Multi-cluster Time Triggers Real-time Ethernet Adaptive Scheduling For Train Control

Switched industrial Ethernet technology has become the development direction of intelligent high-speed train communication network due to its large bandwidth,high speed and good compatibility.The train control system requires train communication Ethernet to provide highly reliable and strong real-time deterministic data transmission services.However,the train real-time data protocol(TRDP)in IEC614753-2-3 does not have a clear data scheduling mechanism and can not meet application needs.In view of the current problem of low real-time and determinacy of TRDP,starting from the actual application scenario of the train,the train communication Ethernet is abstracted into a cluster network structure in groups.A multi-cluster train real-time adaptive schedule model based on time trigger mechanism is proposed,to improve the real-time and certainty of data transmission.It is verified through data set calculation,NS-3 platform simulation and SOC communication platform verification.The main research work of this dissertation is as follows:First of all,the characteristics of train communication Ethernet data transmission and networking topology are analyzed in detail.Combined with its semi-closed dynamic coupling networking feature,the MC-TREAS model based on the time trigger mechanism is proposed.By inserting the cluster porosity,the real-time periodic data is scheduled online by non-strict periodic scheduling in units of clusters.Second,in view of the characteristics of train communication Ethernet dynamic repairability and complex topology,a dynamic fault tree analysis method based on cluster network K-terminal connectivity detection is proposed to evaluate the system data transmission reliability.The impact of data cycle and sorting method on network schedulability is analyzed,harmonic period and flow strict periodic ascending sorting is used to improve the schedulability of the network.The quantum particle swarm optimization algorithm is used to solve the optimization problem of intra-cluster traffic time slice allocation constraints in the MC-TREAS model.The maximum theoretical value of traffic jitter and delay is at the sub-millisecond level,and adaptive adjustment parameters Genetic algorithm and fuzzy control strategy are optimized to improve premature convergence.For the redistribution of inter-cluster traffic caused by train reorganization,an adaptive scheduling algorithm is proposed to match the data and pores online,and the calculation time is within 3s.In order to achieve the minimum delay transmission of event triggered data,the Physarum polycephalum foraging algorithm is used to plan the shortest delay path for the event-triggered data online,and the calculation time was within 800 us.Fourth,the NS-3 simulator was used to simulate the real-time periodic data transmission performance in the MC-TREAS model.Based on zynq-7000 series chips,schedulable real-time Ethernet terminals and communication nodes for train communication network are designed,and a communication platform is built.The platform can implement the TRDP full-scenario communication mode,and the maximum jitter and delay of periodic data are within 1ms.Finally,the completed work is summeried,and the next research and improvement strategies for the existing work defects are put forward.The models and algorithms proposed in this dissertation have been verified by theory or experiment and have engineering practical value,which can bring certain reference value to the real-time and deterministic improvement of train communication Ethernet.