Structure exploiting derivative computation for moving horizon estimation

This paper deals with the solution of the optimization problem resulting from the moving horizon estimation technique. A method is presented for the most crucial and computational costly part, namely the first and second order derivative calculation. It fully utilizes the properties of the sequential single shooting based problem formulation by exploiting the structure in the derivatives as well as in the associated first-order sensitivities. This results in a highly efficient method suitable for parallelization. Furthermore, the generalized problem formulation offers the possibility to con sider computational delays and allows the application of the method to both continuous-time and discrete-time systems. Simulation results confirm the effectiveness of the proposed method.