Safety-Critical Control for Aerial Physical Interaction in Uncertain Environment

Jeonghyun Byun, Yeonjoon Kim, Dongjae Lee, H. Jin Kim · 2025

Read Safety-Critical Control for Aerial Physical Interaction in Uncertain Environment by Jeonghyun Byun, Yeonjoon Kim, Dongjae Lee, H. Jin Kim (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-2501-16719.txt.

Introduction

Jeonghyun Byun and co-authors design safety-critical control for drones that physically touch their surroundings in a 2025 paper. Aerial manipulators fly up to structures to inspect, push, or repair them, trading the safety of distance for useful contact. Contact in uncertain settings risks crashes when forces surprise the controller.

Their approach pairs a disturbance observer, which estimates unexpected forces in real time, with a safety filter. The filter bends the planned path whenever safety margins shrink, for both still and moving structures. Tests put a fully actuated vehicle against real interaction uncertainty. The work matters for power line and bridge work, where drones are asked to do with touch what cameras alone cannot.

Catalog record

CREATORJeonghyun Byun, Yeonjoon Kim, Dongjae Lee, H. Jin Kim
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE38 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2501-16719

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Safety-Critical Control for Aerial Physical Interaction in Uncertain Environment Jeonghyun Byun, Yeonjoon Kim, Dongjae Lee, H. Jin Kim Abstract Aerial manipulation for safe physical interaction with their environments is gaining significant momentum in robotics research. In this paper, we present a disturbance-observer-based safety-critical control for a fully actuated aerial manipulator interacting with both static and dynamic structures. Our approach centers on a safety filter that dynamically adjusts the desired trajectory of the vehicle’s pose, accounting for the aerial manipulator’s dynamics, the disturbance observer’s structure, and motor thrust limits. We provide rigorous proof that the proposed safety filter ensures the forward invariance of the safety set—representing motor thrust limits—even in the presence of disturbance estimation errors. To demonstrate the superiority of our method over existing control strategies for aerial physical interaction, we perform comparative experiments involving complex tasks, such as pushing against a static structure and pulling a plug firmly attached to an electric socket. Furthermore, to highlight its repeatability in scenarios with sudden dynamic changes, we perform repeated tests of pushing a movable cart and extracting a plug from a socket. These experiments confirm that our method not only outperforms existing methods but also excels in handling tasks with rapid dynamic variations. I INTRODUCTION Aerial manipulators have gained popularity in robotics research due to their ability to combine the maneuverability of unmanned aerial vehicles (UAVs) with the versatility of robotic manipulators. Their capability to physically interact with the environment makes them ideal for a wide range of applications involving physical interaction such as drawer-opening [ 1 ] , door-opening [ 2 ] , plug-pulling [ 3 ] , window-cleaning [ 4 ] , multi-manual object manipulation [ 5 ] , non-destructive testing (NDT) [ 6 ] and heavy object pushing [ 7 ] . Such tasks, often referred to as aerial physical interaction (APhI) depicted in Fig. 1 , require consideration of two important factors when designing controllers. The first is the external disturbance generated by the physical interaction between the vehicle and the surrounding environment, and the second is the vehicle’s actuation limit such as motor thrust limit. However, very few studies have focused on controller design for APhI that simultaneously accounts for both disturbance attenuation and actuator limitations. Figure 1: Aerial physical interaction (APhI) experiments conducted for the validation of the proposed method. (a) Pushing a static structure. (b) Pulling an object from a static structure. (c) Pushing a movable object. In [ 8 ] , the control of a fully actuated UAV with actuator saturation was introduced, and in [ 9 ] , the authors presented an actuator limit-aware control strategy for an aerial manipulator physically interacting with a human operator. However, they did not consider the effect of external disturbances or model uncertainties when designing their controllers and only conducted numerical simulations. In [ 10 ] and [ 11 ] , energy tank-based controllers for aerial manipulators interacting with uncertain dynamic environments were presented. They proved the vehicle’s safety under the effect of external disturbances or model uncertainties. Also, [ 12 ] proposed a power-based safety layer for aerial vehicles physically interacting with surrounding environments, and [ 13 ] introduced a control method for human-robot APhI that considers motor thrust limitations and external disturbances. However, these four controllers assume that the interaction wrench can be measured by a force/torque sensor or estimated based on a derived model. This reliance may not be suitable for aerial manipulators not equipped with force/torque operating in windy environments. [ 14 ] presented a saturated robust controller for a fully actuated multirotor taking into account the disturbance rejection