Abstract

Human-to-human interaction across distributed applications requires that sufficient consistency be maintained among participants in the face of network characteristics such as latency and limited bandwidth. The level of inconsistency arising from the network is proportional to the network delay, and thus a function of bandwidth consumption. Distributed simulation has often used a bandwidth reduction technique known as dead reckoning that combines approximation and estimation in the communication of entity movement to reduce network traffic, and thus improve consistency. However, unless carefully tuned to application and network characteristics, such an approach can introduce more inconsistency than it avoids. The key tuning metric is the distance threshold. This paper questions the suitability of the standard distance threshold as a metric for use in the dead reckoning scheme. Using a model relating entity path curvature and inconsistency, a major performance related limitation of the distance threshold technique is highlighted. We then propose an alternative time-space threshold criterion. The time-space threshold is demonstrated, through simulation, to perform better for low curvature movement. However, it too has a limitation. Based on this, we further propose a novel hybrid scheme. Through simulation and live trials, this scheme is shown to perform well across a range of curvature values, and places bounds on both the spatial and absolute inconsistency arising from dead reckoning.

A key factor determining the success of a Distributed Interactive Application (DIA) is the maintenance of a consistent shared virtual world. To help maintain consistency, a number of Information Management techniques have been developed. However, unless carefully tuned to the underlying network, they can negatively impact on consistency. This work presents a novel adaptive algorithm for optimizing consistency by maximizing available bandwidth usage in DIAs. This algorithm operates by estimating bandwidth from trends in network latency, and modifying data transmission rates to match the estimated value. Results presented within demonstrate that this approach can help optimise consistency levels in a DIA.

Scalability is an important issue for Distributed Interactive Application (DIA) designers. In order to achieve this, it is important to minimise the network traffic required to maintain the DIA. A commonly used technique to reduce network traffic is through short-term entity dynamics extrapolation. However, this technique makes no use of a priori information regarding entity dynamics. We have been developing methods to employ this information through a number of techniques, primarily statistical in nature, which have shown great promise in constrained experimental environments. The main tenet of our approach is that user behaviour in real DIAs follows patterns, and through acquisition, analysis and exploitation of these patterns, a reduction in network traffic can be achieved. In this paper, we report on our development of a realistic DIA based on an industry standard SDK in which we have implemented data acquisition routines that allow us to do this. Results are presented for trial runs using the system. These results clearly exhibit patterns of user behaviour consistent with our previous research and suggest that the exploitation of this knowledge can help reduce network traffic.

As Distributed Interactive Applications (DIAs) become increasingly more prominent in the video game industry they must scale to accommodate progressively more users and maintain a globally consistent worldview. However, network constraints, such as bandwidth, limit the amount of communication allowed between users. Several methods of reducing network communication packets, while maintaining consistency, exist. These include dead reckoning and the hybrid strategy-based modelling approach. This latter method combines a short-term model such as dead reckoning with a long-term strategy model of user behaviour. By employing the strategy that most closely represents user behaviour, a reduction in the number of network packets that must be transmitted to maintain consistency has been shown. In this paper a novel method for constructing multiple long-term strategies using dead reckoning and polygons is described. Furthermore the algorithms are implemented in an industry-proven game engine known as Torque. A series of experiments are executed to investigate the effects of varying the spatial density of strategy models on the number of packets that need to be transmitted to maintain the global consistency of the DIA. The results show that increasing the spatial density of strategy models allows a higher consistency to be achieved with fewer packets using the hybrid strategy-based model than with pure dead reckoning. In some cases, the hybrid strategy-based model completely replaces dead reckoning as a means of communicating updates.