Development of Variable-Structure Algorithms for a Satellite Attitude Control

Abstract

Attitude control of satellites is a challenging problem especially if the requirements on transient performance and energy consumption are tight. PID control is by far the most used in practice for attitude control, due to its simple implementation and effectiveness. The present work is divided into three main areas. First, a novel variable-structure (VS) control strategy is proposed to address the attitude control problem. This VS control does not lead to a sliding mode; instead, it consists of two separate PID controllers and a switching logic. The state-space is partitioned into a few domains or areas, and each of these areas is assigned to one of the two PID controllers. Then, based on the current states of the system, one of the two PID controllers is selected. Different configurations of the state-space partitioning are considered. Results show that the novel VS PID provides improvement in transient performance and energy consumption over the conventional PID control for specific operating conditions. Second, non-parametric auto-tuning based on continuous oscillations is investigated; since manual tuning of controllers in-orbit is a tedious process and would generally lead to non-optimal performance. To excite these oscillations a discontinuous closed loop test utilizing two relays is used. Optimal tuning rules are developed to map the measured amplitude and frequency of the excited oscillations to optimal controller gains. For a class of satellites that vary widely is size and dynamics, near-optimal performance is achieved with no more than 4% deterioration in transient performance in the worse-case scenario. Third, an experimental testbed is developed and fabricated to validate the theory and numerical simulations. The testbed utilizes magnetic levitation to provide a single-axis frictionless rotation, which emulates the principal dynamics of a satellite in space. The testbed shows repeatability of the experiments, and results agree with the theoretical analysis.

Date of Award	20 Dec 2023
Original language	American English
Supervisor	IGOR Boiko (Supervisor)