Unknown Input Observer-based Finite-time Frequency Estimator for a Biased Multi-sinusoidal Signal

This paper proposes a novel finite-time parameter estimator for a multi-sinusoidal signal with a constant bias. Via the architecture of unknown input observer (UIO), a parameter model consisting of finite-time estimate of the sinusoid and its derivatives is rendered which can be applied with any existing adaptive identifier. Next, motivated by the recently developed Dynamic Regressor Extension and Mixing (DREM) method, we propose a novel parameter estimator guaranteeing the finite-time convergent property. Besides the finite-time property, the advantages such as modularity with systematic procedures, and fine-tunning gains implicitly manifest the effectiveness of our algorithm, which is verified by simulation.