This paper presents a new algorithm for unknown input estimation which is able to correct input estimation using new observations. This algorithm is based on extraction of more than one input from one observation in each sample time. These input provide a vector then these inputs are analysed and the final value of unknown input would be determined. The innovation of this algorithm is how to extract and augment inputs. For this scope, a new model named “two samples backward model” is used instead of motion and observation models. Although this algorithm is a novel estimation, conventional algorithms are used in its structure. This algorithm includes one input estimator, two data fusion and two state estimators which are would work to each other consecutively. For input and state estimation, MIE algorithm and standard Kalman Filter with known input are used respectively. Results of simulations depict improvement and softness in estimates using this technique.

Keywords