Parameter estimation for nonlinear systems: adaptive innovations model filters vs. adaptive extended Kalman filters

The problem of recursively estimating the states and parameters of a nonlinear continuous-time system with discrete measurements is investigated. As a new method, an adaptive extended Kalman filter is proposed and compared to an existing approach, an innovations model filter. By means of a simulation example, it is illustrated that both methods are capable of estimating the parameters of a nonlinear system, but that due to the time-varying filter gain in the new method, better state estimates are obtained. The new method is therefore considered a valuable alternative to existing methods.