Application of the Sign-Constrained Robust Least-Squares Method to Surveying Networks

Abstract

The least-squares (LS) method is highly susceptible to outlying observations. For this reason, various types of robust estimators have been developed; for example, M estimators. In this paper, it is proposed to use the sign-constrained robust LS (SRLS) method in surveying networks utilizing the shuffled frog-leaping algorithm (SFLA). The robustness of SRLS is directly implemented as constraints. Therefore, a penalty function approach is used to deal with the constraints. In addition, the performance of any stochastic optimization approach such as SFLA strongly depends on the search domain. Hence, a strategy to define the boundaries of the search domain has been developed for use in surveying networks. The results indicate that SRLS yields better results than the LS method even if there are more outliers among the observations.