Joints stiffness

[Julio Jerez]

Couldn't steer it though so i tried to drag it with the mouse, but the mousepicking still generates absurd amounts of force i guess so i managed to rip the treads off. Awesome work

:lol: Normally when that happens with almost any solve when you get is a jittery explosion, but here (at least on my test) the tread just stays there, that a testament that the thong is working correctly. When I was writing I new it will world by the math, but it is always a suproce that work as you actually expect it would.

hey Sweenie do you remember when you where asking a mode that can mix exact solve with
an we add in per island threshold size?

well this is the generalization, if you apply that to the vehicle model that you were making it should give you all you where expecting and I believe even more.

Yes in the demo steering is a problem, in has to do with the friction model. Basically I have to add a callback to align the tangent direction other with the keep getting aligning to the diction o motion and prevent the hold th9ong from turning at all.
Also the side friction should be lower, suspect that is real life track are designed is such way the side area if less that the area on the direction of motion other wise turn is very inefficient.
I will try some of those tricks tomorrow. But so far this is just a matter of modeling.

[Julio Jerez]

I was search the net for confirmation of what I found about track vehicles.
form what I am seen tanks are the worse kind of vehicle there are. the track system is extremely fragile, very prompt to failures, extremely power hungry, and requires complex drive systems.
why the military keep making Tank make no sense whatsoever.

I found these article thar explain some of the thong I found.
http://www.gizmology.net/tracked.htm
By the bottom the guy explain some trick to make the turn possible,

[JoeJ]

For the collision aggregate, would it make sense to give an option to dis/enable self collisions?
Could save some callbacks.

[JoeJ]

Oops. Seems this is already there (NewtonCollisionAggregateSetSelfCollision)

[Julio Jerez]

you mean these

Code: Select all

int NewtonCollisionAggregateGetSelfCollision (void* const aggregate);
void NewtonCollisionAggregateSetSelfCollision (void* const aggregate, int state);

that is part of what I was saying before that aggregate in theory soul be slower, in practice the groping functionality allowed for some shout cuts in the case that apply. example vehicles,
in the demo for example I god the tread in a aggregate, so the do inter link collision, the otherwise will required lots of collision flag in callback. and since they do not add overhead when no used, it is a win.

[Julio Jerez]

Hey Joe, now I figure out how to solve the problem the joint lose stiffens when the are limits.

I suspected that could happened but I thought the problem would be now but it could be quite problematic. by the good thon is that now know to fix it and I do not thong it will be a problem.
I will work on the next week.

[JoeJ]

Great news! Hope it works...

[Julio Jerez]

wow, i know it isn't finished yet but that thing is a monster
Couldn't steer it though so i tried to drag it with the mouse, but the mousepicking still generates absurd amounts of force i guess so i managed to rip the treads off.
Awesome work

Hey Sweeny I added the control for steering left and right. It is no driven by a powered joint.
I have no added the Booming stuff but that is no really a problem
I also made the link and the small wheel 20% bigger.

when I did that I thread became a lithe looser, by if I remove one link then it is too tight, I will have to move the front ore rear tire a lithe to give the tread more tension.

The problem I am facing now Is the how to make it turn easier, the are simple too make contacts with the ground, I tried to play with side friction, by still no lock.
there are I know that this is far for a real racked vehicle, but it still an approximation and I think there has to be some principle in he design of tracked vehicles that allow the for turning and I am missing that.
Basically the side tire friction on the extreme tire, prevent the vehicle from turning,
Next what change the tire fiction according to the design turn rate. maybe the will make better.
as it is now you can turn but it is very hard.
also I would not try to pick with the mouse, the treads are no attaché to the body, so it loses them.
All in all is does work. and it is surprisingly stable.

[Sweenie]

I googled around a bit and several people mention that a tank turns easier if it is as square as possible and also if most of the weight of the tank is placed on the center of the treads.

[JoeJ]

You could try to align friction direction to be tangential to the com,
maybe that works better than reducing side friction only.

It might be interesting to see a real tank operate on hard ground (concrete).
Maybe they work properly only on soft ground (soil).

[Sweenie]

Here is some interesting reading about tracked steering...
http://www.gizmology.net/tracked.htm

It occurs to me how inefficient this design is, ok, it is powerful for climbing steep hills but it's so much power wasted on friction!

[Julio Jerez]

]It is not only a problem with friction, I try to simply the model by take out the track and having moving by the wheels, and the problem is obvious.

It is easy to turn, but is still hard, I can see how making the aspect ratio more square can make tuning easier. It is easy to see a long vehicle the longitudinal friction does no have radius of action to overcome the side friction with has a long.
so make the vehicle wider can help.

There are also other problems, the supposition is no that simple once you want to make no vertical.
in the image, you can see that the vehicle is resting on one wheel, and do no fall, that is because with the brake on, no wheel can roll so the only way the wheel to move is up or down.
[attachment=0]projects.png[/attachment

so for that king of suspension to work it can not move in an arch it has to move vertical, so it either need a way for bar that the wheel is connected too, to compress, which will bring more joint and springs.

So I thud real tank most have the problem too, and guess what Tank suspension are very complex system with up to 20 or more moving parts so that the wheel move vertically.
I appears that in WWII the germane discover that problem with tanks and the adopted a simple suspension system that was first rejected by the Americans. So German take move the wheel vertically and have and advantage over British and American tanks.

Why The military invest on Tanks make no sense to me?, for what I can see tanks are extremely fragile vehicles, with high maintenance cost, extremely inefficient, very expensive, and with all the armor they have to be very heavy so they can not have vey large range. Tank a solutions in search of a problem.

I will change the model a little.

[Julio Jerez]

changing the aspect ration does help, now it turns easy but only if it has not thread.

but the thread problem I believe is caused but the kinematic unfeasibility of the suspensions,
with that kind of suspension, In fact the wheel should not even move, I am surprised it does not blow up.
I comment out the threads for now until I have a feasible model.

I still do not know if I will go for the elaborate suspension it seem to much work for very lithe gain,
I will make vertical with spring /damper and slider contacts

[Sweenie]

Oops, just realised i posted a link to the same page as you did, .
I had the feeling i've seen it before,

You seem to have the same problems as real tanks do which prove how realistic your simulation is

[Julio Jerez]

for each link I filter all the contact with the floor leaving only one, the I found that when I reduce the side friction I made a mistake I set to 5.0 instead of 0.5.
This is not perfect yes my any mean, by is the best yet.
If you have time try it out

Now I found a new problem with tanks.
It have to do how thread moves. Thread are a link bodies with only one degree of freedom.
for the tank in move forward they want to continue moving forward.
however when the tank in going to turn, the fruition between the wheel and the link is not enough to keep the tread on the reel. so eventually the threads fall out the tracks.

I now have to come up with a way to keep then on the track,
I have to say how even came up with tank design I do not know what that person was thinking,
A thread vehicle is an abomination of vehicle locomotion.

there is also another problem a Tanks turns in perpetual kinetic fruition, so thread most be replace quite frequently. By doing this now I know why the Pentagon Budget is 500, billion dollar a year
To me the cost of Operation for a M1A1 must be more than 2 or more million dollar a year.