Study And Implementation Of Save-Restore And Log-Replay In HLA/RTI

High Level Architecture ( HLA ) has become IEEE 1516 in September 2000 and the standard for Distributed Simulation in the world. The distributed simulation based on HLA will be the mainstream of future simulation. Run-time Infrastructure (RTI) is the implementation of HLA specification, and also it is the foundation of design and work for a simulation application.The saving and restoring mechanisms of RTI can be used to recover the error in a simulation, it can facilitate analysis of alternatives, and also it can continue an interrupted simulation. So, it is very worthwhile for the realization of saving and restoring. The logging and replaying mechanisms can be used for a simulation for After Action Review (AAR). So users can utilize the simulation well. It is noticeable to develop an HLA federation-independent tool for logging and replaying. But, there is little information what can be gotten in our country, even from abroad.In the thesis, the background and the evolution of HLA and the basic ideas of HLA standard is introduced first. And then the mechanism and design and implementation of saving and restoring in RTI federation management are presented. After that, the thesis analyses the design of logging and replaying, and it presents the implementation of StarLogger, and also the improvement of StarLogger. At last, the thesis presents the test for the function of saving and restoring of StarLink, and also it presents the test of StarLogger.The innovatory points to be revealed in the thesis is follow: first, the mechanisms of saving and restoring in StarLink is studied, and a method which is used not only by saving and restoring, but also by creating to recover the data of RTI is presented. It makes it easy to implement the saving and restoring and also it improves the stability. Second, the key problems of how to save the complex data to the disk and how to recover it from the disk are studied, so some data such as pointer and class object can be saved and read out from disk file. Third, the method that log and replay the one federate is presented, from it a method that extend the MOM to get the stub is studied so as to directly replay the simulation. The simulation created by these methods resembles the real simulation greatly.