GT Sport - Trailers, Videos and Screenshots

  • Thread starter sk8er913
  • 17,667 comments
  • 2,106,599 views
An adaptive AI script on animations? I don't think that's as easy as you think it might be.
How so? All the game has to learn is the way I downshift and upshift on a given track with a given car. Downshifting from high speed into a corner will be one pattern, downshifting from medium speeds between two corners will be another etc. All the game has to learn is when to take the virtual hand off the gear lever or fingers off the paddle and when to leave it/them there. Sounds pretty simple to me.
 
And yet there's not a single game on the market that does it. Yeah.. must be simple.
Perhaps its because devs feel the time and cpu/gpu access it requires isnt worth the effort, rather than it being difficult to implement.
It reminds me of (a bit off topic) Hideo Kojima in MGS 2 making of dvd said the tanker/rain part of the game, they could have utilized real time drops of rain falling onto snake, but the effort wasnt worth it as it could be "faked" by having the out door rain model of snake have rain splashes as a part of the character model itself.
It let them save cpu/gpu resources in the process.
 
If that was the case, then why is it not available in PC games, even as a enable/disable option?

Oh, I know why.. most of those games don't even feature arms in the cockpit, and have the option to disable wheel on screen completely. Surely if it was so easy, they wouldn't be disabling aspects of the display like that as an option.
 
And yet there's not a single game on the market that does it. Yeah.. must be simple.

If that was the case, then why is it not available in PC games, even as a enable/disable option?

Oh, I know why.. most of those games don't even feature arms in the cockpit, and have the option to disable wheel on screen completely. Surely if it was so easy, they wouldn't be disabling aspects of the display like that as an option.
Didn't say it was simple, I just don't think it's as hard as most other parts of the game. No one had real time dynamic weather until a few months ago either. Name any feature and at one time it didn't exist in any racing game so that's a total non-sequitur . Seems a no-brainer to me that no one has tried adaptive AI script on VR because no VR headset is commercially available yet or in widespread use. Once the technology takes off on the consumer end, there will be a ton of effort put into making the experience as real as possible. That's not to say this feature will ever happen, but you can bet the pressure will be from the community and sooner or later someone will take the bull by the horns and work something out, IMO. Pretty sure Facebook didn't throw $2billion at it to have the technology remain stagnant after release. It's pretty clear the the future of gaming is moving towards a total 360 degree, virtual immersion into the game world, this type of feature would seem a logical advancement in racing simulations.
 
It'll be a long time before 1:1 movements are there with VR and what's seen on-screen. I don't see it happening at all unless they put out some kind of mocap gloves for players that are linked to the game and it's input.
 
...VR, eh? Yeah why not. As long as it doesn't turn into a SAO situation.

However, I can see that doing this in public might not be the best way to meet new people...

oculus_rift_consumer-6.jpg
 
It's only a novelty as of now because no one has had their hands on a retail version with full support of titles yet. I'm looking forward to it. The ability to freely look around in cockpits of both cars and planes (I play War Thunder on the regular on PS4) is about the only use I'd get out of it. My interest in shooters has gone away completely at this point.
It's the same argument as the LED TV vs curved LED TV.
No, no it's not.
 
It's only a novelty as of now because no one has had their hands on a retail version with full support of titles yet. I'm looking forward to it. The ability to freely look around in cockpits of both cars and planes (I play War Thunder on the regular on PS4) is about the only use I'd get out of it. My interest in shooters has gone away completely at this point.

No, no it's not.
That's what I'm meaning. Once they start hitting the mass market they will develop a lot faster and be more and more of a necessity.
 
It'll be a long time before 1:1 movements are there with VR and what's seen on-screen. I don't see it happening at all unless they put out some kind of mocap gloves for players that are linked to the game and it's input.
Input lag on your typical widescreen tv of 30-50 ms will prevent 1:1 from being a reality for most of us. With really low lag gaming monitors you can get much closer. But that's a different point altogether. For me the biggest issue is in shifting and I think that's a relatively easy problem to solve for the most part. All that's really required is for the game to track your shifting patterns on a typical flying lap and when the gap between shifts is less than "X", the hand is not removed from the shift knob, or fingers from the paddles. It will never be an exact match because laps are never identical, but it should be really close the vast majority of the time. If we could program a number ourselves and fiddle with it, that would be even better.

