Effect of planning on multi-touch grasps

The manipulation of virtual objects has a central role in interaction with table-tops. For example, users move and rotate documents and pictures around the sur-face to share them with other users. Graphical designers manipulate information and graphical objects to create new content. Multiple users work collaboratively to create schedules, make decisions, or solve complex problems. In all these sce-narios, users interact with their hands and their fingers; they grasp, translate, and rotate virtual documents as they would do with physical objects.

Researchers in Psychology have studied how people manipulate objects in the physical world. This work has unveiled that people show strong signs of prospec-tive motor planning, i.e., they choose initial grasps that avoid uncomfortable end postures and facilitate object manipulation. A very recent model that considered continuous tasks, such as rotating a physical knob, has been proposed by Herbort [HB12]. This model, which appeared to us as especially relevant to the continuous manipulation of virtual objects, argues that there are various biases that influence people’s initial grasp of an object. In its simplest form, the model can be expressed as follows:

p_initial= w_anti· panti+ w_{def ault}· pdef ault

w_anti+ w_{def ault} (3.1)

According to the model, two different biases contribute to the initial grasp ori-entation pinitial. An anticipatory bias pulls the initial grasp toward a pronated or supinated angular position panti, depending on the intended direction of rotation. A

64 GESTURES

r

Figure 3.16 : The position of interactive objects expressed in polar coordinates (r, θ), where r is the radial distance and θ is the clockwise angle with respect to the vertical axis of the screen. The grip orientation is expressed by the clockwise angle φ defined by the thumb and the index.

second bias pulls the initial grasp toward a preferred task-independent orientation p_{def ault}. The contributing weights wanti and wdef ault of the two biases can vary, for example, depending on the difficulty of the task or the required end precision.

In [18], we describe a series of experiments that test if such prospective motor control also takes place in the context of virtual objects manipulated on a tabletop.

In those experiments, participants were asked to grab a circular start object with the thumb and the index of the right hand, and to manipulate it to make its position and orientation match another circular target object.

In order to test if there is a default bias, we considered different locations for the start object. Regarding the potential effect of an anticipatory bias, we considered two conditions differing in whether users could plan their movements or not:

· target is displayed before participants grab the start (Known Target, planning possible), or

· target is displayed after participants grab the start (Hidden Target, planning not possible).

Figure 3.16 illustrates the six screen positions for both the start and the target objects that we tested. One was located close to the user, centered on the vertical axis of the display, 35 mm from the front edge. We refer to it as the User position.

The other five positions were located around the User position with an angular position θstart of −90^◦, −45^◦, 0^◦, 45^◦, and 90^◦, and a radial distance of r = 314 mm. Our main measure was participants’ grip orientation φinit (i.e., the angle defined by the thumb and index when participants grab the object).

Default bias In order to observe if there is a preferred task-independent way of grabbing an object (default bias), we considered the trials where the start and target configurations were the same. In such trials, the start and the target objects were

Gesturing with multiple fingers 65

Figure 3.17 : default bias. Effect of start position θstarton φdef ault

5 0

Figure 3.18 : anticipatory bias. (Left) Effect of target angular positions θtarget

on the grip orientation φinitfor outward translations. (Right) Effect of the rotation angle β on φinit

appearing at the same position, and the participant had to hold the start object and keep it inside the target. Figure 3.17 reports participants’ grip orientation for those trials φdef ault, and clearly illustrates that φdef ault varied along different angular positions. This supports the fact that, like with physical objects, there is a default bias when manipulating virtual objects.

Anticipatory bias Our observations also corroborate the fact that there is an an-ticipatory bias when manipulating virtual objects on a tabletop. For example, Fig-ure 3.18-(Left) illustrates that φinitis different depending on whether participants were able to plan their manipulation or not. In the presence of translations, the anticipatory bias is dependent on the φdef ault orientation of the target position. In the presence of rotations, the bias is dependent on the direction βdirand the range β of the rotation (Figure 3.18-(Right)). When both rotations and translations are combined, both these factors contribute to φinitbut with different weights.

Overall, our results support that there is some prospective planning when people perform translations and rotations of virtual objects on a tabletop. We have shown that users choose a grip orientation that is influenced by three factors: (1) a pre-ferred orientation defined by the start object position, (2) a prepre-ferred orientation

66 GESTURES

defined by the target object position, and (3) the anticipated object rotation. We have also shown that our results are consistent with Herbort’s WIMB model for physical objects [HB12].

These results suggest that we could possibly infer information about users’

prospective movement to improve user experience during the manipulation phase.

Interface designers could, for example, develop techniques that adapt their graph-ical layout to improve visual feedback, avoid potential occlusion issues [HB04, VC12] or reduce interference [HMDR08] when multiple users interact in close proximity in collaborative settings. We could also derive directions about how to design grips and visual guides to facilitate both the acquisition and the manipula-tion of virtual objects.