My enthusiasm is in Maintenance Mode.
NPC SRV coming with 4.0?
- Thread starter Ian Skippy
- Start date
As we are talking about a hobby here, when that happens to me I go do something else I enjoy instead.
Better than them not doing it ¯\(ツ)/¯
To me it seems they maybe want our NPC crew member to be able to drive our SLF and also SRV around on planets.
Perhaps they want to generate scenarios on the planet surface which require co-op support.
Perhaps they want to generate scenarios on the planet surface which require co-op support.
That’s what I’m doing. Among those things: chatting with folks about how about how amazing the game easily could be, if only the devs were willing to make it so.
I have that same convo about the world in general.
Or maybe they want the NPC security ships that protect planetary installations to be able to launch SRV's so the security commanders can drive to our landed ships and wait for our return to great us with refreshments!
As if my statement were supposed to be the beginning and the end of the conversation as opposed to a reference to an entire Russian novel’s worth of back and forth exploring this very topic. As if the path forward were some abstract ineffable unknown as opposed to well trod and well understood territory outlined explicitly and publicly by the devs themselves before they even made the game.Yeah, if only devs were willing to take the easy option of making the game amazing.
Well and truly the only possible explanation for my or anyone’s displeasure is that We Just Cannot Be Pleased, that Frontier’s plight is an intractable lose-lose-lose situation; Sisyphus by way of Franz Kafka with a touch of Terry Gilliam. That’s what’s happening. They’re not going off the rails ignoring their promises, their audience feedback, and (probably I hope) their better instincts; it’s just that everything is subjective and every subject is a jerk. OK fine keep telling yourself (and everyone else) that, I’m sure FDev are doing the same, so at least you’re in good company, and it’s clearly working out fine for them.
I can't think why they don't listen to you Kao
Oh yes I can, it's because you say stuff like this:
You're clearly classifying this in the 'easy' box. And yet if I were to ask you an easy question like: Name a game with AI vehicles on proc gen height map surfaces, you'd struggle with an answer.
Wanna have a wild guess at why that might be?
Yeah that's right, it's because it's not actually easy...
(There was a great non-FDev dev post on the subject on Reddit a few years back, which unfortunately I can no longer find, but the TLDR is: Pathing is a pain without the manually placed points of reference you get in a designed environment. Doubly so when you're dealing with knitted-together undulating surfaces rather than the relatively flat and squared off environment in a more favourable proc gen environment. Hence FDev went for the 'easy' fudge of floating skimmers).
Most of us have our gears ground up about 'claims vs delivery' & speed of the same with ED. The level of feedback you're displaying in this thread alone is next to useless though. It's just not the constructive crit you seem to think it is.
Last edited:
Hey ZW, have you got any insight into the requirements for vehicle AI on proc gen height maps etc? I know it's been a while but would be interesting to get your take. (IE was it considered more difficult during your dev years / is it something you'd foresee being trickier than standard bespoke locations etc?)
I accept there's a risk reading too much into these FD Careers ads, but this is interesting and relevant:
Concept Artist
Concept Artist
Frontier is the studio behind Jurassic World Evolution, Elite Dangerous and Planet Coaster. We are Britain’s biggest independent publisher and developer of videogames, based among the world-leading technology cluster in the historic city of Cambridge, England.
We are proud to be named one of the UK’s Best Places to Work by GamesIndustry.biz.
Elite Dangerous brings the player a vast host of futuristic environments and hardware, which must not only follow the iconic styling already in place but also broaden the Elite visual canon ready for technological developments coming our way in the near future.
Desirable
And for all the Armchair Devs out there ..... still a few days to apply for:
"We are looking for an experienced Game Designer to help design exciting new content for Elite Dangerous. "
Concept Artist
Concept Artist
Frontier is the studio behind Jurassic World Evolution, Elite Dangerous and Planet Coaster. We are Britain’s biggest independent publisher and developer of videogames, based among the world-leading technology cluster in the historic city of Cambridge, England.
We are proud to be named one of the UK’s Best Places to Work by GamesIndustry.biz.
Elite Dangerous brings the player a vast host of futuristic environments and hardware, which must not only follow the iconic styling already in place but also broaden the Elite visual canon ready for technological developments coming our way in the near future.
Desirable
- Contemporary mechanical, vehicle or product design training
- Graphic design skills
- Concept art skills outside of hard-surface
- Keen interest in games development. Would be useful if you could match your aspiration for the concept to the requirements of the game
- Ability to convey realistic surfaces including the narrative of a vehicle through targeted wear and damage
And for all the Armchair Devs out there ..... still a few days to apply for:
"We are looking for an experienced Game Designer to help design exciting new content for Elite Dangerous. "
Well that reply was a better effort than than,
I'll give you that.
The game could be better, much better. I mean it's not finished for a start and I doubt you'd find anyone disagreing with the idea. But as soon as you reduce that complex, much had conversation down to lazy comments about it being "easily better" but the devs "aren't willing" to make it so, then you're going to get called out for spouting drivel.
I don't know if I many any insights on that. I'd imagine the biggest challenge is not knowing what the landscape is going to be, we've seen some pretty funky cliffs and ravines, often borderline impossible. Stopping them from ending up in some stupid situation where they are plummeting off like lemmings is probably going to be as annoying as hell.
