Let us Talk about Vehicles.

[Sweenie]

Is the tire a rigid body with full collision or is it some raycast "hybrid"?
The reason i ask is although i enjoy fast race cars would this vehicle model work well to simulate for example a jeep or monster truck with big tires?

Raycasting doesn't cope well with vertical surfaces and with small tires you can add fenders to prevent anything from touching the side or top of the tire, but you can't really put fenders around a monster truck tire.
If you drive into a big boulder or a wall with a large raycast tire things usually goes bad.

[Julio Jerez]

NO ray casting in the vehicle model.
Every component of the vehicle is a rigid body with full collision. That goes for the tire and the engine. Believe or no this what makes the model simple and more predictable.

this model is intended to handle, jeeps or monster trucks with big tires, scoters , motorcycles,
if it has while it will handle it, and even if it doe no has wheel stuff like a helicopter sould be possible.
BTW a the tire can be either independent suspension, or attach to a real axel will also be a rigid body with mass.
The possibility are unbounded, as along as if can be define in turned of rigid bodies connected by joints.
the second demo I will made will be a monster truck with real axel and all wheel drive

[Julio Jerez]

Umm I spend the weekend work of the tore model. But for some reason the brush model does no work as I was expecting. The car does not feel good.
I am trying to avoid using PacejKa because it is not really a theoretical model, it a Numerical fitting based on coefficient that supposedly can be found on data sheet from real manufacturer.
But I have never seen a table with these parameters, the only place I's seen the mention is on Books, but to may tire do not these values.
what I have seen is people who make vehicle simulator making their own graphical interface for the user to crate their Own tires. but for me this is way beyond the scope of the vehicle model.

There reason I like the brush model is because is a Analytical theoretical model, the depend only on three parameters called cpx, cpy and pneumatic trail (lateral stiffness, longitudinal stiffness, and pneumatic radio of self aligning torque)
The model can be fitted into the Pacejka equation plus because the coefficients are linear, they are easier to used in the calculation vehicle stability derivatives.

However the results I am getting as so far from realistic, that something must be really wrong, in my implementation. I am going to go over these papers again, I must have something very wrong
http://www.control.lth.se/documents/2007/jsvenPDH.pdf
http://kth.diva-portal.org/smash/get/di ... TEXT01.pdf

The good new is that I do have more than one bug, the bad is that I do not know the will affect the handling. Also there is another parameter that need to be specified, which is call Vehicle understeer Gradient, which play a huge factor in the vehicle design and is also one of the reasons I want to use the brush model.
https://etd.ohiolink.edu/!etd.send_file ... ion=inline

Who know that vehice dynamic was actually quite a complex subject.
It is way of a lot more than just place for tire on a box. Hopefully the works

[Sweenie]

Maybe the vehicle would be easier to steer with some kind of progressive steering?
I'm not a vehicle expert but doesn't modern cars have something called "Progressive power steering" that prevents sharp turns at higher speed?

Also, maybe if you make the wheel gradually turn towards it's target angle instead of turning it in an instant.

I suppose it's tricky to tune the friction when driving with a keyboard. I mean try driving your own car with either zero throttle or full throttle. A pretty dangerous driving experience I would guess . Also if it is a Viper I suppose you'll have a hard time staying on the road.

[Julio Jerez]

yes that is a big part, I am adding a automatic pilot so that the can is driven but a controller that can keep parameter constant.
but there are lot of thong that affect the design of a car, the tire model is only one part of it, although a very important one.
It turn out that there are many other parameter that very important in designing a car.

there are these thing call Gradient stability, transient respond, etc that are use as criteria to design a car. The tire model is part of all of them.

All of the pare and text book use the bicycle model for a design and the use theory of small perturbation to intruded the effect of thongs like roll rate, roll axis, understeering, over steering steering gradients and so on. I always underestimated the bicycle model I though is was a jack, but there is a lot more to it than I though.
The bottom like is that the physic of and object moving a high speed is quite different that the physics is a body moving a very low speed.
My biggest problem is that all the work I seen used the bicycle model to develop that relationship between parameters, and is difficultly to map the parameter to a rigid body simulator with more degree of freedom that the simple 2 dof bicycle model.

