CITEC Central Lab Facilities
CITEC
British Flag   German Flag

Lean-based Velocity

Type: Navigation
Authors: Robert Abel, Andreas Jagel, Katharina Klein
Supervisors: Nikita Mattar, Thies Pfeiffer

Original Design

User Perspective

Tolksdorf et al.[3] describe an empirical study of different modes of travel using the Nintendo Wii Balance Board. They explored two different methods of navigating the 3d environment ― one velocity-based and one acceleration-based input method.
Using the velocity-based method, user input leaning forwards and backwards would translate to their avatar's velocity in the virtual environment, thus leaning forwards and backwards would move their avatar forwards and backwards respectively.
The acceleration-based method would alter the avatar's acceleration. Thus, the user leaning forwards would accelerate the avatar, while leaning backwards would decelerate it. Therefore standing still would not impact the avatar's velocity, which is somewhat counter-intuitive and therefore requires a longer adaption phase.
Both methods let the user change his direction of movement by leaning sideways, thus rotating their avatar's view around its own up axis.

Technical Details

The Nintendo Wii Balance Board features four pressure sensors ― one for each corner. [3] introduced a so-called user vector which was calculated as a function of the four sensor inputs. This way it was possible to get the gist of the user's intended motion without the need of the user to exercise very precise movements.
The sigmoid functions used to calculate the user vector were chosen so that small movements would not alter the user vector drastically while still preserving the swiftness of the users' perceived motion. So moving at higher speeds, rapidly changing directions or moving for a long period of time would still be possible and not strenous.

Our Realization

User Perspective

Just like in [3]'s velocity-based implementation, we let the users steer by shifting their weight to either side and actually move with forwards and backwards leaning motions.
While [3] concluded that the velocity-based method was impractical for long periods of movement, we found that the supplied supermarket 3d scene was small enough for velocity-based movement. The tasks of getting the items have periods of movement followed by periods of selection and manipulation. So in our implementation the user should not need to pay attention to actively not moving while selecting and vice versa.

Technical Details

We worked with the supplied user vector as per [3]. We used the methods described in Mackinlay et al.[2] to determine whether the avatar's motions would still behave intuitively while the user was busy using other devices such as the Nintendo WiiMote or tracking gloves in our CAVE for selection and manipulation purposes. We restrained our model accordingly to enable selection tasks to be perfomed without inadvertent movements.
We thought of implementing a 2d Gaussian surface akin to LaViola et al.[1] for scaling the rotation and velocity outputs, but found its behavior too unintuitive due to the inherent weight-dependency of the Wii Balance Board sensor output.
Instead, we opted for restraining the user vector, which itself is not weight-dependent. We projected the user vector onto the xz plane to determine the users' wish to either rotate or move. Rotation was directly calculated from the degrees on the xz plane, while movement was calculated as a function of the angle between normal vector and user vector.
We set minimum and maximum angles of leaning to either side, so the avatar would behave reasonable while standing still, getting on and off and moving. We compensated for the fact that users were usually able to lean forwards much easier than backwards with different angles.
The user vector itself incorporates an acceleration threshold, so there was still non-linear behavior present. This was expected by most users, i.e. leaning farther should result in faster movement.

Evaluation (details here)

Easiness:★★★☆☆(3.0)
Fatigue:★★★☆☆(3.43)
Fun:★★★☆☆(3.43)
Learnability:★★★★☆(3.57)
Raslism:★★★☆☆(2.71)
This evaluations was done with seven probands who tested lean-based velocity using the image plane selection method.

References

  1. LaViola, Joseph J., Jr. et al. "Hands-Free Multi-Scale Navigation in Virtual Environments." Proceedings of the 2001 symposium on Interactive 3D graphics (2001): 9 - 15. Print.
  2. Mackinlay, J. D., S. K. Card, and G. G. Robertson. "A Semantic Analysis ofthe Design Space of Input Devices." Human-computer interaction 5 (1990): 140 - 190. Print.
  3. Tolksdorf, J. et al. "Navigation in Virtual Reality with the Wii Balance Board." 6th Workshop on Virtual and Augmented Reality. (2009). Web.