Preposition is essentially a 2D horizontal positioning of the DHM. It is defined by translating and rotating the pelvis root of the DHM in the horizontal (XY) plane of the global coordinate system. The main idea of prepositioning is to identify a good starting point for the DHM, given the object to Grasp and the surrounding environment. Prepositioning the DHM saves time when predicting the whole-body posture with the SPE, especially if the location of the previous task is far from the current one. In such cases, when the DHM is prepositioned, it does not collide with all the objects between the current and next task. This process is just like moving from one task to the other while skipping the displacement between the tasks, which is outside the scope of the EWD application.

The current preposition algorithm allows to find a collision-free preposition that is as close as possible to the task and allows a good accessibility to the object(s) with the arm(s), but without explicitly solving the whole-body posture. It may also consider the tasks prior and after the current one when they exist within the planning data. Indeed, these tasks may give a clue of where to put the DHM so that the sequence of tasks make more sense (e.g. stay on the same side of a conveyor).

The accessibility is evaluated at two different levels. First, a "whole-body" accessibility is evaluated with a simplified representation (e.g. bounding box, bounding spheres) of the DHM in a standing straight posture without considering the arm geometries. This representation is moved around the object(s) while trying to find a collision-free preposition. The search begins in a circle with a radius equivalent to some predefined measure (e.g. the forearm-hand distance). If no collision-free preposition is found in that circle, the radius is increased and the search is done again. Once one collision-free preposition is found, it is then used to do a refined evaluation of the accessibility between that preposition and the object(s). The space is discretized and an accessibility score is computed, which gives some clue of how clutter is the environment between the current manikin pose and the target object. The algorithm stops when a preposition with a sufficient accessibility score is found. 

The current preposition approach is performed without knowing exactly where the hand(s) will grasp the object(s) and the exact positions of the feet. Other approaches prefer to work the other way around: They first need the grasp locations (using a grasp planner) to estimate position of the feet (e.g. Zhou et al. 2009​​​​​​​). Grasp planners usually separate the hand geometries from the rest of the body when trying to find a grasp location, which prevent them to evaluate the overall accessibility to the object. The approach of the SPE is more focused on that overall accessibility to the object, which may lead to more plausible prepositions and eventually grasp locations. Once the preposition is found, each of the 4 DHM are moved to that location.​​​​​​​