Reply
Fender Bender
Registered: 12/04/2007
Offline
4305 posts
 

Re: GT5's poor collision physics and AI

Sep 29, 2009
Good analysis bowler but will it be enough?
Message 81 of 121 (4 Views)
Reply
0 Likes
I Only Post Everything
Registered: 08/14/2009
Offline
1080 posts
 

Re: GT5's poor collision physics and AI

Sep 29, 2009

Kenshin71 wrote:

SWERV_GRIFFIN wrote:

 

Ease up on the attitude, you are coming off as very arrogant and argumentative. YOU cannot control what responses are to a thread that you apparently think you own. Show some respect around here, or you will never gain any from the members of this forum. That's some Real Talk. Regards, SWERV


 

I'm not here to gain respect from the members of this forum though. Unfortunately no-one on this forum has any influence on the game at all so it's all academic. I don't go in other threads and derail them or ignore the topic, so I hope others can have the same courtesy.

 

Actually I was hoping there was someone with channels to PD on this board who could maybe pass on the message. Still, I'm vainly hoping someone from SCE might stumble on the thread and I hope other concerned players will add their voice. If people don't want to talk about collision physics and AI and instead go off topic there are plenty of other threads to do it in.

 


i've discussed the collision physics and AI issues several times here in this thread. at no point have you even acknowledged or replied to these points. so unless you have some input of your own on this matter how do you plann to have any 'discussion' about it. all i've read from you on the matter is your OP, since then you seem more interested in moderating the thread rather than contrubuting to it.

Message Edited by meatbag_ on 09-29-2009 12:59 PM
Photobucket
Message 82 of 121 (4 Views)
Reply
0 Likes
Sackboy
Registered: 07/16/2007
Offline
479 posts
 

Re: GT5's poor collision physics and AI

Sep 29, 2009

jlbowler wrote:

Kenshin71 wrote:

Having played GT5P and seen some youtube videos of GT5 I was a bit concerned about the poor collision physics in the game.

 

A good example is here:

 

http://www.youtube.com/watch?v=6Ksfo2L_65U

 

The Ferrari is just bouncing off walls like a bumper car. I don't care so much about the damage modelling, the superficial stuff is fine, but the collision physics are laughable. No sense of impact or momentum. I also saw a video where a bunch of cars just ploughed into a stationary car, none took evasive action. Has anyone ever raised this with Yamauchi or is he untouchable? I wonder if PD even know what sort of comments/criticism their game gets because they really need to lift their game.


I'll discuss the collision physics and debate your points with you.  You say that there is no sense of impact or momentum.  From what is seen in the video, it is very difficult to see if there is any sense of impact.  This would be better view from an actual driving perspective instead of a replay that shows the action from various angles.  As far as the momentum goes, the first wall impact at :03 slows the car from 233 km/h to 168 km/h.  ( I have no doubt that this would have destroyed the car in real life but we're dealing with a demo of a game and the point is for the player to get their full time in).  The car then bounces on the rumble strips and struggles to accelerate ( I can't say if the struggle to accelerate is based on damage incurred, the strips, or the player just not accelerating).  The second wall impact at :09 takes the car from 198 km/h down to 158 km/h.  The trip through the plastic barriers slows the car all the way through the series of impacts.  The car then bounces over the rumble strips again (you can see the car pitching side to side as it goes over the strips.)  The car is then slowed while going through plastic barriers again.  At approx. :45 you can see the car travel over the rumble strips and the two right side tires are off the ground.  This was either caused by the impact with the wall or the rumble strips.  Either way, this is a sense of impact affecting the car.

 

For you, or anyone, to say that the game doesn't have a "sense of impact or momentum" and then to try and prove it with that video is just not very wise.  There are plenty of clues in this video (which by the way is a :40 second clip shown twice from different angles) to show that the game does incorporate changes in momentum based on collision and does have a sense of impact based on collisions with objects.


 

I'm not here to debate. You made some good points but if you can't see that the collision physics are ridiculously bad I can't help you. I think most objective people recognize the collision physics are poor, the question is will PD do anything about it? I've seen collision physics in replays of other games and GT5 is not up to par. 

 

Thanks for your comments anyway. I appreciate people who can stay on topic.

