Cascaded PI-LMS adaptive prediction filter for reducing fluctuations in the output signal of a pendulous integrating gyro accelerometer

Abstract To address the reduction in instantaneous accuracy caused by the fluctuations in the output signal of a pendulous integrating gyro accelerometer (PIGA), these fluctuations were analyzed and primarily attributed to the subdivision errors of the resolver and the dynamic tracking errors of the servo loop. A novel cascaded PI-LMS adaptive prediction filter, characterized by fast convergence and strong noise suppression capability, was proposed to reduce these output fluctuations. The filter’s performance was validated through simulations and further confirmed in both static and dynamic tests. In dynamic tests, the cascaded PI-LMS algorithm achieved 31.3% faster convergence (0.184 s) compared to LMS and reduced the noise power spectral density by 24.44 dB, effectively suppressing PIGA’s output fluctuations and enhancing its instantaneous accuracy.