I have to so some more reading, but one thing is right my control system is terrible, the digital one of zero control is no good at all so that's the first thing I need to solve.

I check code with the car in auto pilot, and you can see how much under steers, so maybe this test I can use to tune the Steering gradient and see if the theory apply to the simulation.

[Julio Jerez]

I committed a test that makes the vehicle travel at about 55 mph, I wanted 80 but it slides out of the map.
you can see how the car oscillates along the path after is take a turn.

I am reading these book: Motor Vehicle Dynamics by Giancarlo Genta, and they talk about the transient oscillation after a small perturbation, they apply a Laplace transform the bicycle model and form there they get the steady state and the transients response.
I always ignored that analysis, because implies Eigen vector and Pole and zero analysis, but this is actually very important.
a good car should not oscillate like the demos does. it should have natural damping an may one overshot, the demos act like a spring system, so that is a sign of a really bad setup.
a Transient respond is how the vehicle react to s small perturbation. the perturbation could be a small change in the steeling angle(my case), by it can be a bump in the road, a gust of wind, even a flat tire, or tire pressure unbalance.
The analysis assume a zero acceleration, so I need to make the vehicle travel a constant speed for that.

when I started this my goal was to have a car that can drift, and now that I have that, be careful with what you wish, it turned out that car stability is way, way more complex and harder to achieved than drifting.

[Julio Jerez]

The second part in that I am still dubious is in the vehicle tire parameters.
all of the vehicle data (dimensions and engine specification) I am take from the Viper data sheet that are every where over the internet, I am actually using data from the 1992 which is actually quite mild (450 hp) compared to todays vipers (650 hp) which are drivable without automatic control.

The part I made up are suspension stiffness and tire specifications and some engine missing data.
The suspensions are not really critical as long as they are stiff enough to minimize roll transfer.

Tires are a different story. Every one talk about the Pacejka Magic formula and keep saying that tire manufacture provide data sheet with the coefficients. But when I look around I have not found a single tire data sheet with Pacejka coefficients. The only data sheet I found is on the appendix of book: "Motor Vehicle Dynamics by Giancarlo Genta" which provides parameters for an unnamed sport car.

The other tire model is the Brush model, and there are some variations.
The brush model is actually not very realistic when is get to the non linear range, but is the only one that is analytic and can be used in the stability equations.

There are lots of unknowns on the vehicle dynamics stuff, lot of stuff is a black art that is done by empirical tests and kept secret.

[Julio Jerez]

basically the test that I need to runs are the one on papers like this.
http://drum.lib.umd.edu/bitstream/handl ... sequence=1

this is simple enough to verify that the vehicle model in the engine satisfy the equations in the book and papers, and to see how close it comes to the predictions. Stuff like low speed and high speed state steady.
but I need a better set of test. Maybe a procedural circular path or a Daytona type path when the auto pilot is easy to implement and not the simple AI stuff that is very erratic.

[pHySiQuE]

My frank opinion is that making a better vehicle implementation is not going to move you forward one bit.

Newton's differentiating feature is stability and accuracy. If you look at the recent game SOMA, there are physics interactions in there that would be completely impossible with PhysX or Havok. Yet Newton lacks the most basic functionality of a working character controller and a fixed joint, or other means of picking up objects and moving them around smoothly. You should focus your effort on making out-of-the-box player movement and interaction as smooth and solid as the rest of the physics are.

BTW, why didn't you ask Frictional Games to add a splash screen or give you some kind of shout out to promote Newton? Amnesia and their other games are extremely popular but you have squandered the opportunity to make any connection between the two products. You get angry about particle demos Nvidia makes with PhysX, but totally ignore your own opportunities. Players interact directly with Newton physics in ways they simply cannot with other physics libraries, but do we hear a word about it? If Nvidia had a feature like that they would be making a big deal about it, but you totally ignore your own killer features.

Your path to success will not be the same as any other product. You have to recognize your own strengths and double down on everything that makes Newton great. The reason Newton isn't more popular is because you lack focus. This is a common problem for programmers who are dealing with low-level details, it can be very hard to step back and see the big picture of why your product is good.

