vehicles

A place to discuss everything related to Newton Dynamics.

Moderators: Sascha Willems, walaber

Re: vehicles

Postby Julio Jerez » Wed Oct 31, 2018 7:20 pm

the model is design for realism, if you make a car with a mass of 100, and tires of forty, it will act like a monster truck, 40 is on the high end for a mass for a typical sport car which weight from 1200 to 1500 kg. the value of 1000 is because this is a light car and the rest of the parts, tire, engine, differentials, etc add to about 300 to 400 kg, so is not even as light as it seems.

to handle the stiff suspension there is a dedicated solver that use a langragian model to integrate the body one step and calculate the acceration at time t plus st, and use it as the acceleration at time t.

this is why you saw it jumping like that, the spring mass system when integrated the masses became part of the response. and they are very stiff for those masses, for a time mass they become do stiff that brake numerical accuracy.

basically the car must has reasonable paramerter set, which many can be found on car magazines.
Julio Jerez
Moderator
Moderator
 
Posts: 12249
Joined: Sun Sep 14, 2003 2:18 pm
Location: Los Angeles

Re: vehicles

Postby Julio Jerez » Mon Nov 12, 2018 11:01 am

Ok I rigged all the bodies and joint, but something is still off somewhere.
the first test is the burn out, I expected it to make a burned out in first gear when pressing the brake and and acceleration simultaneously, but is not doing it instead the drive force over power the tire brake force. These are fix points that can be used to tune the envelop in which the car operates.
another thing that look wrong is the lateral force seem too stiff and prevent the far from tail spinning.

I will go over some calibration this week but for the most part the vehicle interface is completed.

the missing thing are the aerodytnamics forces possible some stability controll, but is should drive witout those.
Julio Jerez
Moderator
Moderator
 
Posts: 12249
Joined: Sun Sep 14, 2003 2:18 pm
Location: Los Angeles

Re: vehicles

Postby Julio Jerez » Tue Nov 13, 2018 1:13 am

In debugging teh car, I suspected I had a bug on teh tire model, and indeed I had one,
after I fixed this is how it looks like.
This is teh Giancarlos Genta brush tire model, and check out hwo cool that look.
Untitled.png
Untitled.png (10.27 KiB) Viewed 8169 times


The sleep ratio if one is teh vale 10.0 on teh horizontal axis and it reach about 80% of the load, this is incredible acurate accurate according to any empirical tire model.

The one thong that is hard to find is a realistic vale for the tires lateral and longitudinal stiffness, which according to teh book is support to be eassy to fin on any sheet, but I has not found any.
after a few trials an error I found that a good value seems to be the vehicle weight divided by the tires count.
anyway I will continue with calibration.
Julio Jerez
Moderator
Moderator
 
Posts: 12249
Joined: Sun Sep 14, 2003 2:18 pm
Location: Los Angeles

Re: vehicles

Postby Julio Jerez » Tue Nov 13, 2018 1:37 am

for the people who want to see how is behaves here is a small video

notices how one of teh front tire act as a pivot that is completely clamped. This is a hard to achieve effect without cheats my friend. check it out.
It still slide too much, there is a still a lot of more calibration to do, mainly the lateral forces, but you can see how start to alone wich each fix point tuning.
Julio Jerez
Moderator
Moderator
 
Posts: 12249
Joined: Sun Sep 14, 2003 2:18 pm
Location: Los Angeles

Re: vehicles

Postby arpio » Tue Nov 13, 2018 7:15 pm

Looks very cool. I hope I'll have some spare time on the weekend, so I can try it myself. :D
arpio
 
Posts: 23
Joined: Wed Oct 24, 2018 4:33 pm

Re: vehicles

Postby Leadwerks » Sat Nov 17, 2018 4:45 pm

Ha, spinning donuts like that is exactly what I used to do in an old Mazda SE5:
Image
User avatar
Leadwerks
 
Posts: 569
Joined: Fri Oct 27, 2006 2:54 pm

Re: vehicles

Postby Julio Jerez » Sat Nov 17, 2018 6:59 pm

ok guys I am now at the point whe the different coordinate system that people use in thier engines is important, hence my interest in a standard file format.
after researching few format specifications, my conclusion that the only format that is practical and general is in fact FBX. what remain is the license part.
this is what the license says
"This software contains Autodesk® FBX® code developed by Autodesk, Inc. Copyright 2008 Autodesk, Inc. All rights, reserved. Such code is provided "as is" and Autodesk, Inc. disclaims any and all warranties, whether express or implied, including without limitation the implied warranties of merchantability, fitness for a particular purpose or non-infringement of third party rights. In no event shall Autodesk, Inc. be liable for any direct, indirect, incidental, special, exemplary, or consequential damages (including, but not limited to, procurement of substitute goods or services; loss of use, data, or profits; or business interruption) however caused and on any theory of liability, whether in contract, strict liability, or tort (including negligence or otherwise) arising in any way out of such code."

