| With the development of smart cars,the problem of vehicle network bandwidth is becoming increasingly apparent.In order to cope with the tight bandwidth of vehicle network,automotive Ethernet has become one of the important trends of vehicle network.Automotive Ethernet adds a lot of protocols to the existing vehicle network,which makes the design and analysis of vehicle network architecture face new problems.In order to improve the accuracy of the analysis of automotive Ethernet data transmission delay,this paper takes Ethernet AVB as the research object,and based on the mathematical model of the AVB protocol,utilized Q reinforcement learning to fit the data forwarding rules at the switch and established a combination Mathematical model of automotive Ethernet data transmission delay characteristics.The main research work of this paper is as follows:Firstly,studied the related protocols(such as TCP / IP,AVB,TSN).Understood the composition structure of data packets and the data transmission rules of automotive Ethernet,and summarized the differences and connections between AVB and "traditional Ethernet" and TSN.Secondly,established a mathematical model for automotive Ethernet data transmission delay.According to the FQTSS queue forwarding rules of the AVB protocol,analyzed the delay composition and influencing factors,and established the mathematical model of the protocol.Then analyzed the shortcomings of the model,the Q reinforcement learning method was used to fit data forwarding rules,and an AVB data transmission delay agent model based on Q reinforcement learning was established.Meanwhile,designed and developed a software to analyze the delay characteristics of automotive Ethernet data transmission based on the mathematical model.The integration of the three platforms of communication scenarios,parameter settings,and calculation results was formed based on the establishment of the main frame structure.And then achieved a visual and dynamic analysis of the vehicle Ethernet data transmission delay characteristics.Thirdly,the time delay simulation analysis is performed on the surround view system.In MATLAB / SIMULINK,combined 360 surround view system model as application scenario.Meanwhile,the data forwarding delay at the switchboard and the data transmission delay of the entire system were analyzed.The results show that,on the one hand,the average delay and mean square error can be used to evaluate the network performance.At the same time,the protocol guarantees the transmission quality of type A and B data frames.On the other hand,it can reduce the delay of the type B data frames in five aspects: the length and transmission frequency of type B data frame,the length of type A data frame,the load of BE data,and link bandwidth.Finally,tested the software of automotive Ethernet data transmission delay analysis.Established an automotive Ethernet simulation system with BMW G38 surround view system and Intrepidcs RAD_Galaxy as the core,based on this test platform,monitored and sent bus data through Vehicle Spy,and analyzed the data delay by wireshark.Meanwhile,analyzed the main program Test and verified with software functions.In verification of main program,the delay value of each frame in the program is compared with the theoretical delay value.In terms of software function testing,when the reserved bandwidth for data of type A is 25.6Mb / s,the reserved bandwidth for data of type B is 25.6Mb / s,and the load of the data of type BE is 25 Mb / s,the deviation is the largest,and it's value is 4.96%.The research results show that the mathematical model established in this paper can reflect the average performance of the delay of automotive Ethernet,improve the accuracy of delay analysis,and has guiding significance for the design of next-generation vehicle network architecture. |
PDF Full Text Request