Sliding-window recursive least squares updates battery circuit parameters from voltage and current data to improve SOC and voltage prediction.
Leaky bitcells with capacitive discharge let a bloom filter forget stored data over time, avoiding rebuilds and counter overhead.
Movable taps and overlapped coefficients cut ISI and noise amplification while reducing equalizer hardware and power use.
Adaptive LMS background calibration corrects pipelined ADC gain and memory errors while pausing training signals to avoid clipping and data loss.
Dynamic sharing of computational blocks cuts adaptive filter power and cost, while offset injection helps preserve coefficient convergence speed.
Adjusting transition and observation matrices preserves non-observable error axes, improving EKF navigation coherence and precision.