【作者】 余满珍； 陈启军；

【Author】 Yu Manzhen Chen Qijun(Department of Information and Control Engineering, Tongji University Shanghai 200092)

【机构】 同济大学控制科学与工程系；

【摘要】 本文介绍了一种基于启发式群聚算法的机器人全局任务调度策略,在任务划分阶段采用一种启发式群聚算法对随机划分的各子任务按能否使总的运行时间缩短进行各子任务间的聚合,以使各子任务粒度和相互之间的通讯量达到一种优化状态,并在此基础上对群聚之后的各子群任务采用集中式动态调度策略,在程序运行过程中实现各处理器的负载平衡,提高各处理器的利用率,缩短总的运行时间。在一个由5个DSP处理器组成的同构型松耦合MIMD并行处理平台上,对平面四自由度连杆机器人在关节锁定下的运动控制任务采用上述先随机划分、再聚合、最后集中式调度的方法进行了并行实时仿真实验,取得了满意的并行性能指标。更多还原

【Abstract】 This paper presents a heuristic clustering algorithm based scheduling strategy for the global task control of robot. Subtasks partitioned occasionally are clustered depending on whether or not the total running time can be decreased, so the subtask granularity and communication consumption between subtasks can be optimized. Based on this algorithm, a centralized dynamical scheduling strategy is used to schedule subtasks clustered to processors, which realizes the load balance of processors, improves the utilization of processors and decreases the total running time. This method is applied to the control of a 4 DOF robot in the presence of joint failure, firstly, the whole control task is partitioned occasionally into subtasks, then these subtasks are clustered according to the clustering algorithm, after that, all the subtasks are assigned to a simple homogeneous loosely coupled MIMD parallel process system with five DSPs using the centralized dynamical scheduling strategy. The real-time simulation results show the effectiveness of the algorithm.更多还原