Strain-Parameterized Coupled Dynamics and Dual-Camera Visual Servoing for Aerial Continuum Manipulators
Niloufar Amiri, Farrokh Janabi-Sharifi · 2026
Read Strain-Parameterized Coupled Dynamics and Dual-Camera Visual Servoing for Aerial Continuum Manipulators by Niloufar Amiri, Farrokh Janabi-Sharifi (2026) 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-2603-23333.txt.
Introduction
Amiri and Janabi-Sharifi model and control a tendon-driven arm on a drone. Such aerial continuum manipulators mix drone agility with a soft arm's compliance. Past coupled models cost compute and skip the drone's underactuation. Their formulation joins a strain-based Cosserat rod model with the drone's rigid-body model in one frame. It avoids long symbolic derivations and runs as plain ODEs on SE(3). They add a dual-camera visual servoing scheme with wide coverage. Tests on a prototype show it works in real settings.
Catalog record
| CREATOR | Niloufar Amiri, Farrokh Janabi-Sharifi |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 51226 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.23333
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Strain-Parameterized Coupled Dynamics and Dual-Camera Visual Servoing for Aerial Continuum Manipulators Niloufar Amiri, Farrokh Janabi-Sharifi Abstract Tendon-driven aerial continuum manipulators (TD-ACMs) combine the maneuverability of uncrewed aerial vehicles (UAVs) with the compliance of lightweight continuum robots (CRs). Existing coupled dynamic modeling approaches for TD-ACMs incur high computational costs and do not explicitly account for aerial platform underactuation. To address these limitations, this paper presents a generalized dynamic formulation of a coupled TD-ACM with an underactuated base. The proposed approach integrates a strain-parameterized Cosserat rod model with a rigid-body model of the UAV into a unified Lagrangian ordinary differential equation (ODE) framework on SE ( 3 ) \mathrm{SE}(3) , thereby eliminating computationally intensive symbolic derivations. Building upon the developed model, a robust dual-camera image-based visual servoing (IBVS) scheme is introduced. The proposed controller mitigates the field-of-view (FoV) limitations of conventional IBVS, compensates for attitude-induced image motion caused by UAV lateral dynamics, and incorporates a low-level adaptive controller to address modeling uncertainties with formal stability guarantees. Extensive simulations and experimental validation on a compact custom-built prototype demonstrate the effectiveness and robustness of the proposed framework in real-world scenarios. I Introduction Uncrewed aerial manipulators (UAMs) extend the capabilities of conventional uncrewed aerial vehicles (UAVs) by enabling physical interaction with the environment. Early UAM platforms predominantly employed rigid-link aerial manipulators (RA-AMs), which benefit from well-established rigid-body modeling and control methodologies [ 18 ] . However, their limited compliance restricts safe and robust operation in uncertain or contact-rich environments. To overcome these limitations, aerial continuum manipulators (ACMs) have emerged as lightweight and compliant alternatives [ 24 , 12 ] . In tendon-driven ACMs (TD-ACMs), the manipulator exhibits distributed deformation, configuration-dependent stiffness, and strong bidirectional coupling between strain and applied forces. Although compliance enhances safety and adaptability, it significantly complicates the modeling, state estimation, and control of TD-ACMs, leaving several fundamental challenges unresolved. I-A Modeling Mechanical models are essential for model-based control design and simulation-based analysis. While kinematic and quasi-static formulations [ 3 , 19 ] provide useful approximations, they are unable to capture the highly nonlinear behavior of TD-ACMs and the dynamic coupling between UAV motion and continuum robot (CR) deformation. The primary challenge lies in incorporating the distributed dynamics of the CR and its strong bidirectional coupling with the underactuated UAV platform [ 31 ] . Unlike RA-AMs, where link inertias and joint coordinates are explicitly defined, CRs exhibit distributed curvature and strain, resulting in effectively infinite degrees of freedom (DoFs), configuration-dependent compliance, and distributed internal forces. These characteristics render simplified geometric models, such as those based on constant-curvature (CC) approximations [ 10 ] or classical beam theories [ 16 ] , insufficient accurately predicting CR dynamics under large deflections, external loading, and dynamic flight conditions. More accurate formulations, such as lumped-mass and Cosserat rod models [ 31 , 21 , 4 ] , often incur prohibitive computational costs. Existing approaches typically adopt decoupled modeling strategies, treating the UAV–CR interaction as a disturbance [ 24 , 7 , 11 ] . This assumption breaks down when the manipulator inertia is non-negligible, during load transportation tasks, or under