Estimating sudden changes of biases in linear dynamic systems

A procedure is presented for estimating sudden changes in the bias vector in a linear dynamic system. This procedure is the result of combining the recently-developed method [1], [2] of tracking a piecewise-constant vector observed directly through additive noise with the method [3] for separating the estimation of the bias in a linear system from the estimation of the dynamic state variables. The application of the procedure is illustrated by a second-order dynamic system with two biases subject to transition. Its performance is demonstrated by a simulation.