An Information-Theoretic Framework for Consistency Maintenance in Distributed Interactive Applications

Distributed Interactive Applications (DIAs) enable geographically dispersed users to interact with each other in a virtual environment. A key factor to the success of a DIA is the maintenance of a consistent view of the shared virtual world for all the participants. However, maintaining consistent states in DIAs is difficult under real networks. State changes communicated by messages over such networks suffer latency leading to inconsistency across the application. Predictive Contract Mechanisms (PCMs) combat this problem through reducing the number of messages transmitted in return for perceptually tolerable inconsistency. This thesis examines the operation of PCMs using concepts and methods derived from information theory. This information theory perspective results in a novel information model of PCMs that quantifies and analyzes the efficiency of such methods in communicating the reduced state information, and a new adaptive multiple-model-based framework for improving consistency in DIAs. The first part of this thesis introduces information measurements of user behavior in DIAs and formalizes the information model for PCM operation. In presenting the information model, the statistical dependence in the entity state, which makes using extrapolation models to predict future user behavior possible, is evaluated. The efficiency of a PCM to exploit such predictability to reduce the amount of network resources required to maintain consistency is also investigated. It is demonstrated that from the information theory perspective, PCMs can be interpreted as a form of information reduction and compression. v The second part of this thesis proposes an Information-Based Dynamic Extrapolation Model for dynamically selecting between extrapolation algorithms based on information evaluation and inferred network conditions. This model adapts PCM configurations to both user behavior and network conditions, and makes the most information-efficient use of the available network resources. In doing so, it improves PCM performance and consistency in DIAs.