电光与控制
電光與控製
전광여공제
ELECTRONICS OPTICS & CONTROL
2011年
2期
39-44,59
,共7页
无线网络%实时系统%调度%可靠性%马尔可夫过程
無線網絡%實時繫統%調度%可靠性%馬爾可伕過程
무선망락%실시계통%조도%가고성%마이가부과정
wireless networks%real-time system%scheduling%reliability%Markov processes
像机载应用这样的安全关键(safety-critical)系统通常具有非常高的可靠性和实时性需求,而无线连接通常被认为是易错的.因此,在考虑设计或者分析一个安全关键的无线网络时,需要新的QoS参数来衡量系统的实时性.一个合适的参数是在给定信道条件下消息的丢包率,即消息不能在时限之前可靠传输的概率.一种应用于实时系统的概率实时可调度的分析方法被引入来分析无线网络中的消息丢包率.使用错误重传的消息被建模为具有可变长度执行时间的任务,消息的调度过程被建模为一个状态转移的马尔可夫过程.通过这一方法,不仅可以得到关于网络中所有消息的平均丢包率,而且可以得到单个消息的丢包率.实验结果表明:1)通过将该方法的分析结果与仿真结果比较,验证了该方法的有效性;2)通过该方法与假设每个消息都将至少传输某一个固定次数的分析方法的比较,说明了后者的分析结果是过于保守的.
像機載應用這樣的安全關鍵(safety-critical)繫統通常具有非常高的可靠性和實時性需求,而無線連接通常被認為是易錯的.因此,在攷慮設計或者分析一箇安全關鍵的無線網絡時,需要新的QoS參數來衡量繫統的實時性.一箇閤適的參數是在給定信道條件下消息的丟包率,即消息不能在時限之前可靠傳輸的概率.一種應用于實時繫統的概率實時可調度的分析方法被引入來分析無線網絡中的消息丟包率.使用錯誤重傳的消息被建模為具有可變長度執行時間的任務,消息的調度過程被建模為一箇狀態轉移的馬爾可伕過程.通過這一方法,不僅可以得到關于網絡中所有消息的平均丟包率,而且可以得到單箇消息的丟包率.實驗結果錶明:1)通過將該方法的分析結果與倣真結果比較,驗證瞭該方法的有效性;2)通過該方法與假設每箇消息都將至少傳輸某一箇固定次數的分析方法的比較,說明瞭後者的分析結果是過于保守的.
상궤재응용저양적안전관건(safety-critical)계통통상구유비상고적가고성화실시성수구,이무선련접통상피인위시역착적.인차,재고필설계혹자분석일개안전관건적무선망락시,수요신적QoS삼수래형량계통적실시성.일개합괄적삼수시재급정신도조건하소식적주포솔,즉소식불능재시한지전가고전수적개솔.일충응용우실시계통적개솔실시가조도적분석방법피인입래분석무선망락중적소식주포솔.사용착오중전적소식피건모위구유가변장도집행시간적임무,소식적조도과정피건모위일개상태전이적마이가부과정.통과저일방법,불부가이득도관우망락중소유소식적평균주포솔,이차가이득도단개소식적주포솔.실험결과표명:1)통과장해방법적분석결과여방진결과비교,험증료해방법적유효성;2)통과해방법여가설매개소식도장지소전수모일개고정차수적분석방법적비교,설명료후자적분석결과시과우보수적.
The safety-critical systems like avionics generally have both very high reliability and real-time requirements,while wireless connection is known to be error prone.To design or analyze a wireless safety-critical network,new real-time Quality of Service(QoS)metrics ale needed.A proper metric is the loss probability that messages cannot be reliably transmitted before deadlines in a given channel condition.The probabilistic schedulability analysis method was introduced to calculate the message loss probability in wireless network.A message that would be retransmitted until it had been successfully received before deadline was considered as a task with stochastic execution time and the message scheduling process was modeled as a Markov process with state transition.Through this method,both the average loss probability of all messages and the loss probability of every message in the network Can be acquired.The first experiment validates the effectiveness of the method by comparing its analysis results with its simulation results.The second experimental result shows that the usual method,which uses the traditional schedulability analysis method to analyze whether the system design requirements ar'e satisfied by assuming every message is transmitted for a fixed number,is too conservative compared with the proposed method.