setting revolute joint position?

[mattdm]

Rapier3d joints have "motors", which ... aren't working for me like I think they ought to, so maybe I'm just really confused overall. But what I want is: a revolute joint which can swivel under normal influences subject to its limits but returns (with configurable torque and dampening) to a set point. I'd like to be able to change that set point dynamically -- causing a turret to swivel, or a head to turn, or a door to open or close.

However, xpbd does not have motors. There is position_lagrange and align_lagrange, but even after reading the theory section I'm not really clear on what these are or why I would want to set them... and anyway this doesn't seem like it.

I guess I could temporarily reduce the angle limits, and then reset them? That doesn't seem right, though. What am I missing?

[Jondolf]

I don't think you can make proper motors right now, it's something that will be added in the future. You might be able to cheat it by changing the angle limits, but I doubt it would be very robust or easy to control

[mattdm]

I am super-naïve here, having little real physics background or experience with other game engines. So, feel free to back away slowly if this is all completely off-base.

I see that RevoluteJoint has align_orientation (well, I guess all joints do, but... starting there)... until there are motors, to make a system that has the same effect, would that be the function to use? That takes a delta_q ... is that angle to try to rotate along each axis? If so, what is this vector aligned to (hopefully not world x,y,z axes!)? Or, is it a torque vector? Or something else entirely?

This function also takes "lagrange" and "compliance" -- aren't these already properties of the joint itself, though? I see that it returns torque — I'm having trouble thinking of things from this (metaphorical) perspective — I'm thinking of a joint as a thing which applies force to the attached bodies to cause them to align in a certain way, and maybe that's all wrong.

Maybe a motor shouldn't be part of a joint at all, but an entirely separate thing which:

• attaches to a rigid body
• applies force or torque to any other rigid body in contact at that point
  • taking into account friction between those bodies
    • possibly with a parameter to "lock" and always use an infinite coefficient of friction, so there is no slip
• and then the bodies would follow the physics model, including being held by any joints

For convenience, it would be of course nice to have a way to perfectly align a joint and motor.

I'm thinking mostly of rotations, where I think the "motor" would be a point (although, interesting if it could be a circle, cylinder, or sphere, actually...). For linear motors, I would expect a line in 2D or rectangle (or other shape?) in 3D. I'm not sure there's a real-world equivalent, but for a game, this would make it easy to make those speed-up ramps in Mario Kart, or moving sidewalks without actually moving a platform.

[mattdm]

Leaving aside motor ideas in xbpd for a second... if align_orientation isn't the right approach, my other idea is to make a system which:

• looks for entities with RigidBody and some component like "Servomotor"
• reads an angle from that
• and either also an entity id or finds a joint and gets the other RigidBody entity from that
• gets the rotational angle between the RigidBody with the Servomotor and the other RigidBody
• if the difference is bigger than some configured precision, and an ExternalForce with torque in the right direction to the other RigidBody
• possibly scale the torque to the delta? I dunno. I will have to experiment.

This seems more straightforward to me, but less... integrated. I'd love to hear opinions on if this is close to sensible. :)
*

[RunnersNum40]

I see that motors aren't implemented yet, but in the interim does anyone have a good way to achieve the same functionality?

I'm curious if there's a possible approach of applying torques and forces to the entities themselves.