Adaptive filters are local estimators from control theory that can track time-varying parameters in a compute-efficient fashion. Robust implementations are not frequent, but are necessary for local estimation. We motivate our claim based on a challenging example of sensor data stream, publicly available.