In the prohibited section the restriction is that you cannot repackage and distribute the SDK without our written consent. You cannot claim the SDK to be your own, that is what that clause prevents.

Mark Davies


so basically what is prohibited is the distribution of the SDK, but we are not doing that, what I have done is that I wrote a tool utility that takes FBX file and convert it to an NGD file.

this way anyone can just install the FBX sdk from Autodesk, and build the tool if they want to try anything in teh sandbox.
All they need to do is install the fbx sdk and set an environment variable them compile the tool
we aren't even distributing compiled version of the tool.

for now I only have the fbx2ngd conversion tool, therefore I added the FBX version of the vehicle to the media folder, this way people can open the file in their prefer modeling package:
Max, blender or whatever.
an check out how the asset are set up in the Newton sandbox demos.

this also allow us to communicate better with asset because anyone can give me any FBX file
and we use it for new demos or bug fixes.

the other part this is good for is that I can now try vehicles on a different coordinate system to make sure the vehicle model can deal with that issue which seems an obstacle that prevent people from using it.

later to be even more safe, I will write the inverse tool ngd2fbx, that way we do not even have to distribute .fbx files only two open source utilities.

But as far as I can see is not illegal to distribute .fbx files.
Julio Jerez
Moderator
Moderator
 
Posts: 12249
Joined: Sun Sep 14, 2003 2:18 pm
Location: Los Angeles

Re: vehicles

Postby blackbird_dream » Sun Nov 18, 2018 5:07 am

Very Nice the demo with the brush model. Have you tried Pacejka model (Magic formula) also?
User avatar
blackbird_dream
 
Posts: 354
Joined: Wed Jun 07, 2006 3:08 pm
Location: France

Re: vehicles

Postby Julio Jerez » Sun Nov 18, 2018 11:39 am

I am using the brush model for one reason.
off all the model that I read in several books about vehicle dynamic, it is the only one that has a nice mathematical derivation that can model accurately a tire in the linear range and drift apart when the tire reach the non linear range (mean the point at which start losing grip.

I learn this model from the book Motor Vehicle Dynamics by Giancarlo Genta.
Chapter 2 is dedicated to tires and then go to derive the mathematics for a brush tire approximation.

The brush model is only accurate within the linear range in which the tire operates but in return we get a model that is well defined over the entire linear range for both lateral and longitudinal slip combined.
This is something the Magic can't do.

The magic formula is better because it model the non linear behavior when a tire loses grip, but this is a fitting curve with no real theoretical foundation and even when using it the model is one dimensional, you still have to use empirical circle of friction projections.

you can get an explanation in this video
https://www.youtube.com/watch?v=We5iNzg6AAA
check out his previews videos as well, this professor explain the theoretical derivation of the brush model in detail.
at about 34:10 he explains that the model only generate decouple longitudinal and lateral forces. these are good for what is called steady state studies, but not for a simulation.
this is why I prefer the brush analytical tire model because provide a combined effect even if it loses accuracy in the non lineal range.

People may interpret that as bad or poor, but this are trades engineers has to do, take for example that modeling of beam and columns for buildings. The model of Young is only acurate in a very small linear range of compression of elongation for material like still and concrete. but again engineers do not make buildings that are suppose to operate in the nom linear range because at that point you have a much bigger problems to worry about.
Similarly vehicle are designed to drive on the
linear range of the tires, even professional race car drivers drive cars on the limit of the linear range, because one you pass that point the tire yield less performance.
It is only people who like to drift, do tire burn out and tail spins, that put the tire in the non linear rage so that the can get different behavior between front and rear tires.

odly enough for what I want to do, the brush model is incapable to producing an in place tire burn out.
this is because the tire model the force in the linear range dos not decay, so when the tire is at a maximum spins the force do not became lower than when the tire is a maximum stress but not sliding.
so maybe after all the tuning and calibration is done I add a Pajecka model as an option.

but in any case the tire model is well encapsulated in a function, it can be updated by anyone with interest, for now I am more concern with a completing the vehicle model and cover a wide range of drive modes from realistic to arcade.
Julio Jerez
Moderator
Moderator
 
Posts: 12249
Joined: Sun Sep 14, 2003 2:18 pm
Location: Los Angeles

Re: vehicles

Postby blackbird_dream » Sun Nov 18, 2018 5:00 pm

Pacejka is a fitting model you re right. Not phenomelogical but quite accurate depending on the data you acquired. Widely used even in racing games. But Brush models are also well known and useful.
User avatar
blackbird_dream
 
Posts: 354
Joined: Wed Jun 07, 2006 3:08 pm
Location: France

Previous

Return to General Discussion

Who is online

Users browsing this forum: No registered users and 9 guests

cron