An Intelligent 3D User Interface Adapting to User Control Behaviors

1
An Intelligent 3D User Interface Adapting to
User Control Behaviors

• Tsai-Yen Li and Shu-Wei HsuComputer Science
  Department
• National Chengchi University
• Taipei, Taiwan, R.O.C.

2
Outline

• Introduction
• Intelligent navigation interface
• How virtual force works
• Customizing virtual force field
• Simulation experiments
• Dynamic adjustment
• Evaluation
• Conclusions

3
The WALK Mode

• WALK (walkthrough) focused navigation mode
• Under the WALK mode, a viewpoint easily gets
  stuck as the avatar collides with obstacles.
• Low frame rates
• Limited view on screen (no side view)
• Need assistance.

4
Assisting Mechanisms (I) Motion Planning Method
73
Motion Planning Algorithms LiTing, VR2000
5
Assisting Mechanism (II)Virtual Force Field
Method

• Using virtual force field to improve navigation
  control
• However, no single assisting mechanism or set of
  tunable parameters is the best for all users.
  (User discrepancy exists.)

49
Virtual Force Field LiChou, CVGIP2001
6
Proposal for Adaptive Force Field

• Objective Personalized virtual force field
• Optimizing navigation efficiency for a given user
• Adjusting virtual forces according to a users
  reaction
• We propose two methods based on virtual force
  field to achieve user adaptation.
• Simulation experiments (off-line)
• Dynamic adjustment (on-line)

7
How Virtual Force Works?
8
Run-Time Adjustable Virtual Forces

• Virtual forces should be adjusted according to
  run-time user interaction.
• Example 1 sensitivity to speed.
• No force when stop
• Example 2 sensitivity to moving direction
• Repulsive force should be gradually reduced if
  the user is moving toward obstacles intentionally.

9
Parameters for Adjusting Forces
LiChou, CVGIP2001

• Adjusting the default virtual force

• Tunable parameter configuration q (k, m, n, l )

What are the best parameters?
10
Finding Optimal Parameters

• Not feasible to ask the user to repeat the
  experiment.
• too tedious
• getting familiar.
• Use simulation
• sampling user input
• repeating the experiments

11
User Input Sampling
Real User
Inputs
3D Interface
3D Display
12
Simulation Experiments (Off-line)
13
Problem of Simulation

• How to simulate high-level user intention without
  visual feedback

14
System Diagram for Simulation Experiments

• Milestone
• Direction Curve

15
Using Milestones to Adjust Input Commands

• Milestones short-term goals along a path

Landmarks
Input Commands
16
Using Direction Curve to Adjust Orientation
Pn original location Hn look-at location Qn
current location
17
Adjusting Input Commands According to User
Intention
Simulation without considering user intention
Simulation accounting for user intention
18
Dynamic Adjustment (On-line)

• Disadvantage of simulation experiments
• Requiring prior sampling and off-line analysis
• Optimal parameters might be scene dependent.
• Objective bringing the simulation on-line
• Ideas to cut down computation
• Do simulation in the window of N steps
• Only move to the best neighboring configuration
  in one step. (converge gradually)

19
System Diagram for Dynamic Adjustment
20
Incorporating Momentum for Configuration Changes

• Neighboring configurations of (A1, A2, A3, A4)
  (A1-D1, A2, A3, A4), (A1D1, A2, A3, A4), (A1,
  A2-D2, A3, A4), (A1, A2D2, A3, A4),
• Problem sensitive to user input and environment
• Objective to avoid frequent fluctuation on
  parameter configurations.
• Attaching a momentum counter to each direction.
• Change configuration only if enough momentum has
  been accumulated.

21
Experiments with Dynamic Adjustment
22
Evaluation

• Test subjects 10 users with various 3D and
  computer experiences
• Average data

23
User Discrepancy Experiment
User A experienced 3D user User H novice
computer user
24
Conclusions

• Virtual force field is a useful assisting
  mechanism to improve 3D navigation. However,
  parameter personalization is important.
• We proposed simulation experiments and dynamic
  adjustment methods with user intention analysis
  to find good control parameters.
• Off-line simulation experiments can find the
  optimal parameters while the on-line version is
  more practical and flexible.
• Experiments show both methods improve navigation
  performance by customized parameters.

25
Future Work

• Incorporating motion planning techniques to
  resolve encountered collisions while considering
  user discrepancy.
• Extensions to consider more environmental factors
  and other human factors.

26
Questions