It's the same argument as the LED TV vs curved LED TV.
27" CRT tv vs. 50" widescreen is more like it.
 
VR headsets are cool, but they still have a significant amount they need to approve. It would be interesting to do a few laps with one.
 
How so? All the game has to learn is the way I downshift and upshift on a given track with a given car. Downshifting from high speed into a corner will be one pattern, downshifting from medium speeds between two corners will be another etc. All the game has to learn is when to take the virtual hand off the gear lever or fingers off the paddle and when to leave it/them there. Sounds pretty simple to me.

Didn't say it was simple, I just don't think it's as hard as most other parts of the game.

Hah.

AI routines not only are costly in terms of processing (for the CPU) as @KilzoneStrife wondered, increasing complexity increases error.
So having it learn the player on a enormous amount of variables (in just one corner you'd have speed of approach, timing of braking, player position in relation to the track, to the other cars, window for gear shifting and so on) just to have the animation look slightly better timed than what they already present would not be worth the expenses.


VR headsets are cool, but they still have a significant amount they need to approve. It would be interesting to do a few laps with one.

It's better than TV because it reduces the amount of information processed by your brain into just the subject they want to present, but yes most sets still have problems with FoV of screens, refresh rate, scuba goggles effect of seeing the frames and well seeing the individual pixels (pixel density).
But still, the amount of noise (the outside of your TV set and surrounding) reduction and addition of depth perception makes it the best for first person simulation.
I'd guess in 5yrs we should be seeing sets that fix most of the issues above and probably a few others like weight and maybe eye tracking (so we don't have to unnaturally move our necks a lot just to see).
 
Hah.

AI routines not only are costly in terms of processing (for the CPU) as @KilzoneStrife wondered, increasing complexity increases error.
So having it learn the player on a enormous amount of variables (in just one corner you'd have speed of approach, timing of braking, player position in relation to the track, to the other cars, window for gear shifting and so on) just to have the animation look slightly better timed than what they already present would not be worth the expenses.
I don't think it's that complicated nor do I think it needs to be done in real time. I'd set it up as a learning process you turn on and off or you can simply pre-program a given time between your downshifts and upshifts before taking your hand off the knob. Most cars have short first gears for example. You could simply tell the game, "keep my hand on the shifter until I select second gear" and it would be right just about all the time. Likewise, your shifting time between the taller gears is usually long, so you could just tell the game, keep my hand on the shifter when downshifting at gear X or below but above X return to the wheel. The animation is already there, there is already some kind of algorithm running in the background for it work. I don't see this as having to be overly complicated to come much closer to realism.
 
I don't think it's that complicated nor do I think it needs to be done in real time. I'd set it up as a learning process you turn on and off or you can simply pre-program a given time between your downshifts and upshifts before taking your hand off the knob. Most cars have short first gears for example. You could simply tell the game, "keep my hand on the shifter until I select second gear" and it would be right just about all the time. Likewise, your shifting time between the taller gears is usually long, so you could just tell the game, keep my hand on the shifter when downshifting at gear X or below but above X return to the wheel. The animation is already there, there is already some kind of algorithm running in the background for it work. I don't see this as having to be overly complicated to come much closer to realism.

That's what games are already doing.
They have breaking zones mapped and intended speed/gear. When you enter the mapped area and start breaking he switches a hand to the shifter so it can perform the shifting animation in case you shift.

What happens when you have one car in front of you controlled by another human player? You don't know if he'll try to shoot for a late break point or maybe an early one and you'll have to swerve. Add another car behind you and then that shifting point becomes another again, and so on.
Increasing the number of variables to take into account to just to gain a hundred milliseconds in execution when most players will be: a) focused on the road b) using a controller or set with only paddle shift c) probably with different hand positions on the wheel already, seems like a lot for too little.

If/when player insertion becomes an issue to be addressed, I figure it'll most likely be done with improved motion sensoring or capturing if costs (production and computing) diminish. That's why most VR controllers are "wands" already.
 
Sony about PGW:
"With the changes this year of E3 taking place a week later and Gamescom taking place earlier, we feel that both PlayStation and the wider industry would benefit from a major news milestone later in the year, during the most important period for gaming."
Can I have the source pls? Thx
 

Latest Posts

Back