机载网络时间触发通信调度设计优化与评价方法

The airborne network is an important component in the avionics system to realize information transmission and function integration. The introduction of the Time-Triggered (TT) mechanism and the arrangement of reasonable communication scheduling can effectively ensure the real-time and reliable communication of airborne networks; however, the problem of selection from multiple scheduling design methods needs to be solved first. Based on a summary of time-triggered communication processes and characteristics, taking the Time-Triggered Ethernet as the analysis object, we present scheduling models guided by different constraints on the basis of the Satisfiability Modulo Theories (SMT), Mixed Integer Programming (MIP) and heuristic algorithm, and further form a scheduling method based on Reinforcement Learning (RL). From the perspective of solving ability and performance guarantee of the scheduling design, evaluation indexes including the solving time, solvable scale, transmission delay and scheduling modes are established. With an industrial scale network as a study case and the evaluation indexes as an analysis instructor, different scheduling methods are verified and compared to search for appropriate application scenarios. Results show that the SMT method can support small-scale cases, while the MIP method has obvious advantages in TT traffic transmission and schedule solving capability. The heuristic method typical of the genetic algorithm and the RL method can support large-scale cases. Despite their slightly higher TT delay than that of the MIP method, their rate-constrained delay is significantly optimized by 7.97% and 12.35%, respectively. The future development trend of the time-triggered communication scheduling design optimization technology in avionics system airborne networks is also discussed.