面向分布实时嵌入式系统的任务优化调度算法

文章针对负载难以准确预知且动态变化的复杂分布实时嵌入式系统,提出了一种基于快速模型预测控制的任务调度算法,通过动态调整分配在各个处理器上的任务QoS,使得每个处理器的CPU利用率保持在参考值附近,保证了任务的实时性;利用多参数二次规划的方法求解模型预测控制中的约束优化问题,可有效减小算法的执行时间.实验结果表明算法有效且负载极小.

An Effective Scheduling Algorithm for Distributed Real-time Embedded (DRE) System

Aim.The introduction of the full paper believes that the existing scheduling algorithms axe ineffective for a DRE system.Subsection 3.1 establishes the model for the DRE system.Subsection 3.2 does three things: ( 1 ) it discusses the design of the model prediction controller (MPC) ; (2) we transform the nonlinear constrained optimization problem for the MPC into a multi-parameter quadratic programming problem; (3) the scheduling algorithm proposed by us maintains the CPU utilization of each processor around reference values,satisfies the end-to-end deadline of a task and thus greatly reduces its execution time.Section 4 did two experiments to verify our scheduling algorithm.The experimental results,given respectively in Figs.2 and 3 for Experiment I and Table 2 for Experiment Ⅱ,and their comparison show preliminarily that the workload and computation time of our scheduling algorithm are much less than those of the existing scheduling algorithms,thus being effective for resource-limited DRE systems.