and rotor thrust saturation. However, they did not consider the situation that reaching the desired pose of the vehicle is not dynamically feasible, e.g., the magnitude of an external disturbance exceeds the motor thrust limit, and only conducted numerical simulations. In this paper, we present a disturbance-observer (DOB)-based safety-critical control of a fully actuated aerial manipulator physically interacting with static or movable structures. To that end, we derive the dynamic model of the aerial manipulator configured with a fully actuated hexacopter and a rigidly attached robotic arm. Then, the control framework including a DOB-based control law and a safety filter is introduced. While the DOB-based control law is designed as presented in [ 15 ] , for the safety filter, we develop a system model that combines the aerial manipulator’s model and DOB structure and formulate an optimization problem to determine the desired pose and twist of the vehicle that conforms to the constraints on motor thrust. In particular, our main contributions are arranged as follows: • We design a safety-critical controller that can conduct various types of APhI with uncertain environments in the absence of interaction wrench measurement or estimation. • We show that the safety set representing the motor thrust limits is forward invariant under the proposed safety filter. • We validate that our controller outperforms existing approaches for APhI control through pushing and pulling experiments with both static and dynamic structures. This paper is outlined as follows; In Section II, we formulate the dynamic model of our fully actuated aerial manipulator, and the control framework that includes DOB-based control law and safety filter is introduced in Section III. In Section IV, we present a theoretical analysis on the safety filter, and Section V covers the experimental validation of the proposed method. Notations: \boldsymbol{0_{i\times j}} , \boldsymbol{I_{i}} , \boldsymbol{e_{3}} and \boldsymbol{E_{i}}\in{\mathbb{R}}^{6} represent the i\times j zero matrix, i\times i identity matrix, [0;0;1] and {\mathbb{R}}^{6} vector where its i^{\textrm{th}} element is one and the other elements are zero, respectively. For scalars a_{1},\cdots,a_{N} , we let ca_{1} and sa_{1} denote \cos{a_{1}} and \sin{a_{1}} , respectively, and diag \{a_{1},\cdots,a_{N}\} represents an N\times N diagonal matrix where the (i,i)^{\textrm{th}} element is a_{i} and and all off-diagonal elements are zero. For vectors \boldsymbol{\alpha} and \boldsymbol{\beta} , we let \alpha_{i} denote the i -th element of \alpha , and if \boldsymbol{\alpha}\in{\mathbb{R}}^{3} and \boldsymbol{\beta}\in{\mathbb{R}}^{3} , [\boldsymbol{\alpha}]_{\times}\in\mathbb{R}^{3\times 3} means the operator which maps \boldsymbol{\alpha} into a skew-symmetric matrix such as [\boldsymbol{\alpha}]_{\times}\boldsymbol{\beta}=\boldsymbol{\alpha}\times\boldsymbol{\beta} . Also, for matrices \boldsymbol{A_{1}},\cdots,\boldsymbol{A_{N}} , \textrm{blkdiag}\{\boldsymbol{A_{1}},\cdots,\boldsymbol{A_{N}}\} represents a block diagonal matrix obtained by aligning \boldsymbol{A_{1}},\cdots,,\boldsymbol{A_{N}} . Moreover, we abbreviate the phrase ”with respect to” to w.r.t.. II Modelling Figure 2: An aerial manipulator configured with a fully actuated multirotor and a rigidly attached robotic arm. The thrust values T_{1},\cdots,T_{6} configure the thrust vector \boldsymbol{T}\in{\mathbb{R}}^{6} , and \mathscr{F}_{W} and \mathscr{F}_{b} represent the Earth-fixed and multirotor frames, respectively. Fig. 2 illustrates an aerial manipulator configured with a fully actuated multirotor with six tilted motors and a rigidly attached robotic arm. Also, two key coordinate frames, the Earth-fixed frame \mathscr{F}_{W} and the multirotor frame \mathscr{F}_{b} , are displayed. The generalized coordinate of the aerial manipulator, \boldsymbol{q}\triangleq[\boldsymbol{p};\boldsymbol{\phi}] , consists of the position of \mathscr{F}_{b} w.r.t. \mathscr{F}_{W} expressed in \mathscr{F}_{W} , \boldsymbol{p}\in{\mathbb{R}}^{3} , and ZYX Euler angles, \boldsymbol{\phi}\in{\mathbb{R}}^{3} . Then, the Euler-Lagrange model of the aerial manipulator is formulated as follows [ 14 ] : \begin{split}\boldsymbol{M}(\boldsymbol{\phi})\boldsymbol{\ddot{q}}+\boldsymbol{C}(\boldsymbol{\phi},\boldsymbol{\dot{\phi}})+\boldsymbol{G}=\boldsymbol{\tau}+\boldsymbol{\tau_{ext}}\end{split} (1) where \boldsymbol{\tau}\in{\mathbb{R}}^{6} and \boldsymbol{\tau_{ext}}\in{\mathbb{R}}^{6} represent the generalized control wrench and external disturbance, respectively, and \begin{split}\boldsymbol{M}(\boldsymbol{\phi})\triangleq&\begin{bmatrix}m\boldsymbol{I_{3}}&\boldsymbol{0_{3\times 3}}\\ \boldsymbol{0_{3\times 3}}&\boldsymbol{Q^{\top}}\boldsymbol{J}\boldsymbol{Q}\end{bmatrix},\\ \boldsymbol{C}(\boldsymbol{\phi},\boldsymbol{\dot{\phi}})\triangleq&\begin{bmatrix}\boldsymbol{0_{3\times 1}}\\ \boldsymbol{Q^{\top}}(\boldsymbol{J}\boldsymbol{\dot{Q}}\boldsymbol{\dot{\phi}}+[\boldsymbol{Q}\boldsymbol{\dot{\phi}}]_{\times}\boldsymbol{J}\boldsymbol{Q}\boldsymbol{\dot{\phi}})\end{bmatrix},\ \boldsymbol{G}\triangleq\begin{bmatrix}mg\boldsymbol{e_{3}}\\ \boldsymbol{0_{3\times 1}}\end{bmatrix}\end{split} with the mass and moment of inertia of the aerial manipulator, m and \boldsymbol{J}\in{\mathbb{R}}^{3\times 3} , and the gravitational acceleration, g

Read the complete text