汽车电子系统中软件组件分配与混合关键性调度研究

The complexity of automotive electronic system is rapidly growing. It has become a typical mixed-criticality real-time embedded system. Its designing and assurance of performance appear to be challenging. The project will research on both allocating software components to ECUs in design phase and mixed-criticality scheduling on multiprocessors in runtime phase, and will have achievements in the following two aspects: ①With the idea of decomposing problem , we provide a multi-step components allocation algorithm based on graph partitioning: to reduce the communication cost, create components partition using graph partitioning theory; according to the outgoing degrees of nodes, the partition applies branch and cut algorithm. The method can achieve higher resource utilization rate and adapt to the increasing in the number of components when we make an optimization matching. ②After the configuration of ECU, a reliable mixed-criticality partitioned scheduling algorithm for ECU with multiprocessor is designed: Computing the task priority value from functionality subsystem level; study the schedulability of mixed-criticality tasks based on EDF policy, generating task partitions, the number is equivalent to the number of processors, iterate the remained tasks, make partition testing, and select a partition to join in to balance the workload of processors; Using the computed priority value, adjust the execution order of some tasks or migrate them to other ECU. The algorithm can improve the scheduling performance and reliability. Finally, we will make the simulation and prototype verification for these two algorithms. The research achievements of the project can provide theoretical support for the development of automotive electronic system, and improve developing efficiency and system performance.

汽车电子系统复杂性急剧增长，成为了典型混合关键性实时嵌入式系统，在开发和性能保证上均面临许多挑战。项目将在开发前期中软件组件到ECU分配、运行期中多处理器混合关键性调度这两个关键问题开展研究，拟取得以下两方面成果：①采用问题分解思想，设计基于图划分的多步分配算法：减少通信为标准，基于图划分理论生成组件划分；根据外连节点出度对划分进行分支切割。实现较高资源利用的优化匹配，更好适应组件数目增长。②完成ECU配置后，对多处理器ECU，设计可靠的混合关键性任务划分调度算法：从功能子系统级计算任务优先性值；研究基于EDF混合关键性任务可调度性，生成处理器数等同的划分，遍历任务并进行划分检验，以均衡处理器负载为标准选择划分；根据子系统级优先值，调整任务执行顺序或进行迁移，提高调度性能和可靠性。最后进行仿真和原型系统验证。通过本项目研究，为汽车电子系统的研发提供一定理论支持，提高开发效率和系统性能。

随着智能化、网络化的发展，汽车电子系统复杂性迅速增长，系统设计和开发面临许多新的问题。课题围绕汽车电子系统设计开发中的软件组件分配、混合关键性任务调度及可靠性保证等几个关键问题开展了研究工作，主要成果如下：1）对于软件组件到车载ECU分配这一NPC问题，以减少在不同ECU间通信开销为出发点，提出了基于图划分的两步软件组件分配算法，首先对组件图进行K路划分，然后对不满足ECU源约束的划分，依据连接度进行切割调整；2）针对汽车电子系统安全关键性需求和ECU嵌入式单元资源受限的特点，提出了满足任务可靠性的、资源开销成本最小化启发式任务分配算法。采用任务分解的思想，将应用的可靠性目标转变到单个任务上，然后设计了启发式任务分配算法，根据任务的可靠性目标为约束，为任务选择资源开销成本最小的处理器。3）对于混合关键性汽车电子系统，从设计者角度出发，为增加任务的可调度数，在高关键任务模型中引入概率参数，结合任务划分方法，提出了基于概率的虚拟截止期EDF混合关键性调度算法。4）为更好的开展理论验证与应用研究，搭建了基本的车载总线网络实验平台，模拟了车载系统的部分功能。项目组发表研究论文6篇，培养硕士生3名，基本完成了项目的预期目标。项目研究成果，在软件到ECU自动分配、车载系统可靠性保证等方面具有一定的理论参考价值。