I am using Bullet 2.78.
I have a model of a robot skeleton where the spine vertebrae are joined by 6DOF constraints.
The upper part of the model (ribcage, head, arms etc) adds up to a considerable weight and when I run the model it fall right over forwards and the rotational and translational limits I have set on the vertebrae constraints are effectively totally ignored.
One solution is to run at an extremely low timestep, but this is prohibitively slow.
Another solution is to reduce the upper weight significantly but this is a model of a real robot so it needs to use the real values.
I have therefore been attempting to tune the parameters of the constraints so that they are able to resist breaking like this.
Changing the error margin BT_CONSTRAINT_STOP_ERP and the "Constraint force mixing factor when joint is at limit" (whatever that means) BT_CONSTRAINT_STOP_CFM does help somewhat and can stop it crashing to the floor but the actual limits that I set are still mostly ignored.
However, there are a number of other constraint parameters offered in the Bullet code that act on the btRotationalLimitMotor that look like they could help.
These are parameters like m_damping, m_limitSoftness, m_maxLimitForce etc
In the code they are clearly heavily used in the btRotationalLimitMotor::solveAngularLimits() and btTranslationalLimitMotor::solveLinearAxis() methods.
However, although I can certainly set these parameters, they do not affect the behaviour in any way, even when set to crazy values.
Looking further, as far as I can see these methods are never actually called by any other code. This explains the lack of response to their tweaking.
So, my questions are:-
* am I correct to assume that these parameters/methods are actually defunct?
* if so, should they not therefore be removed to prevent confusion to other poor schmucks like me?
* Are there any other parameters or tweaks apart from just the 3 setParam(BT_CONSTRAINT_..) variations that I can use to help get the constraint limits to do what I set them to do? What happened to the bias setting that used to be available?