传统最小二乘法对奇异点比较敏感。当样本中存在奇异点时，不能客观反映数据的真实分布情况；而传统最小一乘法计算量太大，不能实时处理数据。针对经典RBS算法在确定节点本地时钟之间的相对时钟漂移率和偏移值时，采用传统最小二乘法会导致时钟同步收敛速度太慢的问题，提出了一种改进型的最小二乘法。该算法能够有效地识别和剔除样本容量中的奇异点。经试验证明，该方法能够改进节点之间的时钟同步效果和收敛速度。%The traditional least square method is comparatively sensitive to singular points, when singular point exists in sample, real actual distribution of the data cannot be reflected;while the traditional least one multiplication method is too computationally intensive, data cannot be processed in real time. Aiming at classical reference-broadcast synchronization ( RBS) algorithm, due todetermine the relative clock drift rate and offset value between local clock of the nodes by adopting traditional least square method may lead to the convergence speed is too slow, thus the improved least square method is proposed. With this method, the singular points in sample size can be effectively recognized and excluded. The tests prove that by using this method, clock synchronous effect and convergence speed between nodes can be greatly improved.
展开▼
