A Reweighted Filtering Algorithm and Its Application to Open Pit Deformation Monitoring

The Global Positioning System (GPS) has been proved to be a capable surveying tool for monitoring the stability of steep walls in open pit cut mines. However, due to restrictions on satellite geometry and severe environment-dependent errors, as well as the uncertainty of the a priori knowledge of system noise variance, sub-centimetric accuracies using existing algorithms cannot always be achieved. This paper proposes a reweighted filtering algorithm which treats GPS deformation monitoring in an open pit environment as a kinematic surveying system. Estimating techniques of a posteriori weights of predicted states and observations are developed to match the weights appropriately and obtain reweighted filter solutions. Two different weight functions to reweight the predicted states and observations are established based on the state residuals and observation residuals. The proposed algorithm can automatically assign the larger weights to the more accurate predicted states and observations and the smaller weights to the less accurate predicted states and observations. Results from simulated GPS deformation monitoring data are shown using the proposed algorithm. Comparisons with the traditional Kaiman filtering method indicate high levels of filter stability and accuracy for the proposed algorithm.