This paper presents a passive monitoring mechanism, loss）, nodes inference （LoNI）, to identify loss）, nodes in wireless sensor network using end-to-end application traffic. Given topology dynamics and bandwidth constraints, a space-efficient packet marking scheme is first introduced. The scheme uses a Bloom filter as a compression tool so that path information can bc piggybacked by data packets. Based on the path information, LoNI then adopts a fast algorithm to detect lossy nodes. The algorithm formulates the inference problem as a weighted set-cover problem and solves it using a greedy approach with low complexity. Simulations show that LoNI can locate about 80% of lossy nodes when lossy nodes are rare in the network. Furthermore, LoNI performs better for the lossy nodes near the sink or with higher loss rates.
针对无线传感器网络的最小二乘定位算法抗差性的不足,提出了一种基于时空滤波(STF)的抗差性加权最小二乘(WLS)节点定位算法——STLS。该算法基于空间域滤波的数据一致性检测算法利用相邻节点间必须满足的几何约束关系,采用优化矩阵操作,剔除粗差邻居节点,其计算复杂度为多项式的平方。通过使用具有2步收敛特性的 DFP算法,最小化目标代价函数,实现节点的快速定位。实验结果表明,在均匀网格拓扑或各向异性 C 型网格拓扑下,该算法均可有效识别和剔除测距低估粗差点,其定位精度明显优于未进行空间一致性检测的加权最小二乘定位算法,当网络平均连通度较低时,该优势表现得尤为明显。
