Safe On-Orbit Dislodging of Deployable Structures via Robust Adaptive MPC
Longsen Gao, Claus Danielson, Andrew Kwas, Rafael Fierro · 2025
Read Safe On-Orbit Dislodging of Deployable Structures via Robust Adaptive MPC by Longsen Gao, Claus Danielson, Andrew Kwas, Rafael Fierro (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-16849.txt.
Introduction
Longsen Gao and co-authors design a controller that can safely free a stuck solar panel on a space station. Their 2025 paper treats the task as a robot arm job in orbit. The jammed panel sits on a hinge whose traits change over time, and the station must not be harmed.
Their method learns the hinge's traits on the fly, within safe bounds, while it plans each move. The cost it weighs blends two drives: probing the system to shrink doubt, and acting well once the traits are known. The team tests the plan in computer runs of the freeing task and in lab trials, both in free fall and on the ground. It beats rival schemes on how well it learns the traits and how well it steers.
Catalog record
| CREATOR | Longsen Gao, Claus Danielson, Andrew Kwas, Rafael Fierro |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 62 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-16849
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Safe On-Orbit Dislodging of Deployable Structures via Robust Adaptive MPC Longsen Gao, Claus Danielson, Andrew Kwas, Rafael Fierro Abstract This paper proposes a novel robust adaptive model predictive controller for on-orbit dislodging. We study orbit dislodging where a servicing spacecraft uses a robotic arm to free a jammed and unactuated solar panel mounted on a hybrid hinge that acts as a time-varying client on a space station. Our method couples online set-membership identification with a robust adaptive MPC to enforce safety under bounded disturbances. The controller explicitly balances exploration to excite the system and shrink uncertainty and exploitation to improve control performance through a dual-mode cost. The feasibility of the developed robust adaptive MPC method is also examined through dislodging simulations and hardware experiments in freefall and terrestrial laboratory environments, respectively. In addition, the advantages of our method are shown through comparison experiments with several state-of-the-art control schemes for both accuracy of parameter estimation and control performance. NOMENCLATURE M Inertia matrix. C Coriolis and centripetal torque term. \mathcal{K} Stiffness parameter. \mathcal{D} Grasp factor related to contact geometry. A State matrix of the linear parametric model. A_{i} Affine decomposition matrix of A(\rho_{k}) . B Input matrix of the linear parametric model. B_{i} Affine decomposition matrix of B(\rho_{k}) . \mathcal{P} Admissible parameter polytope. \mathbb{W} Bounded disturbance set. H_{\rho} Polytope matrices defining \mathcal{P} . H_{w} Polytope description of \mathbb{W} . \mathcal{U} Input constraint set. \mathcal{X} State constraint set. \mathbb{Z} Combined state–input constraint set. \tau_{f} Coulomb friction torque. \mathbf{K} Stabilizing feedback gain used in the tube policy. \mathbf{A}_{\text{cl}} Closed-loop map operator. \Gamma_{j,\varsigma|k} Nonnegative multipliers certifying tube constraints. \mathbb{X}_{T} Terminal set with robust positive invariance in tube variables. Q{,}R{,}\Upsilon Weighting matrix in the cost. \varOmega Collection of decision variables. \lambda Scaling factor used with \mathbb{X}_{T} in the terminal condition. J_{N} Finite-horizon objective. \mu Friction coefficient. r_{e} Effective rotor radius. \epsilon Small positive smoothing constant. \tau_{m,k} External torque disturbance applied by the manipulator. \delta_{\tau} Bound on the external torque disturbance. \varsigma Step index along the prediction horizon. \alpha_{k} Angle between gripper direction and rod axis at time k . d_{g} Dislodging contact position measured from the hinge axis. r_{\alpha} Wrench direction vector used in the torque transfer term. \boldsymbol{\rho} Parameter vector. \zeta_{k} Viscous damping parameter. \phi_{\rho} Bound on the rate of change of \boldsymbol{\rho}_{k} . v_{\varsigma|k} Auxiliary input in the tube policy. c_{\varsigma|k} Auxiliary input used in the convex reformulation. \ell Stage and terminal cost function. I Introduction On-orbit failures have frequently occurred ever since the first artificial satellite was launched into space [ 1 , 2 , 3 , 4 ] due to multiple factors including orbital debris [ 5 ] , thermal cycling and temperature extremes [ 6 ] , radiation-induced degradation and surface charging [ 7 ] , vacuum-induced cold welding and stiction [ 8 ] , and electromagnetic interference [ 9 ] . These failures have caused countless economic and scientific losses, which have become