humanoid-path-planner / hpp-model Goto Github PK

In Body::computeDistances, the global placement (Transform3f) of the body is explicitly recomputed, while it is already computed in the forward kinematics (Device::computeForwardKinematics). So I guess that the exact same computation is run twice. Is my understanding correct?

Is anybody using the body member functions to modify the inner object list, outside of the URDF parsing function? Should I keep these accessors in the new hpp-pinocchio API?

The currentTransformation of the rootJoint of a device initializaed with a "freeflyer" exhibits an identity rotation for a non-1,0,0,0 quaternion. Could you explain or check the behavior?

E.g. for romeo:
q = -0.541726 0.40124 -0.366266 0.533858 -0.436099 -0.691464 0.216079 -0.552687 0.302264 0.0213719 0.942931 -0.439916 0.0922138 0.438537 -0.773439 -0.0570331 0.18508 0.888636 -0.0981649 -0.327298 0.695369 -0.130973 -0.993537 -0.310114 0.196963 0.666487 -0.532217 0.350952 -0.0340994 -0.0361284 -0.390089 0.424175 -0.634888 0.243646 -0.918271 -0.172033 0.391968 0.347873

Placement from Pinocchio:
R =
-0.0496261 0.372383 -0.926751
0.833805 0.526255 0.166809
0.549825 -0.764451 -0.336611
p = -0.541726 0.40124 -0.366266

Placement from hpp::model
rotation matrix: [(1 0 0)(0 1 0)(0 0 1)]
rotation quaternion: (1, 0, 0, 0)
translation: (-0.541726 0.40124 -0.366266)

In class Device, methods are implemented to add and remove collision pairs, and to access the list of collision pairs. Collision pairs are pairs of joints (or equivalently bodies). When a pair is added, all the inner objects of joint 1 are added as outer objects of joint 2. The information contained in the pairs is somehow duplicated in the inner/outer lists: we can remove some outer objects without removing the corresponding pair of joints. Device::CollisionPair may then be outdated without any trouble.

Pinocchio does not maintain such a CollisionPairs structure.

I believe that this structure of hpp::model is weakly used, if not at all. Should I try to mimic it in hpp-pinocchio? Or can we consider this function as deprecated?

In hpp::model, two lists are kept for pairs of object that must be tested for collision, and those for which distances must be computed. Is it important to keep these two sets explicitly? Or can we have only one set, while either collisions or distances are computed by calling the respective method?

Is anybody using a model where the two lists InnerObjects(DISTANCE|COLLISION) are different (and resp. the two lists outerObjects(DISTANCE|COLLISION) are different)?