]

True. Man, would love to have more time for this.

Boy do I know that feeling!

Its primary focus should be on protecting important structures like engines, maximizing attack potential should be secondary.

That, i'm not so sure of anymore. Yes, self-preservation first and foremost makes sense. BUT it appears to be heavily criticized at points, and can translate to bad gameplay. (See complaints regarding AI ships fleeing pre-4.10).

Yes, it's definitely a balancing act. What I meant is that if you make an Arawn attack say a station, its main defense against the Targon Tracers is movement. If it lets the engines get taken out right away, it'll be destroyed swiftly. So I'm not saying it should necessarily fly away and not get hit, but it should first and foremost try to move in a way which protects the engines.

Presenting the best angle for attack comes secondary to that, because there's no point in being at a good attacking angle if it has been blow up. If it's attacking a group of fighters then protecting the engines is not so much of an issue, as long as no other enemy capships or stations are nearby of course.

Of course this is clearly a point where captain skill could factor in. Send a one star captain into a tricky battle and expect tears.

Anyway, @all, always looking forward to your feedback, critical or otherwise.

One thing I love about Egosoft is that you devs are actively engaging the community. It's rare and highly appreciated, even if you don't always say the things I want to hear

I'm also fully aware that AI coding is a highly complex subject with a lot of edge cases to consider and balance. I've done some "core wars"-like coding, which is a much easier task than this, so I appreciate the non-triviality of the matter.

As mentioned earlier, do I think making more distinct separation of anti-fighter and anti-capship weapons, where the anti-capship weapons have longer range (~20km effective range) would make it easier on the AI for cap vs cap battles as it would allow for more room around each ship.

I also think it would feel more appropriate, and allow for more distinct roles of the ships.

Thanks again!

[ZaphodBeeblebrox]

My thoughts on this are:

There are too many players with too many expectations.

No matter how good a programmer you are, you will never meet everybody's needs.

There needs to be more player controlled variables.

So that each player can have some input on ship AI behaviour.

For example: I want my captain to boost away / jump away immediately his shields / hull hit a a certain %age.

Trying to anticipate every eventuality in code just ain't going to work.

Yes factor in the skills of the crew. But please let the players have some
control over AI behaviour.

[Snafu_X3]

The ships aren't orienting correctly to use their firepower effectively;

The decision to keep big ships upright is a design decision; which is proper, i think.

Umm.. except when it hinders their manoeuvrability (tactical & strategic)? In such cases (ie IZ fights) you /need/ those ships to respond to orders effectively & efficiently while sustaining minimal dmg. It's less important while transiting/patrolling or taking place in combat OOZ, but IZ it's critical that ships respond effectively to <player's> orders IMO.. otherwise, why have them respond to orders at all?

it's not an insurmountable problem. Just, well, tricky in some cases. Failure to do so in all cases lies, of course, with implementation. Working on it [...] Although the solution that is in should solve most of the cases.

Thanks! Looking forward to it

"'Eureka' is Greek for 'this bath is too hot'." -The Fourth Doctor (Who)

Nah. It's Greek for "I need a towel!" (Terry Pratchett - Small Gods)

Ultimately, for X, i think weapons physically moving on ships' hulls are one of those major things requiring work on multiple levels (asset work, design, physics, engine-level logic, AI, possibly UI) that won't really ultimately add much in the larger scheme of things.

I agree to a large extent (although it would look cool IMO ) There are a couple of other things you missed too, such as ship balance WRT weight/mass distribution (/really/ important in space, but even c18 wooden hulled warships had problems when firing full broadsides*, hence one reason why the hulls became bigger & less manoeuvrable.. leading to huge 'ships of the line' that weren't really very useful except in pitched set-piece battles due to their lack of manoeuvrability

*Not due to recoil directly (from the mass of shot discharged), but from the sudden massy wep (cannon) shift away from centre-of-mass of the ship (it's a speculated reason as to why the Mary Rose, Henry VIII's flagship, capsized). Set piece fire options required a ballast shift prior to engagement to make the optimal hit ratio; this would work in disfavour once manoeuvring (of the enemy) started. Nelson's charge thru the centre of the enemy wall (a bold new tactic for the time) partially offset this, along with the ability to fire both broadsides at once

[RAVEN.myst]

My thoughts on this are:

There are too many players with too many expectations.

