search

robotology-legacy / mex-wholebodymodel

Matlab MEX interface to the iWholeBodyModel interface.
11 stars 9 forks source link

[interface] Base part of wbm_getState is broken and needs tests #47

Closed traversaro closed 8 years ago

traversaro commented 8 years ago

In particular, the base state was undefined because it was taken from the returned value of a function that was not returning anything (see https://github.com/robotology/mex-wholebodymodel/commit/0437e086ba93754111f219b1a86aae8113722958#commitcomment-13937981 ). I guess the function is always returning the identity transform in practice. Fixed in https://github.com/robotology/mex-wholebodymodel/commit/0437e086ba93754111f219b1a86aae8113722958 , still need to be fixed in master.

traversaro commented 8 years ago

cc @naveenoid @gabrielenava

traversaro commented 8 years ago

To make sure that everything is properly working, I suggest that we do some consistency tests together with wbm_setWorldFrame,wbm_updateState and wbm_forwardKinematics .

traversaro commented 8 years ago

Anyway the effect of the simulations is limited, as getState was used only to get the initial configuration of the base.

gabrielenava commented 8 years ago

I don't think I have really understand this problem. I mean, in the balancing controller wbm_getState is used only at the beginning and only to obtain the identity since wbm_setWorldLink sets exactly the identity transform. But if I put wbm_getState in forwardDynamics just to see what happens the quaternion returned by getState changes, it's still a valid quaternion and the associated rotation matrix is not the identity so I don't think I understand this problem properly

traversaro commented 8 years ago

I'll explain what I found in the code: wbm_getState calls (with some magic ) ModelGetState::computeFast that itself calls modelState->getRootWorldRotoTranslation() to get the rototranslation, but this function is not returning anything!

The fact that you mention wbm_setWorldLink (that should not be used in simulation, see https://github.com/robotology/mex-wholebodymodel/issues/43) make me wonder if perhaps ModelGetState::compute is called, and that function seems is not affected by the bug.

Anyway a general comment: the fact that a computed quantity changes and is not the defuault does not mean that it is correctly computed!

naveenoid commented 8 years ago

Tackled in https://github.com/robotology/mex-wholebodymodel/pull/52