Using State Estimating to Enhance Real Time and Fault Tolerance inDistributed Simulation

摘要

In simulation field, researchers pay morernattention about characteristic of real time and faultrntolerance. For WAN and distributed environment, thernnetwork transport delay and network partition problemsrnare difficult to be handled, more over HLA, as a currentrnpopular framework for distributed simulation does notrnsupport real time and fault tolerance. According tornanalysis of simulation models, we divide system modelsrninto continuous state model and discrete state model tworntypes. Hermite interpolating and Markov process theoryrnare proposed to handle network delay and partitionrnproblems by substituting real computing data withrnestimated data, so simulation system has the ability ofrnpassing through the transient failure or data transportrndelay. Consequently real time and fault tolerancerncharacteristics of whole simulation system are improved.rnIn order to guarantee the accuracy of simulation result,rnwe use optimistic recovery to restore system andrnsubstitute estimated data with delayed real compute datarnwhen the data estimating error exceeds the error limit. Atrnlast experimental results show that proposed method isrneffective with condition of continuous system model andrnsmall step (look-head).