Solar Cruiser Disturbance Torque Estimation and Predictive Momentum Management

TL;DR

The paper tackles solar sail momentum management for Solar Cruiser by integrating real-time disturbance estimation with a model predictive controller that coordinates an AMT and RCD actuators across a 4-RW assembly. A Kalman filter estimates unmodeled disturbance torques and model errors, feeding a linearized, discrete-time prediction model used in a convex MPC that enforces RW saturation constraints and soft momentum bounds via slack variables. The approach also employs a pseudo-inverse RW allocation to map predictions to individual wheel commands and uses PWM quantization to realize on-off RCD actuation, enabling real-time onboard implementation. Simulation results show the KF-augmented MPC outperforms NASA's threshold-based method, enabling larger slews with lower AMT movement and robust RW desaturation, thus extending mission capability and longevity for solar-sail platforms. The work establishes a practical benchmark for MPC-based momentum management in solar sails employing AMT and/or RCDs and indicates clear paths toward flight-testing and deployment on future missions.

Abstract

This paper presents a novel disturbance-torque-estimation-augmented model predictive control (MPC) framework to perform momentum management on NASA's Solar Cruiser solar sail mission. Solar Cruiser represents a critical step in the advancement of large-scale solar sail technology and includes the innovative use of an active mass translator (AMT) and reflectivity control devices (RCDs) as momentum management actuators. The coupled nature of these actuators has proven challenging in the development of a robust momentum management controller. Recent literature has explored the use of MPC for solar sail momentum management with promising results, although exact knowledge of the disturbance torques acting on the solar sail was required. This paper amends this issue through the use of a Kalman filter to provide real-time estimation of unmodeled disturbance torques. Furthermore, the dynamic model used in this paper incorporates key fidelity enhancements compared to prior work, including the Solar Cruiser's four-reaction-wheel assembly and the offset between its center of mass and center of pressure. Simulation results demonstrate that the proposed policy successfully manages angular momentum growth under slew maneuvers that exceed the operational envelope of the current state-of-the-art method. The inclusion of the disturbance torque estimate is shown to greatly improve the reliability and performance of the proposed MPC approach. This work establishes a new benchmark for Solar Cruiser's momentum management capabilities and paves the way for MPC-based momentum management of other solar sails making use of an AMT and/or RCDs.

Figures (7)

• Figure 1: Depictions of the Solar Cruiser model used in this paper (not drawn to scale) highlighting (a) its attitude control and momentum management actuators, including four RWs (light blue), AMT (red) and RCDs (brown and orange); and (b) the definition of key bodies and, such as the sail $\mathcal{S}$ with CM $s$ (also the CP of the entire sailcraft) and the bus $\mathcal{P}$ with CM $s$, as well as the entire sailcraft's CM $c$.
• Figure 2: Illustration of the MPC soft constraint design with a prediction horizon of $N=5$, where no penalty is incurred for responses satisfying $h^\text{soft}_\text{min} \leq h \leq h^\text{soft}_\text{max}$, while a quadratic penalty appears in the MPC objective function when $h^\text{soft}_\text{max} \leq h$ or $h \leq h^\text{soft}_\text{min}$. The response labeled "MPC design 1" indicates a design that violates the soft constraint, while "MPC design 2" does not.
• Figure 3: Simulation results using NASA's Solar Cruiser momentum management strategy from Inness2023MMTyler2024, featuring RW saturation under a $4^\circ$ slew compared to a $3^\circ$ slew. The black dashed line in (c) denotes the activation threshold ($50\%$ of the soft constraint in MPC) , and the green dashed line denotes the deactivation threshold ($50\%$ of the activation threshold).
• Figure 4: Simulation results using the proposed MPC momentum management strategy under a $6$ deg slew with and without (nominal) the disturbance estimate knowledge in prediction model. The black dashed lines in (b) denote the $25\%$ soft constraint on 4-RWs angular momentum.
• Figure 5: Threshold tuning using $10\%$ AMT threshold and $30\%$ RCD threshold (blue), and using $20\%$ AMT threshold and $60\%$ RCD threshold (red). The zoomed-in plot in (d) demonstrates the first $2000$ seconds of slew maneuver and the PWM-quantized RCD actuation pulsing at every time step.