Adaptive Process and Measurement Noise Covariances for Kalman Filter-Based Speed Estimation with Incremental Encoders: A Preliminary Study

High-accuracy angular speed measurement using incremental rotary encoders is essential for the optimized operation of electric drives and the improvement of energy efficiency in industrial applications. Conventional encoder-based speed estimation relies on frequency- or period-based methods, whose accuracy is limited to complementary operating regions. Hybrid schemes based on stationary Kalman filters have been proposed to fuse both methods; however, since their noise covariances remain fixed, they achieve near-optimal performance only around a single rotational speed, and their accuracy deteriorates over the rest of the speed range. To overcome this limitation, this work proposes two adaptive Kalman filter variants that fuse frequency- and period-based measurements with online covariance adaptation: the measurement-adaptive Kalman filter (MA-KF) and the dual-adaptive Kalman filter (DA-KF). Both are evaluated in simulation and benchmarked against a stationary Kalman filter (S-KF) and the conventional frequency- and period-based estimators. The results demonstrate that Kalman-based estimators achieve significantly lower relative errors than classical methods across the entire speed range, with the maximum relative error of the DA-KF variant remaining below 0.27% while providing a dynamic response comparable to or faster than period-based techniques. Since the analysis is based on an ideal encoder model, these figures should be interpreted as an upper bound on the achievable performance, which experimental validation will further refine.