Phase 1
Home Project Overview People Resources UCL Code

 

Up
Scenario 1
Scenario 2
Scenario 3

Phase 1 - Design and Analysis of Interaction Metaphors for the Single User

Contents

Overview

Summary

Qualitative Differences in Interaction in IPTs

Experiment Design

Movies, Pictures and Other Resources

Results

Overview

This was the first experimental phase of the React\Interact project, and its aims were twofold:-

 

  1. Firstly, we wanted to implement some common interaction techniques. At this initial stage we simply built the two most popular selection/manipulation techniques - Virtual Hand and Ray Casting

  2. Secondly, we wished to show experimentally that the performance of common interaction techniques can vary significantly when used in different VR systems (say an IPT vs a HMD). given this we may argue that designing interaction techniques for IPTs based on techniques developed for other types of immersive systems is a flawed approach.

Summary

An experiment was undertaken January 2003 to explore how common interaction metaphors would perform on the IPT [21 ]. The experiment compared HMD and IPT systems and the virtual hand and ray-based selection metaphors [2 ,8 ,12 ] on a range of selection and manipulation tasks

Forty subjects took part in the experiment. It was a between-groups design with twenty experiencing the HMD, twenty the IPT. Each person used both the virtual hand and the ray-casting technique. Each undertook five selection and manipulation tasks. The order of tasks was balanced.

The first result was that on close-by manipulation tasks the HMD and IPT were very close in performance. Ray-casting was significantly faster on the IPT than virtual hand on the IPT, but there were no other differences. Even this difference was small – roughly 15% in mean time on tasks taking 25-30 seconds.

The second result was that on selection tasks, the IPT was far superior with both interaction styles. Selection is much faster that manipulation, taking 2-3 seconds per selection. However the difference now was 25% and highly significant. The variance of the two was remarkably different, with selection on an IPT being relatively constant time compared to the IPT. The HMD user would often spend a long time looking for objects in near or far space which were eccentrically placed.

The overall conclusion was that for close-by interaction there was little to choose between the two displays, but for simulations of surrounding environments, the IPT would be preferred.

Qualitative Differences in Interaction in IPTs

At the moment, interaction techniques used in IPTs are derived from those used in HMDs. A crucial result from previous research into interaction techniques is that the interaction methods depend critically on the actual input devices and display methods chosen. This is obvious at some levels since a desktop interface simply affords different degrees of control than an immersed user.  It is therefore our contention that interface methods for IPT systems will need to be designed and studied independently since there are qualitative differences in atomic interaction between IPT and HMD.

Our experience in the UCL ReaCTor has shown that interaction metaphors for IPTs should not necessarily be the same as those proposed for HMD systems. There are several qualitative differences.

  1. An IPT is a multi-user system. However only one user can be tracked, and thus the view is only correct for this user, and only this user can interact with the virtual environment.

  2. Since the user can see their own body and can't see their own virtual body, they can not see the virtual object that is used for selection and manipulation (commonly collision detection with the user's virtual hand indicates selection).  A side-effect of this is that if there is significant tracker lag the user might mistakenly believe that selection is effected by touching an object with their real hand, since they can not see the virtual hand that is the system's most up to date reading of the user's hand position and the selection point. NOTE:- in experiments we discovered visual feedback was needed in IPT in form of a virtual hand to alleviate the above problem. And in later experiments even greater feedback was needed in form of highlighting potentially selected object.

  3. The field of view is much larger, and thus interaction might occur with objects in the periphery of vision. Feedback techniques might need to be more explicit to confirm selection.

  4. Since the most common configuration of an IPT has only three walls, the virtual environment does not wrap completely around. Thus a joystick is usually provided whose primary purpose is to rotate the IPT about the user in virtual coordinates. A two-axis joystick is usually used, so the second dimension is usually mapped to locomotion forwards and backwards. However we have observed users forgetting to turn using the joystick and turning so far that they face a blank wall at which point they remember to use the joystick. Note that a similar problems occurs with locomotion both in the HMD and IPT. Both operate in a constrained space with HMD users testing the length of the tethered cables before using the locomotion metaphor, and IPT users colliding with walls.  

Experiment Design

This first experiment was designed to explore the differences between our Reactor IPT and a typical HMD with a single tracked user for the reasons outlined above. To achieve this we designed a set of trials involving various interaction tasks which subjects must perform in a virtual environment. 

These trials involved 2 popular interaction techniques:-

Virtual Hand 

Ray Casting

and 3 separate scenarios in which the user must perform certain tasks:-

Scenario 1 - Medium Space Selection and Manipulation.

Scenario 2 - Near Space Selection.

Scenario 3 - Medium Space Selection.

See the Investigators script for the exact ordering of the trials performed by each subject.

40 subjects completed the trials, 20 performed the trials in our in Reactor IPT and 20 in the HMD. (get techs of machines, trackers HMD used). 

Movies, Pictures and Other Resources 

Movies

QuickTime movie of Virtual Hand User in Reactor IPT (6.2mb)

QuickTime movie of RayCasting User in Reactor IPT (1.4mb)

QuickTime movie of Virtual Hand User in HMD (4.6mb)

Pictures

Picture 1 - User in Reactor IPT, using Ray Casting interaction technique and performing Scenario 1 selection Manipulation of Teapots.

Picture 2 - User in HMD, using Virtual Hand interaction technique and performing Scenario 1 selection Manipulation of Teapots.

Other Resources

Investigators Script

Description of Scenarios

Information given to Reactor IPT user

Information given to HMD user

Results

See individual scenario descriptions for results

Scenario 1, Scenario 2, Scenario 3