an enormous challenge in the past decades. On-orbit servicing leveraging autonomous robotics systems is an increasingly popular research topic for rescuing on-orbit failure [ 10 , 11 , 12 , 13 ] . Additionally, servicing missions confront two fundamental challenges. Model uncertainty from partially known and time-varying client parameters and failure modes degrades prediction and planning, while operational limits including limited sensing [ 14 ] , tight actuation bounds [ 15 ] , and strict real-time requirements shrink the feasible set of safe control actions [ 16 ] . Dislodging jammed components, where off-nominal friction temporarily raises the torque demand above the joint’s nominal requirement, remains a common on-orbit maintenance task [ 17 , 18 , 19 ] . It involves safely manipulating objects that may be stuck or restrained due to various factors, such as microgravity [ 20 ] , stiction [ 18 ] , or mechanical failure [ 2 ] . Dislodging a jammed component often requires applying significant, sudden force, which could potentially damage a delicate component. Few studies have considered using on-orbit robotics systems to address the problem. Previous work in [ 21 ] presented an adaptive control algorithm using a multi-robot system in a dislodging task for a solar panel. However, the algorithm did not consider safety constraints, which can lead to damage of both the servicer and client spacecraft. The risk of damage is exacerbated during the transient period when the adaptive controller is learning the system parameters. Although the adaptive controller is guaranteed to eventually learn the correct parameter, during the learning process, the adaptive controller can exert excessive force to dislodge the jammed component, potentially causing damage. Online parameter estimation can be leveraged to address the challenge of uncertainty during servicing missions to guarantee safety. Adaptive control for updating controller parameters using real-time measurement data has been extensively studied and gained popularity in recent decades [ 22 , 23 , 24 ] . Traditional Model Reference Adaptive Control (MRAC) [ 25 ] updates controller parameters to track a reference model but is sensitive to unmodeled dynamics and does not natively enforce state or input constraints , which can be unreliable in practical implementation due to unmodeled dynamics. Failure to enforce constraints on both the manipulator and manipulated object may lead to component damage during the servicing [ 26 ] , which can cause significant costs and lost time due to the need to replace aerospace components via relaunch. Model predictive control (MPC) is widely adopted in terrestrial industries for explicit handling of constraints and for stability assurance, while within spacecraft guidance and control it is an limited but worthy of development and expanding research area with selective applications based on its strong safety and reliability. In [ 27 ] , an adaptive MPC algorithm for linear systems was introduced using a recursive approach, and its extension in [ 28 ] employed a tube-based framework to handle parametric uncertainty and additive disturbances in nonlinear systems. However, these methods rely on passive identification, preventing full exploitation of concurrent adaptation and control. A learning-based MPC in [ 29 ] used set-membership identification composed of two online phases: an adaptation phase with rigid tube-based robust MPC and a learning phase for uncertainty estimation, but only considers constant parameters, making it unsuitable for time-varying applications such as those found in aerospace. To address these limitations, this paper considers an adaptive controller that integrates set-membership identification to iteratively refine parameter bounds and employs robust MPC to ensure constraint enforcement and stability under worst-case conditions, ultimately improving performance as uncertainty estimates become less conservative. In this paper, we propose a novel robust adaptive MPC with set-membership parameter estimation for the problem of safe dislodging. We demonstrate its implementation on the servicer during a dislodging task based on the client. The main contributions are summarized: • Our robust adaptive MPC method provides robust guarantees on constraint enforcement despite the parametric uncertainty of the client and unpredictable failure modes. • We derive a state-space representation of the hinge model and implement the proposed MPC for the dislodging task. Comparative studies against two baseline methods evaluate feasibility and performance. • We develop a novel cost function that incorporates the time-varying parameter set with control input and state to improve the performance of parameter estimation during the control process. The