+1 (even if I may be one of them)

No matter how good a programmer you are, you will never meet everybody's needs.

+1

There needs to be more player controlled variables.

So that each player can have some input on ship AI behaviour.

+1 and +1

For example: I want my captain to boost away / jump away immediately his shields / hull hit a a certain %age.

Indeed - perhaps sliders or selection radio buttons (Dark Reign did this extremely well, a looooooong time ago, by allowing players to tweak variables such as aggression level (self-explanatory) and damage tolerance (how little/much damage before triggering a 'return to repair bay' task), along with behaviour-setting stances such as "scout", "harass" (hit once then withdraw - like 'scout' but with a bit of a bite), and "search and destroy", in addition to more conventional ones such as "hold ground" or "defend area".)

Trying to anticipate every eventuality in code just ain't going to work.

+1

Yes factor in the skills of the crew. But please let the players have some
control over AI behaviour.

+1

[UniTrader]

Yes factor in the skills of the crew. But please let the players have some
control over AI behaviour.

+1

maybe like the Skills determine the range/possible options for these Settings (eg a Manager with high navigation can order trades to farther away objects, but he is still bound to the Players Settings - which are just less restrictive with higher Skill
or the Morale of a Captain decides the Range of which the Flee behavior can be set: for 0 Stars he practically flees when either seeing red dots or maybe after the first hit (range for the Player Setting) - with max morale the Range for this goes from Shields at 47% over Shields Down, light damage, moderate damage, critical damage to fight to the death* - and the morale-values in between allow settings between those Ranges)

*i know there is no overlap in these examples - the min enemy contact setting also rises with higher morale, so it makes sense to not just get the employees with the highest Skills in everything, though their Setting Range is bigger, in this case their pride gets in the way

[Nikola515]

What we also need is improved drone AI as well.... Main problem is what I have is that they don't stay clear of exploding ships. It is too expensive to use them at this point..... Also Skunk drones are somewhat slow in returning so if possible time should be shortened. I understand that they shouldn't just teleport but it is boring gameplay especially when you are on constant move.

[RAVEN.myst]

Yes factor in the skills of the crew. But please let the players have some
control over AI behaviour.

+1

maybe like the Skills determine the range/possible options for these Settings (eg a Manager with high navigation can order trades to farther away objects, but he is still bound to the Players Settings - which are just less restrictive with higher Skill
or the Morale of a Captain decides the Range of which the Flee behavior can be set: for 0 Stars he practically flees when either seeing red dots or maybe after the first hit (range for the Player Setting) - with max morale the Range for this goes from Shields at 47% over Shields Down, light damage, moderate damage, critical damage to fight to the death* - and the morale-values in between allow settings between those Ranges)

*i know there is no overlap in these examples - the min enemy contact setting also rises with higher morale, so it makes sense to not just get the employees with the highest Skills in everything, though their Setting Range is bigger, in this case their pride gets in the way

Yes, I proposed something similar a while back, where different numbers of 'stars' would unlock various additional controls - they'd be greyed out if the crewmember's skill isn't high enough - 5 stars would have everything available. And as you suggest in your good example, ranges of variables could be opened up with higher skill. It would also make for a very clear-cut scaling of crew skills, very transparent and obvious. An additional advantage of this would be that the player could then decide: "does this captain/station manager/whatever NEED 5 stars in this particular skill, or can I hire this lesser-skilled applicant instead?" The player could then allocate staff with greater precision, theoretically.

[RAVEN.myst]

What we also need is improved drone AI as well.... Main problem is what I have is that they don't stay clear of exploding ships. It is too expensive to use them at this point..... Also Skunk drones are somewhat slow in returning so if possible time should be shortened. I understand that they shouldn't just teleport but it is boring gameplay especially when you are on constant move.

Agreed - very much so! In fact, it should also be possible to control drone AI to some extent: DOs need to be instructed whether to use or not use drones by default, and forced launch and recall orders need to be available; also, drone bahaviour, even if fairly broad: for example, it would be useful to be able to set drones to "protect home ship" - that way, they will stay close and engage incoming fighters, but not go out far; a "seek and destroy" mode would have them be more aggressive and go out looking for targets - in this case, they might get caught in the blast of exploding ships, but that's the cost/benefit decision that the player gets to make. At the moment, drones do their own inscrutable (and often stupid) things, without any control from the player, and often insufficient predictability.

