讨论了 GSS 算法的最优性，通过理论分析和反例证明了该算法不能保证解的最优性，并在系统状态维度大于观测向量维度的情况下，针对 GSS 算法起点随意性导致解变差的问题，提出了一种结合穷举搜索的改进方法．实验结果表明，GSS 改进算法在实际中具有较好的效果，算法的最优率在 90%以上，在未获得最优解的情况下，其所获解与最优解之间的误差在 0．35%以下．文中最后以无线传感器网络中的目标跟踪问题为实例进行了仿真实验，结果表明 GSS 改进算法比原算法更优．

In this paper,first,the optimality of GSS algorithm is discussed,and a theoretical analysis as well asseveral counter- examples is presented to prove that the traditional GSS algorithm does not guarantee the obtaining ofoptimal solution.Then,in the condition that the state vector dimension is larger than the observation vector dimen-sion,an improved GSS algorithm combining the exhaustive search is proposed to avoid the selection result degrada-tion due to the randomness of initial solution selection.Experimental results indicate that the improved algorithm iseffective in practice because it is of a high optimal selection rate of more than 90% and a low solution error of lessthan 0.35% even when the optimal selections are suboptimum.Moreover,the simulated results of the target track-ing in wireless sensor networks demonstrate that the improved GSS algorithm is superior to the traditional one.