btMultiBodyGearConstraint::setGearRatio doing nothing

[qhdwight]

Hello,

I just switched my robot simulation to use multibodies (Featherstone). So far it has been working MUCH better! However I am running into some issues porting over the gears.

I have an axle (index 0) connected to a wheel (index 1). The linkage between them is revolute.

The axle (0) has a btMultiBodyJointMotor on it. There is a btMultiBodyGearConstraint connecting 0 (parent, axle) to 1 (child, wheel). When I spin the axle the wheel moves. But if I set the gear ratio to -1 they still move the same way. Furthermore, if I increase the ratio to say 50 (which it should be in my example), they still rotate at the same speed.

Not sure what I am doing wrong. I am using the racecar_differential.py example as reference. I make sure to only make my constraints after finalizing all links first.

Link to code: https://github.com/qhdwight/mrover/blob/073b0650cdf7728089d8db93036bc0b49c2a84ae/src/simulator/urdf.cpp#L201
Rover URDF: https://github.com/qhdwight/mrover/blob/sim-multibody/urdf/rover/rover.urdf.xacro

I am using Bullet 2.88 provided by Ubuntu 20.04. I also tried compiling and linking 3.25 from source but that also did not work.

[qhdwight]

@erwincoumans I also noticed that the constructor to btMultiBodyGearConstraint ignores the last four arguments. I see how they are not needed for a gear constraint but I believe it is a bit misleading.

[qhdwight]

For example here is a gearing ratio of -50. You can see that the wheel and axle (smaller cylinder) rotate at the same rate and in the same direction.

simplescreenrecorder-2024-01-04_14.37.39.mp4

[qhdwight]

Update... I tried a simpler model with two cylinders connected via a revolute joint. Can't get them to rotate with the proper ratio if they are coplanar. For some reason the mimic joints example only works if they are orthogonal. Not sure what is going on.