pb with hingeActuator joint in Dynamic

[blackbird_dream]

I made a simple vehicle composed of a Newton box with 4 hinge actuators positioned at the wheel's locations (rotating axis aligned with y global, gravity axis).
The 4 wheels are Newtonbodies linked to the box through the hingeactuators.
Then I input via a script the vertical reaction at each wheel (elastic damped force in direction +y).
Then I add a longitudinal traction force at the 2 front wheels (constant force in direction +z) when hitting the "w" key.

I display 2 zooms of the Left front wheel at 2 instants, at rest and when applying the traction force.
In the capture I display the gizmo of the wheel newtonbody , the wheel-axle of the mesh child of the wheel Newtonbody and the downVector attached to the wheel newtonbody. All axes should intersect at the wheel Newtonbody center. It does at rest:


, but it doesn't when I apply the traction force (hitting "w" key) :


here is the schematic diagram of a quarter model :


and the global view of this simple model :


and the files :
https://ufile.io/x7n60

My question is : what is the reason of the drift between axes?
Thks

[blackbird_dream]

when I update the position of the mesh in the OnApplyForceAndTorque event

Code: Select all

MeshOfTire.transform.position = body.Position;

the 2 axes (mesh's center & newtonbody's center) coincide but there is still a gap with the Newtonbody gizmo's center.
I don't understand why the mesh position has to be updated this way and why the gap.
It seems like problems of synchronizing objects and joint's softness combined.

[JernejL]

when I update the position of the mesh in the OnApplyForceAndTorque event

Code: Select all

MeshOfTire.transform.position = body.Position;

the 2 axes (mesh's center & newtonbody's center) coincide but there is still a gap with the Newtonbody gizmo's center.
I don't understand why the mesh position has to be updated this way and why the gap.
It seems like problems of synchronizing objects and joint's softness combined.

Just to focus on this part: This is wrong, you should update position from the transform callback, not the forceandtorque callback.

http://newtondynamics.com/wiki/index.ph ... rmCallback

Edit: i would recommend use imgur.com to upload images - no watermarks, no waiting times, just go to website and hit ctrl+v to paste a image to upload it.

[blackbird_dream]

Are you talking about ND in Unity ? I thought I hadn't to update it as it was performed by the plugin.
Thks for the link.

[blackbird_dream]

Sweenie
It seems the issue is similar to what you raised in my previous post about the demo_06.
Joints have becomed softer in the last build of ND ?

[Sweenie]

Yep, demo 6 behaves very different.
There has been a big update of the joints in Newton so I will need to check if something needs to be changed in the plugin.

[Julio Jerez]

I will try to update the unity plugin this weekend and see what is wrong, the joint should not be any softer than before, if any thing I expect then to be better.

[blackbird_dream]

good news. Looking forward to trying the new build.

[Sweenie]

Demo 6 (Several connected hingeactuators) is a bit extreme as they rotate at an angular rate of 100 which causes the outermost bodies to move very fast(almost like whiplash movements), but before the recent joint refactoring no drifting occured even at high angular velocities. But now the joints drift apart a bit. If i lower the angular rate of the actuators it's look ok though.
If I change the joint solver mode to 1 or 2 the joints wriggle around like a worm instead.

I can't record now but managed to get some screenshots...

9 connected hinge actuators with an angular rate of 5
After a couple of seconds


After another couple of seconds(joints keeping together fine)


Now with an angular rate of 100 after a couple of seconds
Here the joints start to separate at the outer parts of the "chain"

[blackbird_dream]

ok but I need stiffer joints for my application like it was with v1.5.
See my demo the wheel should not move apart from the chassis. In cornering conditions the gap is more severe and the wheel goes like crazy.
And it's like a chain reaction as ground interactions are computed from the instant speed and position of the wheel.
Or maybe there is something else wrong in my simple demo (not to be ruled out).
Anyway the Demo06 was stiffer before the updates.

[Sweenie]

Trust me, the joints usually are rock solid in Newton so something must be wrong.
Hopefully i've only missed something in the plugin.

I found a compiled version of demo 6 from last year that shows how solid the joints normally are.
http://www.svenberra.net/NewtonActuatorDemo.zip

In the demo some actuators can't keep their angular position when the whiplash effect happens but that's expected since i've set each actuator to withstand a maximum torque of 10000 and I guess the whiplash generate torque forces way above that.

[blackbird_dream]

where can you change the joint solver mode ?

[Sweenie]

It's not exposed in the plugin so i changed it directly in dNewtonJointHinge.cpp

[blackbird_dream]

your version dated 15/3/18 was working with which Newton Version ?

[Sweenie]

Should work with a newton version from the same date. The recent experiments with parallell solvers seems to cause "explosions" in the plugin with NaN values in the body transforms as a result