[UniTrader]

.....
Agreed - very much so! In fact, it should also be possible to control drone AI to some extent: DOs need to be instructed whether to use or not use drones by default, and forced launch and recall orders need to be available; also, drone bahaviour, even if fairly broad: for example, it would be useful to be able to set drones to "protect home ship" - that way, they will stay close and engage incoming fighters, but not go out far; a "seek and destroy" mode would have them be more aggressive and go out looking for targets - in this case, they might get caught in the blast of exploding ships, but that's the cost/benefit decision that the player gets to make. At the moment, drones do their own inscrutable (and often stupid) things, without any control from the player, and often insufficient predictability.

my own approach to this is to make drones just plain stupid. they can only perform one Task, which is directly given by the DO*, and otherwise just fly in formation around the mothership. They will never decide something themselves (except blowing up or returning to their Ship if their Command structure fails)

*Currently implemented in my DO Script are:
Missile Defense (iirc for Intrepid Drones) and Figher Intercept (for Interceptor Drones)
Planned for this are:
Bombard Capital (Overrun drones are the ones planned for this i think) and the possibility to give support for other Ships than the Mother Ship (with the previously mentoined options, but controlled by the Pilot, not the DO) - they will return to their mothership though..

[dholmstr]

While this is an old discussion, it still seems valid. So I'll continue my "quest" finding answers and problems alike. With the VR version coming and maybe a 4.2 on the way, is this being looked upon for that release? Minor tweeks and upgrades for the AI and shipsmovement etc. ?

[j.harshaw]

Not at the moment. Focused on X4 development and support-work for XRVR. There is one thing in particular that i'm rather keen to get feedback on, but am saving that for non-VR 4.20 beta if and when.

But by all means, keep discussing. The fact that an idea won't be in the next update doesn't mean it won't be considered for eventual inclusion somewhere in some shape or form. Plus just throwing around ideas is good.

[ZaphodBeeblebrox]

My take on this is:

No matter how good a programmer you are, or how well the AI performs, you will never satisfy everybody. Most of the gripes on this forum over AI can be boiled down to this simple statement.

"It doesn't do what I expected it to do, therefore, it is a bad AI."

The player needs to have greater control over the parameters that control AI behaviour. Exposing parameters as controls will allow the player to adjust the AI so that its actions are closer to what each individual wants.

This will make the AI "better." Also exposing these parameters so that modders can build controls for them would give even greater flexibility.

[RAVEN.myst]

Most of the gripes on this forum over AI can be boiled down to this simple statement.

"It doesn't do what I expected it to do, therefore, it is a bad AI."

The player needs to have greater control over the parameters that control AI behaviour. Exposing parameters as controls will allow the player to adjust the AI so that its actions are closer to what each individual wants.

This will make the AI "better."

I think this is an astute observation, and agree fully. While "more casual players" may want a generic "turn it on and it simply works" system (automatic transmission in a car, for example), the "power users" prefer to have the means to apply their own ingenuity through more control (manual transmission), and modders want the ability to tweak things even more (actually get under the hood) both for themselves and in the interests of what they build for others.

I think, in fact, that investing some development into greater control options for AI behaviour could result in less head-scratching as far as trying to define "optimal" behaviours (of course, it won't eliminate it, as you still need the "optimal" defaults for those players who don't want the detailed control.) In light of the numbers of ships involved (assuming precedent from previous titles), I would also like to propose the inclusion (I know, this may be a bit of work, but bear with me) of a way to save a few behavioral templates (or at least, define a few, and allow them to be edited by the more advanced players), so that a ship can be assigned a customised role with just a few clicks or whatever, so that an intricate setup need not be repeatedly (and potentially erroneously) reapplied.

[Nanook]

The simplest thing to do would be giving the player override control on any AI actions, and not let the AI revert to the default (bad?) behaviour unless the player so chooses. For example, let the player assume helm control on capital ships when the captain starts doing something stupid. Another would be letting the player set factory freighters to a certain task when the manager fails to do so, such as gathering/selling specific wares.