motor constraint isn't working #113234

New Issue

michael campbell · 2023-10-04T05:30:22+02:00

michael campbell commented

2023-10-04 05:30:22 +02:00

System Information
Operating system: Windows-10-10.0.19045-SP0 64 Bits
Graphics card: NVIDIA GeForce GTX 1070/PCIe/SSE2 NVIDIA Corporation 4.6.0 NVIDIA 536.99

Blender Version
Broken: version: 4.0.0 Beta, branch: blender-v4.0-release, commit date: 2023-10-01 16:56, hash: 4d3534401060
Broken: 2.91.0
Worked: 2.90.1

Caused by 1aa54d4921

Short description of error
motor constraint has no effect

Exact steps for others to reproduce the error
open attached blend file and press play. Move the cone empty on the x axis to change the speed of the tanks back wheels which are driven by the motor constraint. The empties are aligned so the x axis oriented correctly with rotation as described in the user manual.

**System Information** Operating system: Windows-10-10.0.19045-SP0 64 Bits Graphics card: NVIDIA GeForce GTX 1070/PCIe/SSE2 NVIDIA Corporation 4.6.0 NVIDIA 536.99 **Blender Version** Broken: version: 4.0.0 Beta, branch: blender-v4.0-release, commit date: 2023-10-01 16:56, hash: `4d3534401060` Broken: 2.91.0 Worked: 2.90.1 Caused by 1aa54d4921c2e8d7114f463a940c169ee573f557 **Short description of error** motor constraint has no effect **Exact steps for others to reproduce the error** open attached blend file and press play. Move the cone empty on the x axis to change the speed of the tanks back wheels which are driven by the motor constraint. The empties are aligned so the x axis oriented correctly with rotation as described in the user manual.

physics driven tank.blend

5.8 MiB

michael campbell added the

labels 2023-10-04 05:30:23 +02:00

Philipp Oeser commented

2023-10-04 10:00:10 +02:00

Can confirm, will check

Philipp Oeser added

and removed

labels 2023-10-04 10:00:29 +02:00

Philipp Oeser commented

2023-10-04 11:00:25 +02:00

Pretty sure this was caused by 1aa54d4921

OK, so while there are a couple of dependency cycles detected, (such as the following), it was still somewhat working prior to above commit :

Dependency cycle detected:
  SCScene/RIGIDBODY_SIM() depends on
  OBCylinder.016/TRANSFORM_SIMULATION_INIT() via 'Object Transform -> Rigidbody Sim Eval'
  OBCylinder.016/TRANSFORM_EVAL() via 'Transform Eval -> Simulation Init'
  OBCylinder.016/TRANSFORM_CONSTRAINTS() via 'Constraint -> Transform Eval'
  OBEmpty.032/TRANSFORM_FINAL() via 'Limit Distance'
  OBEmpty.032/TRANSFORM_SIMULATION_INIT() via 'Simulation -> Final Transform'
  OBEmpty.032/TRANSFORM_EVAL() via 'Transform Eval -> Simulation Init'
  OBEmpty.032/TRANSFORM_PARENT() via 'Eval'
  OBEmpty.032/TRANSFORM_LOCAL() via 'ObLocal -> ObParent'
  OBEmpty.032/TRANSFORM_INIT() via 'Transform Init'
  OB000hull/TRANSFORM_FINAL() via 'Parent'
  OB000hull/RIGIDBODY_TRANSFORM_COPY() via 'Rigidbody Sync -> Transform Final'
  SCScene/RIGIDBODY_SIM() via 'Rigidbody Sim Eval -> RBO Sync'

So #87085 is related (as is the fix for forcefields f4d8382c86)
And in a bold move to rectify this, if I up the Target Velocity & Max Impulse in the motor constraint by a factor of 81 (see #87085 for the reasoning), I am getting comparable results.

Tempted to set this to High prio for a bit, to get input from @ZedDB and @LukasTonne to see if this is something that can be fixed easily. Situation now (that the influence Target Velocity & Max Impulse have on the simulation depends on the substeps) is pretty much broken.

