Requirements for pointing accuracy in Earth observation missions have tightened tolerable microvibration levels on satellites. Various mitigation techniques, including active, passive, and hybrid methods, have been proposed to prevent microvibration propagation. The microvibration damping (MIVIDA) platform was developed to isolate sensitive optical payloads from disturbances, using passive dampers, proof mass actuators (PMA) for 6 DoF force tensor creation, and an interface for different payloads. Accelerometer measurements at the payload interface are used in closed loop to reject vibrations.
This PhD thesis presents methodologies for designing data-driven robust linear parameter-varying (LPV) controllers for adaptive disturbance rejection. The goal is to isolate sensitive payloads from unknown sinusoidal microvibration sources with varying harmonic frequencies. Integral Quadratic Constraints (IQCs) represent and analyse various uncertainty types, including parametric, rate-bounded, time delays, and non-linearities. IQC-based stability analysis for LPV control design with frequency-domain methods is proposed, allowing the determination of a range for the time-varying scheduling parameter for which closed-loop stability is ensured.
An LPV controller synthesis approach is developed for the adaptive rejection of sinusoidal disturbances with varying harmonic frequencies, reducing computational complexity. The LPV controller, designed using the frequency response function (FRF) of a linear time-invariant (LTI) multiple-input multiple-output (MIMO) system, stabilises the system for fast disturbance frequency variations. Global stability is achieved using IQCs to represent frequency variations. The LPV controller is computed by solving a convex optimisation problem in the frequency domain. An LPV controller for rejecting unknown frequency-varying sinusoidal disturbances is implemented on the MIVIDA platform, demonstrating effective disturbance rejection with up to 50.40 dB attenuation and stability against fast or rate-bounded disturbance frequency variations.
Robust performance objectives for LPV controllers are proposed, significantly increasing attenuation during transitory phases and when the estimated scheduling parameter deviates from the true value. A composite linear matrix inequality (LMI) constraint based on an equivalent IQC combines performance and robustness channels, ensuring robust performance.
A novel robust controller synthesis approach is proposed that covers model uncertainty with an elliptical set. Given a set of FRFs of LTI MIMO systems, the approach determines the best linear nominal model and corresponding elliptical uncertainty set. Using a novel split representation, the uncertainty set is represented as an equivalent IQC, integrated into a data-driven frequency-domain controller synthesis method using convex optimisation. The experimental results show an improved attenuation performance compared to the classical methods.
A robust adaptive control scheme for active microvibration rejection is designed. An end-to-end performance analysis, from microvibration to line-of-sight (LoS), is key. A test campaign using a laser-based payload assesses LoS stability, improved by 34.18 dB. In addition, a feasibility experiment evaluates the methods for fault-tolerant control, showing great potential for the development of more robust active systems in space.