Optimization Strategy Of Multi-service Schedule In Ethernet-based Train Communication Networks

With the development of information technology,advanced technologies,such as artificial intelligence,big data,cloud computing and Internet of things,are deeply integrated with rail transit equipments.The high-speed EMU embodies core characteristics of intelligence and information.Besides carrying control data,train communication network(TCN)will generate,integrate,transmit and process larger amount and higher dimension data from multi-source for multi-operation and multi-service.Although the introduction of real-time Ethernet technology has greatly expanded the bandwidth of TCN,currently the real-time guarantee method of communication in TCN is still low traffic shaping and isolation by single-network single-service.TCN based on Ethernet puts forward new requirements for transmission model and scheduling mechanism of multi-service communication to ensure the real-time,security and reliability of multiple data types and improve the utilization of network resources.Transmission models,scheduling algorithms and real-time evaluation methods of three data types,namely periodic real-time data,non-periodic real-time data and streaming media data are proposed in this dissertation.Main research results include the following aspects:Firstly,a real-time periodic data scheduling optimization based on time-triggered mechanism is proposed.A TCN periodic data scheduling model based on time-trigger mechanism and its delay composition of Ethernet TCN is analyzed,as well as a unified temporal constraint description of TCN periodic data.A scheduling optimization model based on data jitter and load balancing is proposed to take control and scheduling performance into account.According to the model,an adaptive fuzzy-control improved quantum particle swarm optimization algorithm and a satisfiability modulo theory based on schedulability sorting are proposed to calculate the time-triggered schedule table of periodic data.Schedule calculation and performance evaluation are carried out based on the random traffic on topology of TCN to prove the effectiveness of the algorithm.Secondly,a real-time apriodic data scheduling optimization of queueing method is poposed.Based on the characteristics of TCN priority,a transmission mechanism of real-time aperiodic data and time-triggered data is established with a new two-level scheduling scheme of DSRR-EDF.The polling weight is comprehensively determined by factors such as queue length,queue priority,error packet loss number and minimum deadline.It can help to ensure the fairness among queues,avoid the port blocking by high priority queue and improve the overall delay performance of network services.Through the EDF scheduling in second level,emergency data can be transmitted first in the same priority queue.Thirdly,theoretical calculation and delay testing method optimization of real-time aperiodic data are proposed.In terms of theoretical delay calculation method,the stochastic network calculus theory is used to calculate delay boundary of different priority queues of real-time aperiodic services by allowing services to exceed the statistical boundary under specified probability under the condition of fully considering influence of time triggered data traffic curve.In the field test method,a prior probability distribution of end-to-end delivery delay rate of real-time data is established.By counting the number of delayed frames in the test sample exceeding deadline in a limited time,a posterior probability model of delay rate based on Bayesian rules is established.The delay test problem is transformed into a statistical confidence problem,which provides a theoretical basis for the selection of testing time and sample number of real-time indicators in industrial field.Fourthly,a bandwidth allocation and scheduling strategy of streaming media data based on quality of experience is proposed.Based on the time-triggered communication model,a TCN based streaming media data transmission model and its real-time performance of stream is analyzed.In view of the limited bandwidth resources of TCN and characteristics of streaming media data,a comprehensive evaluation model of streaming media data network utility based on quality of experience,network transmission performance and cache requirements is proposed.Based on the implementation theory and pricing mechanism of economics,the problem of bandwidth allocation of streaming media data is transformed into the solution of Nash equilibrium of non-cooperative game.For the privacy and decentralization of utility characteristics in streaming media system,a decentralized pricing mechanism is designed to negotiate the bandwidth of multiple links,so as to fully adapt to the TCN structure based on multi-hop switched Ethernet.The effectiveness and usability of the algorithm are proved by experimental simulation.Finally,a TCN multi-service transmission experiment platform based on train real-time data protocol(TRDP)is designed and inplemented with a typical Ethernet TCN two-level structure and network segment configuration as the prototype.The real-time communication equipment on the platform adopts the TRDP network interface cards.The message data source is Ethernet-TCN state perception data based on TRDP-MIB.The streaming media data system adopts the train streaming media broadcast system based on MPEG DASH.Experiments are carried out on the platform to test the delay,jitter and service stability of multi-service communication under different network loads,which verifies the effectiveness of algorithms proposed in this dissertation.Problems for further research are presented on the basis of summary and consideration of the dissertation.