Pretty sure this was caused by 1aa54d4921c2e8d7114f463a940c169ee573f557 OK, so while there are a couple of dependency cycles detected, (such as the following), it was still somewhat working prior to above commit : ``` Dependency cycle detected: SCScene/RIGIDBODY_SIM() depends on OBCylinder.016/TRANSFORM_SIMULATION_INIT() via 'Object Transform -> Rigidbody Sim Eval' OBCylinder.016/TRANSFORM_EVAL() via 'Transform Eval -> Simulation Init' OBCylinder.016/TRANSFORM_CONSTRAINTS() via 'Constraint -> Transform Eval' OBEmpty.032/TRANSFORM_FINAL() via 'Limit Distance' OBEmpty.032/TRANSFORM_SIMULATION_INIT() via 'Simulation -> Final Transform' OBEmpty.032/TRANSFORM_EVAL() via 'Transform Eval -> Simulation Init' OBEmpty.032/TRANSFORM_PARENT() via 'Eval' OBEmpty.032/TRANSFORM_LOCAL() via 'ObLocal -> ObParent' OBEmpty.032/TRANSFORM_INIT() via 'Transform Init' OB000hull/TRANSFORM_FINAL() via 'Parent' OB000hull/RIGIDBODY_TRANSFORM_COPY() via 'Rigidbody Sync -> Transform Final' SCScene/RIGIDBODY_SIM() via 'Rigidbody Sim Eval -> RBO Sync' ``` So #87085 is related (as is the fix for forcefields f4d8382c86ba760c5fccf7096ca50d6572b208bd) And in a bold move to rectify this, if I up the `Target Velocity` & `Max Impulse` in the motor constraint by a factor of 81 (see #87085 for the reasoning), I am getting comparable results. Tempted to set this to High prio for a bit, to get input from @ZedDB and @LukasTonne to see if this is something that can be fixed easily. Situation now (that the influence `Target Velocity` & `Max Impulse` have on the simulation depends on the substeps) is pretty much broken.

Philipp Oeser added

and removed

labels 2023-10-04 11:00:52 +02:00

Philipp Oeser commented

2023-10-04 11:01:48 +02:00

Tempted to set this to High prio for a bit, to get input from @ZedDB and @LukasTonne to see if this is something that can be fixed easily.

Of course feel free to lower priority again (since this area is not well maintained -- and it is quite an old bug)

> Tempted to set this to High prio for a bit, to get input from @ZedDB and @LukasTonne to see if this is something that can be fixed easily. Of course feel free to lower priority again (since this area is not well maintained -- and it is quite an old bug)

Lukas Tönne commented

2023-10-04 11:44:21 +02:00

I can't even run 2.90 on my linux work machine any more, hard to tell what this would actually look like before 2.91. I don't intend to spend much time on this, but likely that motors also have some kind of update that needs to be repeated each substep, similar to force fields.

Philipp Oeser added

and removed

labels 2023-10-04 12:30:16 +02:00

michael campbell commented

2023-10-04 16:15:50 +02:00

I'm not sure if it's any use, but here's a much simpler version. In this one the motor does begin to work, but after a few iterations it causes all constraints to just explode. I've checked all the margins are small enough, and also increased the steps and substeps quite high. Without the motor constraint no explosions happen. Could it be a conflict of some type between the spring code and the motor code? There are no dependency cycles in this one.

car simple.blend

1.1 MiB

Lukas Tönne commented

2023-10-05 09:56:26 +02:00

I'm pretty sure the instability in the car example is caused by a known issue with the "Generic Spring" (aka 6DoF) constraint: When the angle reaches 90 degrees it goes into gimbal lock and becomes unstable.
https://pybullet.org/Bullet/phpBB3/viewtopic.php?t=12159

I attached a modified version that uses simpler hinge constraints and works just fine.

I'm pretty sure the instability in the car example is caused by a known issue with the "Generic Spring" (aka 6DoF) constraint: When the angle reaches 90 degrees it goes into gimbal lock and becomes unstable. https://pybullet.org/Bullet/phpBB3/viewtopic.php?t=12159 I attached a modified version that uses simpler hinge constraints and works just fine.

car simple2.blend

1.1 MiB

michael campbell commented

2023-10-05 16:40:38 +02:00

Thanks @LukasTonne that's cool but no suspension. I think i've found the cause of the bug though, it works fine if you add the constraints via the object menu, but not when adding them manually. I guess some bit of necessary code is present in the object menu implementation but not when adding the constraints from the empty's physics tab?

Here's the working version created by selecting the wheel then the car body, going to the object -> rigid body -> connect, and choosing spring generic with location selected. The motor was also added from the object menu.

Thanks @LukasTonne that's cool but no suspension. I think i've found the cause of the bug though, it works fine if you add the constraints via the object menu, but not when adding them manually. I guess some bit of necessary code is present in the object menu implementation but not when adding the constraints from the empty's physics tab? Here's the working version created by selecting the wheel then the car body, going to the object -> rigid body -> connect, and choosing spring generic with location selected. The motor was also added from the object menu.

car simple3.blend

929 KiB

Iliya Katushenock added the

Download

What's New

Blender Studio

Manual

Developers Blog

Documentation

Benchmark

Blender Conference

Development Fund

One-time Donations

motor constraint isn't working #113234