Message 83 of 121 (4 Views)
Reply
0 Likes
Sackboy
Registered: 07/16/2007
Offline
479 posts
 

Re: GT5's poor collision physics and AI

Sep 29, 2009

meatbag_ wrote:

Kenshin71 wrote:

SWERV_GRIFFIN wrote:

 

Ease up on the attitude, you are coming off as very arrogant and argumentative. YOU cannot control what responses are to a thread that you apparently think you own. Show some respect around here, or you will never gain any from the members of this forum. That's some Real Talk. Regards, SWERV


 

I'm not here to gain respect from the members of this forum though. Unfortunately no-one on this forum has any influence on the game at all so it's all academic. I don't go in other threads and derail them or ignore the topic, so I hope others can have the same courtesy.

 

Actually I was hoping there was someone with channels to PD on this board who could maybe pass on the message. Still, I'm vainly hoping someone from SCE might stumble on the thread and I hope other concerned players will add their voice. If people don't want to talk about collision physics and AI and instead go off topic there are plenty of other threads to do it in.

 


i've discussed the collision physics and AI issues several times here in this thread. at no point have you even acknowledged or replied to these points. so unless you have some input of your own on this matter how do you plann to have any 'discussion' about it. all i've read from you on the matter is your OP, since then you seem more interested in moderating the thread rather than contrubuting to it.

Message Edited by meatbag_ on 09-29-2009 12:59 PM

 

Sorry meatbag, I probably missed your comments amongst all the garbage posts. Can you link me or what was the gist?
Message 84 of 121 (4 Views)
Reply
0 Likes
I Only Post Everything
Registered: 08/14/2009
Offline
1080 posts
 

Re: GT5's poor collision physics and AI

Sep 29, 2009

let me correct myself, i've only posted in this thread twice re: the topic, but heres what i posted FWIW:

 

i think most people on here would agree the AI is not exactly Sterling Moss. it is something we hoed would get addressed but it looks at the moment that the AI isnt going to be very hardcore. thats why this discussion has quickly becomed framed as something online. many of us race online for the most part. thats where the challenge is. when you've got human competition, how good does the AI need to be? i'd like good AI as much as the next guy but as long as there is a better alternative i wont complain much.

 

as far as collision modeling, you kinda have to take collisions and damage together because they are so closely related inherently. i think there are probably two approaches to doing it (although i'm not a game designer so you get what you pay for here)

 

on the one hand, you could try to replicate how to materials interact when they collide given specific density and force involved. the problem with this is that you get exteremly variable results with this, since the attitude, speed, mas, etc are never the same, therefore the way the two objects interact can never be predicted, it would have to be calulated as it happens. i dont know how you would go about adjusting the physical model in the game to reflect this on-the-fly, it would reaquire a LOT of overhead from the CPU. probably way more than it could generate if that were its only dedicated task. they do computer simulated crashes at the IIHS, but even they use real cars because with their supercomuters they still cant replicate what you get in real life.

 

on the other hand you can do what all the other game developers do. they define beforehand what effect a given object (like a concrete barrier) will have on any object that touches it. so if your car hits a wall at a 5 degree angle like when you accidentally brush the wall it will push the car in return  with y-force multiplied by 5 degrees. the games physics are probably setup to accomodate this scenarrio realistically, but if you change the equation to say, 45 or 90 degrees and multiply  y-force by 45 degrees it skews the equation radically and you get the pin ball effect. i think there is probably only one formula used to simulate a wall impact, and by definition it cannot be suitable for all wall impacts, so they tune it to what they think will happen most. all the damage i've ever seen in any game is pre-modeled. even the Burnout series uses a predefined series of models for each crash. they may have 50-100 different models of each car to give variety, but for the most part i think the damage modeling is rudimentry at best. doing thison a scale as big as GT5 with the level of detail that would be required, would be a monumental task. 


 i just wanted to point out that these barriers are not concrete, they are plastic. i have seen some people say they were concrete, but they are not

 


 

 

these are what they look like in real life

 

 


 

 


 

 

Photobucket
Message 85 of 121 (4 Views)
Reply
0 Likes
Fender Bender
Registered: 02/07/2008
Offline
2967 posts
 

Re: GT5's poor collision physics and AI

Sep 29, 2009

Kenshin71 wrote:
Seriously guys, if you're not here to discuss GT5's poor collision physics and AI can you move along to another thread? Thanks.

The point of a discussion is, an intelligent exchange of ideas.

This falls short of being that when one party seemingly refuses to accept one simple fact, each person who takes the time to voice thier opinion has an absolute right to that opinion, and to voice it, regardless whether you agree or not.

I tell my kids this, do not ask a question unless you are prepared to consider ANY answer given.

