vis2006present v4

A Trajectory-Preserving Synchronization Method for Collaborative Visualization: 

A Trajectory-Preserving Synchronization Method for Collaborative Visualization Lewis W.F. Li* Frederick W.B. Li** Rynson W.H. Lau**

Overview: 

Overview Introduction Related Work Methodology Experiment Results Conclusion

Part I: 

Part I Introduction

Introduction (1/2): 

Introduction (1/2) Collaborative visualization Geographically separated users to be connected over the network to visualize and manipulate dataset for problem solving Examples Fluid dynamics visualization Volume visualization Medical data visualization

Introduction (2/2): 

Introduction (2/2) Characteristics of collaborative visualization User is allowed to interact with the visualization dataset continuously over time Dataset updates should subsequently be distributed to remote users over the network Problems Due to network latency, each remote user may receive updates with a different amount of delay User’s ability in performing desirable collaborative tasks will be affected, due to the induced view discrepancy among remote users

Objectives of This Work: 

Objectives of This Work Provide a more synchronized view of visualization changes to collaborating users Develop procedures to correct motion trajectories of dynamic objects Prevent discontinuous motion Address false positive and false negative collision detection problems

Part II: 

Part II Related Work

Related Work: 

Related Work Traditional Applications Easy to work well provided that state updates are received by remote sites in a correct order Time gap between two consecutive updates is typically large as compared to network latency Collaborative Applications State updates occurs continuously Unfortunately, updates need to present to remote users timely or at least within a very short time Existing solutions: - User or system side adaptation - Local Lag mechanism

Part III: 

Part III Methodology

Methodology: 

Methodology Relaxed Consistency Control Model Gradual Synchronization Trajectory-Preserving Synchronization

Relaxed Consistency Control Model: 

Relaxed Consistency Control Model Observation: Users generally pay more attention on the trajectory of dynamic objects rather than their individual states Given that the states of a replicated object at two remote sites at time t are si(t) and sj(t), the state discrepancy D of the object between the two sites during any time period Ta and Tb should be smaller than an application specific tolerance, ξ. Hence,

Gradual Synchronization (1/2)ACM Multimedia 2004: 

Gradual Synchronization (1/2) ACM Multimedia 2004 Trade accuracy of individual state of a dynamic object for preserving their state trajectory Run a reference simulator on the server for each object in a client-server environment Note: 1st order simulator: 2nd order simulator: When a client receive or initiate a new motion update of an object, the client will align the motion of the local object against its reference simulator

Gradual Synchronization (2/2) ACM Multimedia 2004: 

Gradual Synchronization (2/2) ACM Multimedia 2004 Contribution: This method effectively reduces the latency of a client to obtain a state update from a double round-trip time delay to a single one Limitation: High discrepancy occurs between the period when an interaction has just occurred and before the update message reaches a remote client Apparently, such discrepancy appears shortly for each time, but would become serious if interactions occur frequently

Trajectory-Preserving Synchronization: 

Trajectory-Preserving Synchronization Extends from our gradual synchronization method Consider the characteristics of spatial changes and interactions of dynamic objects are affected by network latency A set of procedures are developed to correct motion trajectory of dynamic objects Handle false positive and false negative collision detection problem

Client-Server Trajectory-Preserving Synchronization: 

Client-Server Trajectory-Preserving Synchronization Client A (avatar) and the server

Client-Client Trajectory-Preserving Synchronization: 

Client-Client Trajectory-Preserving Synchronization Server and client B (observer)

Arbitrary Moment Trajectory-Preserving Synchronization: 

Arbitrary Moment Trajectory-Preserving Synchronization Client A (avatar) and the server

Handling Object Collisions Trajectory-Preserving Synchronization: 

Handling Object Collisions Trajectory-Preserving Synchronization Interpret the collision response as motion commands Resolve inconsistent collision problem into two sets of simpler problems

Handling Object Collision Trajectory-Preserving Synchronization: 

Handling Object Collision Trajectory-Preserving Synchronization False negative collisions Collisions detected in the avatar but not in the server (case (b)) Inhabit the avatar to perform collision detection until motion remediation process has finished False positive collisions Collisions detected in the server but not in the observer (case (f)) Inhabit the observer to perform collision detection until motion remediation process has finished

Part IV: 

Part IV Experiment Results

Experiment I (1/4): 

Experiment I (1/4) Demonstrate user’s navigation at an avatar, the server and an observer Compare the performance of different methods Dead Reckoning Original method New Method Here, focus on comparing dead reckoning and the new method only Full and other Demos http://www.cs.cityu.edu.hk/~kwfli/vis2006/vis.html

Experiment I (2/4): 

Experiment I (2/4) Dead Reckoning

Experiment I (3/4): 

Experiment I (3/4) New Method

Experiment I (4/4): 

Experiment I (4/4) Focus on comparing several motion changes Dead Reckoning New Method

Experiment II (1/5): 

Experiment II (1/5) Focus on the motion of selected object (the green ball) in the virtual environment Compare the position discrepancy in between Client A and the server The server and client B Client A and client B

Experiment II (2/5): 

Experiment II (2/5) Screen shots of our prototype for collaborative visualization

Experiment II (3/5): 

Experiment II (3/5) Client A and the server

Experiment II (4/5): 

Experiment II (4/5) The server and client B

Experiment II (5/5): 

Experiment II (5/5) Client A and client B

Experiment III (1/3): 

Experiment III (1/3) Focus on the accuracy of the new method in handling object collisions Compare the position discrepancy between server and four users with different network latencies

Experiment III (2/3): 

Experiment III (2/3)

Experiment III (3/3): 

Experiment III (3/3)

Part V: 

Part V Conclusion

Conclusion (1/2): 

Conclusion (1/2) Propose a trajectory-preserving synchronization method to support collaborative visualization Handle unpredictable user changes Handle collision detection problem

Conclusion (2/2): 

Conclusion (2/2) Limitations Assume using connection-oriented network Message loss is not considered Future Works Consider difference types of network Support haptic interface and rendering

Thank you!: 

Lewis Li: kwfli@cs.cityu.edu.hk Frederick Li: Frederick.Li@durham.ac.uk Rynson Lau: Rynson.Lau@durham.ac.uk Thank you! Questions and Answers Contacts http://www.cs.cityu.edu.hk/~kwfli/vis2006/

Appendix Clock Synchronization: 

Appendix Clock Synchronization Two common approaches Backward correction Forward correction

Appendix Dead Reckoning: 

Appendix Dead Reckoning Client A and client B

AppendixGradual Synchronization: 

Appendix Gradual Synchronization For each motion Motion timers Ts and Tc are maintained at the server and client simulator, respectively Assume position updates in every Δt Estimate the round-trip time, Test Adjust every Δt in client for Tc based on Test Synchronized when Tc is the same as Ts

AppendixGradual Synchronization: 

Appendix Gradual Synchronization Client A and the server

AppendixGradual Synchronization: 

Appendix Gradual Synchronization Server and client B