The height field maps I worked with were very predictable, most games use markers and routes for NPC's to move around known paths. With procgen, there's going to be the world 'generated' around where the player is and then there's the rest of the planet that only exists as a formula - that's got to be a nightmare for path-finding. Maybe you have a unit and you need it to move somewhere hundreds of miles in that direction, how do you map a path there? Go around a mountain range or a canyon? With any other crafted map based game, you know it's there in advance. Sure you have the 'rules' for generating that bit of the map, but simply knowing the "height" at one point isn't going to help you, it's all relative to the area around you, is there a sudden dip over that way? Is there a mountain that way? etc.
You could "look ahead" to find paths - but that's an expensive operation, pick some random point and "generate" the landscape to see what obstacles are there in advance. Sure that can be done maybe offline for all the maps that have AI units, maybe render some points selected at random until you find some "good" paths. The landscape doesn't change, so some pre-baked paths can be loaded from the server for each planet. Another possibility is to use the rule-set to generate a low-res map or the entire planet, they already do that for the zoomed out map view. So you could write some routines to generate paths from that data. Maybe you could use something more abstract like a neural net to learn how to place paths based on that information - using relative high and low points as 'avoidance' areas. That would be neat because you could factor in gravity as an input, so on some worlds the units can do the crazy canyon jump while on others they learn to stick to flat safe "low-grav friendly" routes - that's the kind of thing I'd be looking at - but then I did my AI degree at the end of the nineties
As with most thing in games the real trick is to fool the player into thinking that these things have happened rather than implementing it all in advance (imagine an online pet that loads into memory when you log on, did it do anything while you weren't playing, or did it just calculate the date since you weren't there and present the world as if it had been there all along? The later is way less expensive to implement) I'd imagine the trick is to fool the player into believing the units are moving between two viable locations. They will probably spawn in within an instance, but how far can the player follow their tracks? You might have to render the 'track' path back to a viable origin. What if the player follows the units themselves, they can't just go off randomly until they get stuck somewhere. Then there is the player interaction behaviour to think about, what happens when they are attacked - maybe that's not going to be massively different from what happens with the existing NPC ship combat but coupled with some of the above challenges.
Last edited:
I don't know if I many any insights on that. I'd imagine the biggest challenge is not knowing what the landscape is going to be, we've seen some pretty funky cliffs and ravines, often borderline impossible. Stopping them from ending up in some stupid situation where they are plummeting off like lemmings is probably going to be as annoying as hell.
The height field maps I worked with were very predictable, most games use markers and routes for NPC's to move around known paths. With procgen, there's going to be the world 'generated' around where the player is and then there's the rest of the planet that only exists as a formula - that's got to be a nightmare for path-finding. Maybe you have a unit and you need it to move somewhere hundreds of miles in that direction, how do you map a path there? Go around a mountain range or a canyon? With any other crafted map based game, you know it's there in advance. Sure you have the 'rules' for generating that bit of the map, but simply knowing the "height" at one point isn't going to help you, it's all relative to the area around you, is there a sudden dip over that way? Is there a mountain that way? etc.
You could "look ahead" to find paths - but that's an expensive operation, pick some random point and "generate" the landscape to see what obstacles are there in advance. Sure that can be done maybe offline for all the maps that have AI units, maybe render some points selected at random until you find some "good" paths. The landscape doesn't change, so some pre-baked paths can be loaded from the server for each planet. Another possibility is to use the rule-set to generate a low-res map or the entire planet, they already do that for the zoomed out map view. So you could write some routines to generate paths from that data. Maybe you could use something more abstract like a neural net to learn how to place paths based on that information - using relative high and low points as 'avoidance' areas. That would be neat because you could factor in gravity as an input, so on some worlds the units can do the crazy canyon jump while on others they learn to stick to flat safe "low-grav friendly" routes - that's the kind of thing I'd be looking at - but then I did my AI degree at the end of the nineties
As with most thing in games the real trick is to fool the player into thinking that these things have happened rather than implementing it all in advance (imagine an online pet that loads into memory when you log on, did it do anything while you weren't playing, or did it just calculate the date since you weren't there and present the world as if it had been there all along? The later is way less expensive to implement) I'd imagine the trick is to fool the player into believing the units are moving between two viable locations. They will probably spawn in within an instance, but how far can the player follow their tracks? You might have to render the 'track' path back to a viable origin. What if the player follows the units themselves, they can't just go off randomly until they get stuck somewhere. Then there is the player interaction behaviour to think about, what happens when they are attacked - maybe that's not going to be massively different from what happens with the existing NPC ship combat but coupled with some of the above challenges.
Thanks, really interesting post!
The baked-in paths definitely sound like a feasible way to get trade routes going between bases etc.
The job spec is pitching high with the "understands and explores the surfaces of procedurally generated planets" bit though for sure. Definitely sounds like they want 'drive anywhere' potential. A neural net approach would def be rad, so long as they end up with scarab spawn that stay upright
I do wonder how they'll deal with PvE stuff too. Taking cover, pursuit behaviours etc. Lotta aspects there if they're out in the wilds it seems.
(PS were neural nets of practical use in 90s game dev? Or was that more on how you'd do it with current tools?)
Cheers for the reply
Last edited:
Or the person who has been doing that or a similar job for a number of years has left and they're looking for a replacement?