Filtering Policies Applied Research In A Collaborative Virtual Environment

Due to limited network bandwidth, state updates filtering became an effective method to improve the scalability and consistency of CVEs. In this paper, filtering policies and their effects on task performance were studied. To quantify the regularity of state updates, the concept of Irregular Degree was introduced and four different filtering policies were brought forward. In order to study the filtering policies in CVEs, we designed a desktop collaborative virtual envoirnment, named JBCVE, which had very short reponse time and can surport highly interactive task. In addition, it can be quickly adjusted according to the requirements of experiment. Therefore, it is a good experiment platform.To study the effects of four filtering policies on task performance, two experiments were designed. In the first experiment, three filtering policies based on Irregular Degree were compared. After the first experiment, the filtering policy that achieved the best task performance was selected to compare with other filtering policies in experiment two. The results of the first experiment indicate that the stability of state updates per second is a more important impact factor than Irregular Degree on task performance. Under the circumstance of stable state updates per second, transmission of the state updates with larger Irregular Degree can slightly improve the task performance. The results of the second experiment indicate that for different filtering policies, the variety of state updates per second will result in different taks performance.According to the influencing research of filtering policies on task performance, an adaptive-filtering policy was brought forward, which could adaptively select the fittest filtering policy corresponding to current motion state of the virtual object in CVEs. An experiment was designed to compare this filtering policy with two other filtering policies on task performance in improved JBCVE. Objective and subjective evaluations were recorded in the experiment. The experiment results show that according to different mean state updates per second, adaptive-filtering policy couldachieve different task performance.