Disturbance-Compression Extended State Observer With Noise Insensitivity: Application to Electro-Optical Tracking System

While augmenting the observer gain can diminish the estimation errors of the extended state observer (ESO), it inevitably leads to an amplification of the observing noise, thus creating the trade-off between high-gain observation and noise sensitivity. To address the trade-off, this paper introduces the disturbance-compression extended state observer (DC-ESO) from a frequency-domain perspective. By employing disturbance compression, the DC-ESO modifies the disturbance characteristics encountered by the observer, effectively diminishing the upper bounds of the disturbance and its derivative. Frequency-domain analysis indicates that the DC-ESO necessitates substantially lower observer gain than the ESO while maintaining the same estimation errors. The principles of disturbance compression and its effects on disturbance estimation and noise amplification are elucidated, revealing that the DC-ESO effectively alleviates the inherent conflict between high-gain observation and susceptibility to noise. The parameter tuning methodology and frequency-domain stability criteria are user-friendly for practicing engineers. The effectiveness of the proposed approach is empirically verified through its application to an electro-optical tracking system. Note to Practitioners—This work tackles the disturbance rejection problem in control systems, which encompasses disturbance including mechanical vibrations, friction, and parametric perturbations. Existing extended state observers (ESOs) rely on high gain (or high bandwidth) for disturbance suppression, which unfortunately results in substantial amplification of driving noise. In this research, we incorporate a disturbance compression channel that divides the compensation of disturbance into two separate stages. This approach effectively pre-compresses the disturbance, thus reducing the observational burden on the observer. We reconstruct the ESO in the frequency domain as a transfer function and present a frequency-domain disturbance compression ESO that is easily grasped by engineers. Accordingly, the proposed method is highly practical and effective according to experiments. Future research may consider expanding the targeted suppression to the specific harmonics disturbance and nonlinear system.