aggressive maneuvers. Fully coupled models have been proposed using finite-dimensional Lagrangian formulations [ 26 , 25 ] or augmented rigid-body representations [ 30 ] ; however, these approaches either rely on the CC assumption or suffer from computational inefficiency. To the best of the authors’ knowledge, a fully coupled, geometrically accurate, and computationally efficient dynamic model for TD-ACMs has not yet been reported in the literature. I-B Low-Level Model-Based Control A control framework capable of handling uncertainties is essential for reliable operations of TD-ACMs. They are subject to substantial uncertainties arising from unmodeled dynamics, backbone material variability, tendon friction and hysteresis, and aerodynamic disturbances. Furthermore, the underactuation of commonly employed UAVs introduces additional coupling between translational and rotational dynamics. These factors, combined with strong UAV–CR coupling, render kinematic [ 6 , 19 ] and decoupled dynamic control strategies [ 24 , 11 ] inadequate for achieving robust performance. Existing coupled approaches often rely on simplified CC-based models [ 26 , 25 , 30 ] , which limit their accuracy under large deflections, distributed external loading, and high-speed maneuvers. Furthermore, fully coupled control methods remain largely confined to simulation-based validation due to the absence of computationally tractable dynamic models. In summary, TD-ACMs require coupled, model-based dynamic control frameworks that balance accuracy and computational efficiency while ensuring robustness to system and environmental uncertainties. Such unified solutions, however, remain an open challenge. I-C High-level Visual Servo Control While low-level dynamic control can mitigate TD-ACM parametric uncertainties, successful aerial manipulation ultimately depends on reliable perception and task-level regulation, for instance, to track a target with uncertain or time-varying position. For localization, conventional UAM systems often rely on motion-capture systems, the global positioning system (GPS), or simultaneous localization and mapping (SLAM), whose performance may degrade in GPS-denied, cluttered, or feature-poor environments. In the absence of joint encoders and explicit forward kinematic formulations, these challenges are further amplified in TD-ACMs due to indirect and uncertain state estimation of the CR. Visual servoing (VS) offers a compelling alternative by reducing reliance on global localization systems, enabling direct regulation of the end-effector pose (e.g., relative to the target), and improving robustness to calibration errors, particularly in image-based visual servoing (IBVS) [ 5 , 13 ] . For TD-ACMs, VS is especially attractive because it circumvents the need for precise tip-state reconstruction from uncertain strain measurements. However, VS design for TD-ACMs differs fundamentally from that for RA-AMs. In eye-in-hand (EIH) configurations, for instance, the camera pose depends on the deformable continuum shape, resulting in a more complex and nonlinear image feature mapping [ 27 ] . Moreover, CR flexibility may induce image distortion during motion [ 17 ] , while frequent occlusions and field-of-view (FoV) loss further complicate control. Eye-to-hand (ETH) configurations improve global visibility but typically reduce spatial resolution and maneuverability [ 32 , 28 , 36 ] . Dual-camera architectures, combining EIH and ETH configurations, leverage complementary advantages by enabling supervisory FoV management and high-resolution feedback for accurate local tracking. While dual-camera VS has been applied to rigid manipulators [ 9 ] and fixed-base CRs [ 2 , 35 ] , it has not yet been developed for TD-ACMs. I-D Approach and Contributions This work addresses the aforementioned challenges through an integrated vision-based control framework for TD-ACMs. The main contributions are summarized as follows: • A unified, multi-level dual-camera vision-based control framework for robust TD-ACM operation, integrating a high-level IBVS scheme with a low-level adaptive coupled controller. • A dual-camera VS architecture that leverages the robustness of IBVS to modeling errors, mitigates FoV loss, and compensates for UAV underactuation directly in the image space. • An adaptive, model-based coupled controller with formal stability guarantees, enabling accurate low-level regulation under CR parametric uncertainties and strong UAV-CR coupling. • A geometrically exact, strain-parameterized coupled dynamic model on \text{SE}(3) that, unlike existing Lagrangian formulations [ 25 ] , preserves CR stiffness and damping, accounts for distributed body wrenches, and enables efficient computation without requiring computationally intensive symbolic derivations. • Extensive simulation and experimental validation on a custom TD-ACM platform, demonstrating robustness to parametric uncertainties and environmental disturbances, such as mechanical vibrations. II Strain-Parameterized Coupled Modeling Fig.