三维无线传感器网络定位的可行方向算法

A Feasible Direction Algorithm for Solving 3D Sensor Network Localization

针对内点算法求解半定规划进行三维无线传感器网络定位无法满足实际的需求,利用变量变换将半定规划问题转换为非线性规划问题,提出了解决非线性规划化问题的可行方向算法.在此基础上,取非线性规划问题变量的列数为3,阻止高秩解的产生.为了进一步提高计算效率,采用限制未知节点的度对三维网络图进行稀疏.仿真结果表明,可行方向算法是行之有效的,而且计算速度优于已有的稀疏半定规划内点算法.

By using the change of variables,the semidefinite programming（ SDP） problem for solving the senor network localization（ SNL） in 3D was reformulated to be a nonlinear programming（ NLP） problems. Feasible direction algorithm was proposed to solve the problem. The number of columns of the variables in the NLP is chosen to be equal 3,so as to avoid the higher dimensional solutions. Computational efficiency is improved by exploiting the sparsity of graph,in which the degree of each sensor node are restricted to a small positive integer. Experiments show that the proposed is efficient and robust,and the speed is faster than the existing interior-point algorithms for the SDP.