Abstract: While designing multi-task real-time control systems，thread pool technology is an effective solution，and overtime is a chief avoidance of real-time control systems.In order to reduce the overtime rate of thread completeness，Half-Sync/Half-Async thread pool framework is applied to establish the thread pool of real-time control system，and Least Square Support Vector Regression（LSSVR） is used to estimate the executing times of threads.Also，a scheduling priority algorithm of thread pool is designed based on the estimation results.In performance test，wireless image sensor network is implemented to test the overtime ratios of LSSVR and other thread pools in different states，and test result indicates that LSSVR thread pool has visible superiority of overtime avoidance in most applications.

Key words: Least Square Support Vector Regression（LSSVR）, real-time control, thread pool

摘要： 在进行多任务实时控制系统设计时，采用线程池技术是一种有效的解决方法，但必须首先避免超时的发生。为了降低线程完成的超时发生率，采用Half-Sync/Half-Async线程池架构建立实时控制系统的线程池，利用最小二乘支持向量回归机（LSSVR）对线程执行时间进行预测估计，再基于估计结果对线程池线程的分配调度优先级别算法进行设计。性能测试以无线图像传感器网络节点为对象对所设计的LSSVR线程池及其他线程池在不同状态下的超时发生率做了比较，结果表明在大多数应用情况下LSSVR线程池在抑制超时方面具有明显的优越性。

关键词: 最小二乘支持向量回归机, 实时控制, 线程池