Time-Sensitive Task Scheduling Research For Unmanned Vehicles

With the change of in-vehicle network communication in recent years,higher data transmission rate and real-time communication have become the urgent needs of in-vehicle systems.For advanced driver assistance systems,the traditional CAN bus is far from satisfying their communication requirements,and the in-vehicle Ethernet-based TSN network is a better solution,which guarantees transmission bandwidth while ensuring data timing,very low latency and reliability.Starting from the scheduling problem in unmanned driving,this paper combines the problems of time synchronization and traffic control in time-sensitive networks,and conducts an in-depth study in the direction of task scheduling for unmanned systems from three aspects:scheduling algorithm theory,network environment simulation,and development board protocol real testing.The main research contents and contributions of this thesis are summarized as following:1)A task dynamic scheduling algorithm based on dual criticality levels is proposed.Based on the traditional priority scheduling strategy and with reference to the automotive safety integrity ASIL level standard,this paper divides the possible tasks in the in-vehicle system into critical levels according to the sensors and electronic control components of the sources,and also gives the instance critical levels by the computation time and deadline of the tasks,and integrates the dual critical levels for dynamic preemptive scheduling.The experimental results show that according to the task values set in this paper for tasks of different critical levels,the dual-critical-level dynamic scheduling strategy improves the completion of the task set by 13 times compared with the traditional static monotonic rate scheduling algorithm(RM)and 23.8%higher compared with the dynamic earliest deadline priority scheduling algorithm(EDF)tasks,which can schedule the actual task set well and meet the needs of unmanned scenarios.2)The priority queues were constructed in the OMnet++ simulation platform,and the algorithm was simulated and tested in the simulated environment of a time-sensitive network.Considering the limited transmission load and poor real-time performance of traditional in-vehicle networks,time-sensitive networks TSNs are increasingly used in in-vehicle systems.In this paper,the priority-based task sequence in the theoretical algorithm is constructed in the OMnet++environment by combining the time-accurate protocol and the traffic control protocol in TSN,and it is tested that it can achieve reasonable scheduling for the task sequence based on the time synchronization and discard the time-out tasks,and the improvement in the total task value is about 30.3% compared with the traditional single-queue EDF and 14.1 times compared with the single-queue RM improves by 14.1 times.3)A platform supporting hardware timestamps is constructed and tested for time-sensitive network communication and data transmission.In this paper,based on theoretical research and simulation experiments for the embedded system development board also carried out practical tests,to build a system supporting hardware timestamp on the LS1028 A development board,activate the time-sensitive network capability of the development board,and conduct tests on both time synchronization and traffic control.To avoid the heterogeneity problem of the actual in-vehicle system,data distribution service was deployed on the development board to assist data transmission and also to enhance the migratability of the system.A comprehensive test on task scheduling was conducted based on the time-sensitive network and data distribution data,and finally a 31.9% total task value improvement was obtained compared to the traditional EDF method and 16.2 times higher compared to the RM method.