[Julio Jerez]

These days people who use the Newton engine run away from it faster than Dracula run away from the sun light in movies or, in real life, faster than a Democrat politician running for congress run away from Obama in 1014. This is what they say in Steam.

"SOMA also makes use of a brilliant physics engine, where every object in the world can be picked up, rotated, examined, thrown, etc."

they refer to some "Brilliant, physics engine", some how not named.
My guess is that the do not want to create flame wars by mentioning Newton and I do not blame them. When you do google search to see who is using the engine, these is the * that I get

Anonymous28 April 2015 at 13:58
I see you are leaving OpenAL behind, in favor of FMOD. Are you gonna do the same with Newton, in favor of other physics engine ?
Reply
Replies
Thomas7 May 2015 at 12:24
We tried some other engine's but they lacked the stability Newton has, which is crucial for our games, so gonna stick with for a while. Now that it is open source it is extra nice too. That means we have source for a crucial part of the game. If we wanna preserve our games, that is an important bit.

This "Anonymous" fan was only interested is that they do not use the Newton Physics Engine.
that has been the story, is almost like people get paid for dissuading anyone from even taking a look that. I do not have the energy or the resources to continuing fighting that.

[pHySiQuE]

Just devote 10% of the time you spend writing code to promoting Newton. No one will stand up for you if you can't even stand up for yourself. Have you even talked to Frictional Games about it?

Sorry to be so blunt, but I'm trying to help.

[pHySiQuE]

GDC would be a perfect venue for you because it's almost like an academic convention. Your work would be very highly regarded there.

[Stucuk]

The reason why random people would say things like "Are you going to ditch newton in favour of <Insert Something Else>" is because the general public doesn't have any clue about newton and its capabilities. The general public only knows about the Physics Engines used in AAA games and assume that as EA/Activision/Etc publish AAA games with certain physics engines that those have to be the best.

Ignore the general public (Most humans lack common sense after all) and ignore the "Self Proclaimed Experts". Listen to the people who have actually used newton.

[pHySiQuE]

they refer to some "Brilliant, physics engine", some how not named.
My guess is that the do not want to create flame wars by mentioning Newton and I do not blame them. When you do google search to see who is using the engine, these is the * that I get

That's a customer review, not the devs. SOMA players have no idea what Newton is because you haven't put that information out there.

In marketing, you can't expect the public to come to a conclusion on their own about your product. You have to explicitly tell them why it's great. (You also have to have the substance to back up what you're saying, or your message won't be accepted.)

If you'd like help with marketing I would be willing to volunteer some time. If you are stubborn and ignore the need to promote your product it will not catch on, no matter how much better it is. If you just want Newton to be a small hobby that's okay, and I definitely appreciate you making it available for us to use, but if that's the case please stop complaining when it gets passed over in favor of inferior libraries. It's annoying because you have a great product and there is a solution to your problem, but you refuse to address it. I have seen this among really smart programmers, they just can't think about anything except the code.

[Julio Jerez]

That's a customer review, not the devs. SOMA players have no idea what Newton is because you haven't put that information out there.

I can't force anyone to say they use the engine in their products. Game cost lot of money to make.
In the large majority of the cases the people who make a game are not the people who fund the game.
There are a lot of legal people who's only job is to check what is beneficial to the game and what is not. These legal people know nothing about technologies, in the case of opens source libraries all the do is search the internet to see if the libraries they are using has a favorable opinion.
and if they do not see more positive that negatives, then they will not list it.

What you see the result of years of propaganda from self appointed experts who originally started from ODE who felt the indignity that Newton was better than both ODE and Novodex and they could not have that. They did not care about Tokamak because it was fast but just as bad or worse than Novodex, but Newton was real thread because Newton was using real physic equations to generate plausible results so they have to discredited it.

These things are not uncommon, like they say, survival of the fittest does not means the best will survive, it means the one that fit better to its environment will survive.
This is how you see VHS beat Betamax, Intel 8088 beat Motorola 60000, Windows beat Unix, an so on.
Newton did not adapted well and it is not anyone fault but mine.