Techniques for "natural" interaction

in multi-user CAVE-like environments


This presentation reports work being done at SICS on novel interface techniques to large projection environments (CAVE-like environments such as the SICS Grotto)intended for multi-user applications. We have been exploring the notion of "naturalness" as it relates to interfaces. We use the term interface in a restricted way, covering the gap between the user and the multi-user virtual reality platform DIVE, i.e. instantiations of different techniques of interaction within a virtual world. The goal is to create interfaces which are compelling while at the same time, more effective and less encumbering than merely extending the standard 2D GUI tools.

Designing such interfaces presents a series of design choices, centered on granting user control, number of degrees of freedom to be presented and how large to make the human-computer 'language.' We have set out to make these interfaces as "natural" as possible. We are far from final conclusions, but are instead presenting here a number of our experiments.

The SICS Distributed Interactive Virtual Environment (DIVE) is an experimental platform for the development of virtual environments, user interfaces and applications based on shared 3D synthetic environments. Dive is especially tuned to multi-user applications, where several networked participants interact over an internet.


# Scrolling Viewpoint

Navigating  with the full six degrees of freedom has shown to be difficult for people not accustomed to a 3D mouse. When using the scrolling mechanism, navigation is restricted to a horizontal plane (often used with gravity). The avatar, and thus the viewpoint, is moved forward or backward when the cursor, controlled by pointing with the 3D mouse, is located above or below the three screens. The travel velocity is related to the distance the user points outside the screen boarders. In a similar way, the user rotates by pointing off the screen to the side.

# Ball Navigation

The  ball navigation technique is implemented using a magnetic tracker attached to one side of the ball (an ergonomic sitting ball, approximately 80 cm in diameter). A user sits on the ball and leans in the desired direction. At the same time as the user is  navigating, the user can employ the 3D mouse to select and grasp objects. This is an attempt to reduce the workload of the hand. This navigation method seems to be intuitve as it appeals to an everyday notion of locomation as well as being less fatiguing than other more active mechanisms.

# Flying Cursor

In order to give a greater feeling of depth, we have implemented a "flying cursor". The flying cursor, a flat partly transparent hexagon, is placed on the surface of an object that is being pointed at, and tilted according to the normal of the surface. This provides extra spatial cues, allowing the user to 'feel' the surface of an object as well as get another depth cue through perspective.

# Unencumbered Video Tracking

This is perhaps the most powerful and least encumbering of all the devices. Having used magnetic trackers for most interaction techniques, one realizes that it would be an improvement to unencumber the user of cabling and sensors Toward this end we have implemented a navigation technique that uses the Pfinder system developed at MIT Media Lab. Pfinder extracts information from video images of a person in a scene, and gives 2D+ data about the location of the hands, head, and center of the person's body. This information can be used to navigate and manipulate objects in a VE. We have implemented a way of navigating using a simple body gesture command language. Raising your arms increases the speed, lowering one of your arms makes you turn in that direction. This navigation is based on the metaphor of "gliding", much the way a gull might turn and bank in the air.

# Trampoline Surfing

The trampoline navigation interface uses an ordinary workout trampoline with a magnetic tracker attached to the bottom. It is controlled in a similar way to the ball and is an attempt to take some of the interaction away from the hand. The user stands on the trampoline and leans towards the direction he wants to move. This technique has, as well as the ball technique, proven to be intuitive (almost like a surfboard/skateboard). However, the current implementation has proven quite hard to characterize and can therefore be straining for the leg muscles as it works with very small angle changes. A disadvantage of this interface is that it is very sensitive to natural changes in position, like changing the weight of your body to another foot.

# Path Planning Map

In a virtual environment, one way to let the user gain a larger context of the environment is by using the WIM (World In Miniature) metaphor, which in our implementation is called a Path Planning Map (PPM). This provides the user with a hand held miniature representation of the world, somewhat like a map in the real world. The virtual map metaphor benefits from the notion of everyday maps, something most people are familiar with, and thus, a concept that can be grasped quickly. The PPM can be used for locating and orienting the user, searching for other users, and for moving directly to a point on the map. 

One mode lets the user draw a path on the map, which will be replicated in the virtual environment and displayed to those that choose to see it.  A user can follow the path freely or be taken on the path automatically. A tracker can also be attached to a physical plate directly linked to the virtual map, allowing the user to hold a real representation of the map at the same time as controlling the virtual PPM.   [This work is done in conjuction with Emmanuel Frécon.]

# Tank Navigation

When constructing this interface we looked at how a tank driver steers a tank. The two wands control the speed of the virtual tracks. Magnetic trackers are attached to two sticks to simulate tank navigation. Rotating both sticks forward will give you a forward velocity. Rotating the left stick back makes you turn left. The directional velocity is determined by averaging the sticks rotation and the angular velocity is computed by looking at their rotational difference. An advantage of this interface, apart from being intuitive, is that it only uses 1 degree of freedom (pitch) which prevents a beginner from producing unexpected movements. The fewer degrees of freedom an interface uses, the shorter the learning period needed since the 'vocabulary' is limited. However, a very limited 'vocabulary' prevents use of the interface in a general sense. The 'Tank' interface is appropriate for navigational purposes. Using the base mechanism of the angular control sticks, the tank vehicle is extendable to other vehicles such as a virtual fork lift. The shifting of mode would depend on the application.

#Touch Desktop

To achieve more direct manipulation and more natural environments (in the sense of manifest affordances), we are working on novel projection displays in connection with touch screens as a method of achieving alternatives to 'fishbowl' interfaces. The above photos illustrates one such setup where we have a horizontal 'desktop' with a touch screen interface, and a vertical 'communication wall' that we can use for conferencing. Both surfaces provide an interface into the virtual environment/workspace.  The horizontal projection surface also  provides an elegant mechanism for local groupwork.  In the image on the right, the desktop is used in conjunction with the multi-screen display to provide a view of the virtual desktop which is located in the VE. Objects on the desktop can be manipulated by direct touch.

# Cyber Monkey

A tracker is placed on a puppet's body and another in it's right arm. The puppet is then used as a control agent in the VE.  In navigation mode, by turning the puppet, the avatar's position, and thus view, changes accordingly. The tracker in its arm is used for specifying direction, velocity or manipulating objects. A microphone, placed in the puppet's ear and used with voice recognition, and a speaker in its mouth, encourage the illusion of an embodied virtual agent. There are many extension applications that can build on the puppet-agent technology - such as a tour guide in the VE. When picking up the puppet in the CAVE-environment,  it would know what world you're in (and where you are), you could then ask it to take you where you want to go. Along the way, the puppet¹s arm could be used to point at interesting objects and questions could be posed.


Par Hansson / Anders Wallberg