Implications from Human Motor Control

The geometric relation of an object with the surrounding environment is only one aspect that affects movement planning and coordination during its manipulation. Like all human movements, object manipulation requires control over multiple limbs, each with several degrees of freedom and particular constraints. As a result, many different movements can be performed in order to achieve the same goal. This free- dom is certainly an advantage as it facilitates the avoidance of obstacles, however, it also complicates control. The neurophysiologist Nikolai Bernstein highlighted the “degrees of freedom problem” in human motor control [33] and inspired many subse- quent researchers to explore this issue. One plausible solution is the cost containment theory (see [290] for an introduction). The theory states that human motor control aims to to minimize various costs related to postures at the start, the end, and during a movement. Rosenbaum et al. suggested that depending on the situation different weights are assigned to the costs that can be relevant to the required motor perfor- mance [290]. These weights allow the actor to put an emphasis on accuracy, speed, collision avoidance, style, or any other factor that may be considered relevant.

Rosenbaum et al. proposed internal representations of posture as a basic building block of human motor planning, because these allow simpler internal representa- tions than trajectories [290, p.178]. This view is supported by studies demonstrating that humans have difficulties in memorizing movement, while postures can be easily recalled. Moreover, postures can be specified in terms of equilibrium points for the muscles, i.e., “a set of muscle lengths for which muscle tensions balance out”. Rosen- baum et al. argue, that “when an equilibrium point is specified and the starting point is known, the trajectory to the equilibrium point comes for free, making detailed plan- ning of the trajectory unnecessary.” They also refer to the “end-state comfort effect” which describes a behavioral tendency to grasp objects in an uncomfortable posture in order to achieve a comfortable posture at the end of the manipulation process. A bottle lying on the floor, for example, will generally be taken with an underhand grip in order to facilitate its rotation to an upright orientation (Figure 3.2 a). Therefore, simultaneous rotation and translation of objects may be inefficient in terms of motion trajectories in world coordinates, but it may be the result of efficient limb coarticu- lation (see also [286, p.22]). If computer interfaces do not enable similar handling of virtual objects, our motor planning and operative skills may be impaired.

The central nervous system seems to exploit various strategies to simplify effective motor control (see [34] for an overview). The choice of the most suitable frames of references for motion planning and operation is one example. Berthoz mentioned egocentric and allocentric (related to external coordinates) reference systems. He em- phasized the role of gravity as a reliable natural reference and argued that our brain exploits reference frames connected to the limbs in order to simplify motion control. The action of pointing towards a remote target, for example, can be simplified to the control of two polar coordinates centered in our shoulder joint [34, p.107]. Moving an object with a similar strategy implicitly involves its rotation relative to an exter- nal coordinate system, while it remains stable in relation to the operating hand and arm. If further variables such as the distance or the orientation of an object in the hand should be kept stable, kinematic constraints provide implicit coordination of the limbs. One of these internal control loops is the movement of connected limbs in phasic opposition to satisfy a certain motion constraint, e.g. drawing a straight line: “When the angle of the arm increases in relation to the body, the angle of the arm in relation to the forearm decreases by an equal amount” [34, p.144]. These insights in- dicate that human motor control may facilitate simultaneous rotation and translation as a byproduct of maintaining relations among the involved limbs.

The manipulation of objects is not only a mechanical task, but it requires cognitive effort for planning the action and monitoring its execution. This can be particularly difficult for 3D rotation. Translational movement of objects along the shortest path across 3D space can be readily imagined. Understanding the most efficient 3D rota- tion path, instead, has been proven to be difficult [257]. The same studies of Parsons also showed that the mental process is considerably simplified if the rotation axis co- incides with one of the principal axes of the egocentric reference frame or the object’s shape (see mental rotation tasks in Figure 2.4). Note that the handling of everyday

Integrality and Separability in Manipulation Tasks 37 objects primarily involves rotations of this type – and above all rotations about the vertical axis of gravity. When actively handling objects instead of only thinking about it, the task seems to be alleviated by continuous visual and proprioceptive feedback of the manual process and the continuously changing state of the object (see for exam- ple [378]). Mundane objects furthermore have constraints embedded in their shape and their distribution of mass. We exploit gravity to let objects swing into new ori- entations and let them align through collisions with planar surfaces like tabletops. Handling real-world objects is an iterative learning process that eventually enables playful interactions as in balancing, throwing and juggling.