 

If your asking a question with an open mind, you are looking for an intelligent exchange of ideas, if your mind is not open, then your just trying to tell everyone how you feel, and expecting them to feel the same way.

 

Its all in the approach.



Message 86 of 121 (4 Views)
Sackboy
Registered: 05/22/2008
Offline
392 posts
 

Re: GT5's poor collision physics and AI

Sep 29, 2009

Kenshin71 wrote:

jlbowler wrote:

Kenshin71 wrote:

Having played GT5P and seen some youtube videos of GT5 I was a bit concerned about the poor collision physics in the game.

 

A good example is here:

 

http://www.youtube.com/watch?v=6Ksfo2L_65U

 

The Ferrari is just bouncing off walls like a bumper car. I don't care so much about the damage modelling, the superficial stuff is fine, but the collision physics are laughable. No sense of impact or momentum. I also saw a video where a bunch of cars just ploughed into a stationary car, none took evasive action. Has anyone ever raised this with Yamauchi or is he untouchable? I wonder if PD even know what sort of comments/criticism their game gets because they really need to lift their game.


I'll discuss the collision physics and debate your points with you.  You say that there is no sense of impact or momentum.  From what is seen in the video, it is very difficult to see if there is any sense of impact.  This would be better view from an actual driving perspective instead of a replay that shows the action from various angles.  As far as the momentum goes, the first wall impact at :03 slows the car from 233 km/h to 168 km/h.  ( I have no doubt that this would have destroyed the car in real life but we're dealing with a demo of a game and the point is for the player to get their full time in).  The car then bounces on the rumble strips and struggles to accelerate ( I can't say if the struggle to accelerate is based on damage incurred, the strips, or the player just not accelerating).  The second wall impact at :09 takes the car from 198 km/h down to 158 km/h.  The trip through the plastic barriers slows the car all the way through the series of impacts.  The car then bounces over the rumble strips again (you can see the car pitching side to side as it goes over the strips.)  The car is then slowed while going through plastic barriers again.  At approx. :45 you can see the car travel over the rumble strips and the two right side tires are off the ground.  This was either caused by the impact with the wall or the rumble strips.  Either way, this is a sense of impact affecting the car.

 

For you, or anyone, to say that the game doesn't have a "sense of impact or momentum" and then to try and prove it with that video is just not very wise.  There are plenty of clues in this video (which by the way is a :40 second clip shown twice from different angles) to show that the game does incorporate changes in momentum based on collision and does have a sense of impact based on collisions with objects.


 

I'm not here to debate. You made some good points but if you can't see that the collision physics are ridiculously bad I can't help you. I think most objective people recognize the collision physics are poor, the question is will PD do anything about it? I've seen collision physics in replays of other games and GT5 is not up to par. 

 

Thanks for your comments anyway. I appreciate people who can stay on topic.


Maybe you missed my point.  You stated that there is no sense of impact or momentum and linked a video to support your statement.  I analyzed the video and contend that your statement is incorrect based on the information availiable in the video.  I agree that the collision physics could be improved over what was shown in the video.  This is a discussion thread, if you're not willing to discuss or debate a topic then maybe you shouldn't have started it.  Maybe you should just send an email to PD and wait and see what happens.  You will probably be just as satisfied with that result as you will by any result you're going to get here.

DIsclaimer: My opinions are just that, MY opinions. They in no way shape or form express the opinions of anyone else unless they happen to agree with me.

Actual quote from a driver: "I never brake check anyone, but if I see you right behind me, I'm going to brake as hard as I can"
Message 87 of 121 (4 Views)
Uncharted Territory
Registered: 11/24/2008
Offline
1972 posts
 

Re: GT5's poor collision physics and AI

Sep 29, 2009
Thanks to Nuttz, JKbolwer and Meatbag for some very illuminationg posts re the collision modeling see so far in GT5.
Old age is not for sissies.
Message 88 of 121 (4 Views)
Reply
0 Likes
Sackboy
Registered: 07/16/2007
Offline
479 posts
 

Re: GT5's poor collision physics and AI

Sep 29, 2009

meatbag_ wrote:

let me correct myself, i've only posted in this thread twice re: the topic, but heres what i posted FWIW:

 

i think most people on here would agree the AI is not exactly Sterling Moss. it is something we hoed would get addressed but it looks at the moment that the AI isnt going to be very hardcore. thats why this discussion has quickly becomed framed as something online. many of us race online for the most part. thats where the challenge is. when you've got human competition, how good does the AI need to be? i'd like good AI as much as the next guy but as long as there is a better alternative i wont complain much.

