Research And Implementation Of Virtual Emergency Drill System Server-Side Key Techniques

Virtual emergency drills based on virtual reality are widely used in chemical, file safety and other industry. Virtual emergency drill system provides identical virtual environment, simulation of the disaster scene and supports users to complete drill by cooperation. However, lack of support of large scale users, lack of assurance of system availability and lack of efficient environmental simulation are three major problems affecting the training effect. Then, three key technologies were studied to solve the problem, including high performance network framework, server-side service infrastructure and virtual drill simulation system. The main works of this paper are:1. Researching and implementing a high performance network communications framework for virtual emergency drill. First, framework used Protobuf to implement an efficient data transfer protocol. Then, by using IO multiplexing, one-loop-per-thread and thread pool pattern, the core function of the network framework for high concurrency, compute-intensive and low-latency required system was implemented. Furthermore, framework provided a buffer implementation for asynchronous IO and other network framework auxiliary function modules like timer, log system and file transfer system.2. Focusing on improving availability, scalability and fault tolerance of the system, a RPC (Remote Procedure Call, RPC) framework was designed for drill services communication and a distributed message queue system was designed for frequent message exchanging. Furthermore, a system hierarchical architecture was given in this paper. Also a position synchronization algorithm and a system clock synchronization algorithm were designed to smooth synchronization effects and improve system data consistency.3. Researching the communicating, user migrating and load balancing problems of multiple scene-server architecture. A simulation system for virtual emergency drill was designed to solve the efficient message passing problem and large scale simulation model computing problem. Furthermore, a dynamic self-adjusted task scheduling algorithm LET was designed for heterogeneous platform. Experimental results show that LET algorithm improved the performance in comparison with WRR and also improved the computing power of the system.At last, the study result was improved to be effective, reliable and scalable with the application in real company fire drill training.