Before solving the whole body posture, the SPE must know where to put the hand(s) on the object(s) and which Grasp mode to use (e.g. precision vs. power grasp). First, the bounding box (BB) of each the object will be computed. This BB will be used to label the different faces (front, back, left, right, bottom, top) of the objects, based on the chosen preposition.

The grasp mode and location will first depend whether the object is a tool or an arbitrary part and if one or two hands are used on the same object. An arbitrary part will be grasped by one of the faces of its bounding box (BB). When grasping an object with one hand, the choice of the face to grasp and the grasp mode both depend on the object's BB dimensions, on the accessibility to its faces and on the object's weight. The different grasp modes and their underlying characteristics (e.g. grasp size and weight) used by the SPE are based on the inspiring work of Feix et al. 2016.

Here is an example of how grasp mode is determined: If a given object face (e.g. top) is accessible, sufficiently small and the object is sufficiently lightweight to allow a precision grasp with the fingertips, this grasp mode will be used. The rationale is that precision grasp with the fingertips allows to reach farther, whether enveloping grasp involving all the fingers and the palm reduce the reachability of the upper limb. If the same power grasp would always be used, the result could be that the object is too far to be grasped, although a tip pinch would allow to reach the object.

This example shows how the choice of grasp mode is determinant in providing relevant information to the user to evaluate its design for reachability. Of course, if the object is too heavy, it is not plausible to use a pinch grasp. Thus, when the user also provides the real weight of the object and this weight is higher than the weight allowed for a pinch grasp, a precision grasp with more fingertips (e.g. precision cylinder with 5 fingers) or a power enveloping grasp may be chosen to ensure posture plausibility. As a general guidelines, we suggest to provide the real weight of the object, unless the user is comfortable with the resulting DHM posture or knows that the object is sufficiently lightweight (e.g. max weight of pinch grasp is around 250g, Feix et al. 2016). If the object is too big to be grasped by any of its size, a default side (e.g. top) may be chosen, but the hand closure may not be satisfactory. If this happen, the user can always decide to modify the worker task to grasp the object with both hands, in which case the grasp locations will be opposed (e.g. left/right or front/back).

Tools are a special kind of objects that afford specific grasp modes and locations. They are grasped using dedicated processes (see Patents, Ref [5] for an example of such processes). The EWD application allows to grasp tools with one hand. The action of grasping a single tool with both hands has been added to the list of features to be implemented.

Once the SPE knows the grasp mode(s) and location(s) for the used hand(s), two frames are defined for each hand-object couple. The first frame is defined on (or close to) the hand. Its exact location is determined through non-linear regression, based on the grasp mode and on the grasp dimension. The position and orientation of that frame with respect to the wrist is fixed. For arbitrary parts, the second frame is defined at the center of the grasped BB face, in such a way that the palm will be somehow parallel to that face. For tools, the second frame is defined using the grasping cues presented in Patents, Ref [5]. An example of grasp frames determination from grasping cues can be found in Patents, Ref [6]. Once the gasp modes and locations are found, the whole-body posture can be computed by matching the frames on the hand(s) to those defined on the object(s).