 

as far as collision modeling, you kinda have to take collisions and damage together because they are so closely related inherently. i think there are probably two approaches to doing it (although i'm not a game designer so you get what you pay for here)

 

on the one hand, you could try to replicate how to materials interact when they collide given specific density and force involved. the problem with this is that you get exteremly variable results with this, since the attitude, speed, mas, etc are never the same, therefore the way the two objects interact can never be predicted, it would have to be calulated as it happens. i dont know how you would go about adjusting the physical model in the game to reflect this on-the-fly, it would reaquire a LOT of overhead from the CPU. probably way more than it could generate if that were its only dedicated task. they do computer simulated crashes at the IIHS, but even they use real cars because with their supercomuters they still cant replicate what you get in real life.

 

on the other hand you can do what all the other game developers do. they define beforehand what effect a given object (like a concrete barrier) will have on any object that touches it. so if your car hits a wall at a 5 degree angle like when you accidentally brush the wall it will push the car in return  with y-force multiplied by 5 degrees. the games physics are probably setup to accomodate this scenarrio realistically, but if you change the equation to say, 45 or 90 degrees and multiply  y-force by 45 degrees it skews the equation radically and you get the pin ball effect. i think there is probably only one formula used to simulate a wall impact, and by definition it cannot be suitable for all wall impacts, so they tune it to what they think will happen most. all the damage i've ever seen in any game is pre-modeled. even the Burnout series uses a predefined series of models for each crash. they may have 50-100 different models of each car to give variety, but for the most part i think the damage modeling is rudimentry at best. doing thison a scale as big as GT5 with the level of detail that would be required, would be a monumental task. 


 i just wanted to point out that these barriers are not concrete, they are plastic. i have seen some people say they were concrete, but they are not

 


 

 

these are what they look like in real life

 

 


 

 


 

 


 

OK, thanks for that. Maybe I just saw 'damage' and my eyes glazed over. Disagree about AI and online. The online should not be a crutch for poor AI. There will be plenty of people who will hardly or perhaps never go online so the AI needs to be pretty good. A lot of people started playing GT before online existed, many will avoid it because of the clowns that deliberately crash etc.

 

Regarding the comment about having to take collision physics and damage together, not really. You can develop a collision physics model that is independent of damage, in fact I'd say that's how most devs do it. I read in an interview Yamauchi said they started working on damage only a couple of months ago. They should have been working on a new collision physics system way before that.

Message 89 of 121 (4 Views)
Reply
0 Likes
Highlighted
MVP Support
Registered: 12/04/2003
Online
44760 posts
 

Re: GT5's poor collision physics and AI

Sep 29, 2009

Kenshin71 wrote:
I'm not here to debate.

No kidding.

It also seems as if you are not here to listen to what people try to tell you, either. For all of the following posts in which members tried to help you out and explain things, in varying degrees, in their own way, you hardly even commented on what was in most of the responses. In most cases you either flat out ignored what people tried to tell you or you didn't understand what they really wrote. This makes respondents jobs much more difficult.

#2 - sort of
#5 - directly
#36 - sort of
#40 - sort of
#56 - directly
#64 - directly
#65 - directly
#66 - directly
#70 - sort of directly
#73 - sort of directly
#76 - sort of directly
#78- sort of
#80 - directly
#87 - directly


Kenshin71 wrote:
I think most objective people recognize the collision physics are poor, the question is will PD do anything about it?

Read through the above posts again. Most of us recognise that what we've seen is only a snapshot of where they were at the time of TGS, that's all. We are the ones being objective here. I don't remember anyone saying that what we've seen is "good enough", it's OK. We also have found out that it seems the damage engine, which involves the collision physics that so concerns you, was made in only a couple of months prior to recent unveiling. Given that PD is going to take their sweet time about things, expect improvements. They probably won't tell you what improvements are coming, but expect them anyway.

On top of all of this, only some cars will have deformation and that should affect the physics. Knowing PD's past releases, I can't guarantee that you will ever be satisfied, though.


Kenshin71 wrote:
I've seen collision physics in replays of other games and GT5 is not up to par.

Considering you've only seen a few demo clips and have not played the demo yourself yet, what's your point?
Wait for the finished product to say it isn't up to par.

Cheers,
Message 90 of 121 (4 Views)