不等式约束最小二乘问题解及统计性质.pdfVIP

36 1 Vol. 36, N o. 1 2007 2 A CT A G EODA ET ICA et CART OG RA PHICA S IN ICA Feb. , 2007 : 2007) : P207 : A 1, 2 1 2, 3 1 彭军还 , 张亚利 , 章红平 , 刘 星 1. , 400045; 2. , 100000; 3. , 200030) The Solution of Inequalityconstrained Least Squares Problem and Its Statistical Properties 1 , 2 1 2, 3 1 PENG Junhuan , ZHAN G Yali , ZHAN G Hongping , LI U X ing ( 1. School of Civil Engineer ing, Chongqing University, Chongqing 400044, China; 2. Department of Land Science and Tech nology , China Geology University , eij ing 100000, China; 3. Shanghai Astronomical Observatory , Shanghai 200030, China) Abstract: T o the inequality constrained least squares adju stment problem , this paper conv erts many inequality con str aints into one equality constr aint by using ag gregate function of nonlinear pro gramming ; a basic augmented La gr angean algor ithm can obtain the solutions for equality constrained nonlinear programming problem and the solu tions are identical to those obtained by the Bayesian method and/ or simplex algor ithm. T he new appro ach f eatures t he advantag e that the solutions are the ex plicit ex pression of observations, and thu s the statist ical pr operties can be easily determined and the general conclu sions about the super iority of the est imator can be made. A numer ical exam ple on GPS po int positioning is g iven to show the inter est ing method. Key words: Inequality Constrained L east Squares ICLS) ; Bayesian estimation ; nonlinear pro gramming ; ag gregate constraint method; GPS : 对不等式约束最小