Aerial Manipulators: UAVs with Dexterous Arms

• Aerial manipulators are integrated robotic systems that combine UAVs with onboard arms to enable dexterous interaction with complex environments.
• They utilize varied mechanical architectures, such as cable-driven continuum and tendon-driven designs, to enhance flight stability and optimize payload efficiency.
• Recent research advances focus on robust nonlinear control, whole-body motion planning, and sensor-based compliance to achieve high-precision contact tasks.

Aerial manipulators are integrated robotic platforms that combine unmanned aerial vehicles (UAVs) with onboard manipulators—either discrete-link or continuum—enabling dexterous interaction with the environment from flight. These systems leverage the mobility and spatial access of UAVs but face distinct challenges in flight stability, dynamic coupling, and control during manipulator operation and contact tasks. Recent advances have focused on mechanical design (including cable-driven continuum arms, tendon-driven anthropomorphics), robust nonlinear control, optimization-based whole-body planning, and model-based compliance. The following sections synthesize the state of the art in aerial manipulators across mechanical architectures, modeling, control, planning, compliance, and integration, with reference to foundational and recent research.

1. Mechanical Architectures and System Design

Aerial manipulators comprise a flight base (multirotor or tilt-rotor for full actuation) and a manipulator—ranging from rigid-link serial arms to continuum cable-driven structures and tendon-based anthropomorphic arms.

• Cable-Driven Continuum Arms: Lightweight, modular continuum manipulators use a superelastic NiTi backbone with evenly spaced spacer disks, actuated by four cable tendons routed at $90^\circ$ intervals (pitch circle radius $r\approx 12$ mm, backbone length $L\approx 180$ mm). The arm is driven by compact servos (Dynamixel XL430) and coordinated by an onboard microcontroller (Teensy 4.0), achieving sub-500 g system mass and 0.2 kg tip payload for mid-size UAVs. Modular connectors allow stacking for variable reach (Zhao et al., 2022).
• Tendon-Driven Anthropomorphic Manipulators: ATDM integrates a hexrotor UAV with a 4-DOF tendon-driven arm, optimizing links via FEA and lattice design for minimal inertia and maximized stiffness. All actuation is base-mounted, minimizing coupling disturbance. Tension-amplification is achieved via multi-wrap pulley designs, increasing joint stiffness and output torques. Adjustable tendon pre-tensioning addresses cable relaxation and maintains interactive compliance (Xu et al., 2024).
• Aerial Continuum Manipulator (AeCoM): Implements a five-segment tendon-driven continuum arm on a quadrotor, with sensor-based kinematic modeling and IMU-based feedback for variable payload adaptation and tendon-slacking prevention. This enables robust manipulation in cluttered and confined environments (Peng et al., 2021).
• Suspended Manipulator Platforms: SAM decouples the main aerial carrier from the manipulator by suspending a powered manipulator platform on a cable, drastically improving safety in obstacle-rich environments. The suspended platform is stabilized by onboard thrusters and winches, supporting a 7-DoF manipulator with fine position and disturbance recovery (Sarkisov et al., 2019).

2. Kinematics, Dynamics, and Modeling

Aerial manipulators are represented as floating-base, multi-body systems with tightly coupled dynamics.

• Continuum Kinematics: For single-segment continuum arms, constant-curvature modeling yields kinematic mappings: configuration vector $\psi = [\theta, \delta]^T$ (bending angle $\theta$, plane $\delta$), tendon length changes via $q_i = r\,\theta\,\cos(\delta + (i-1)\beta)$, and task-space transforms $T_B^G = T_B^1 T_1^E T_E^G$ (Zhao et al., 2022, Zhang et al., 2022). Differential kinematics provide mappings between configuration velocities and tendon or end-effector velocities.
• Stiffness and Compliance: Static equilibrium is captured by $$J_{q\psi}^T\,\tau + J_{x\psi}^T\,w_\text{ext} = \nabla E(\psi)$$. Configuration-space and task-space stiffness are derived from tendon pre-tension and backbone elastic energy, enabling explicit compliance modeling essential for contact tasks and perching. Compliance in continuum manipulators substantially reduces flight disturbances and impact forces (Zhao et al., 2022, Zhang et al., 2022).
• General Floating-Base Modeling: Hybrid Lagrangian-quaternion formulations provide singularity-free models for the full aerial manipulator, with all dynamic quantities computed via a global system Jacobian. This enables real-time computed-torque feedback linearization and supports complex joint topologies (Kremer et al., 2022).
• Whole-Body Planning and Differential Flatness: The coupled UAV+arm system is differentially flat with respect to base position, yaw, and manipulator joint angles; full state and input can be parameterized using flat outputs and derivatives (Wei et al., 2021). This underpins optimization-based motion planning frameworks and allows tractable enforcement of system constraints.

3. Nonlinear Control, Robustness, and Compliance

Control strategies in aerial manipulators address dynamic coupling, external disturbances, and precise interaction.

• Geometric Robust Controllers: Fully actuated bases (e.g., 6-tilt-rotor platforms) enable control at arbitrary 6D poses in SE(3), leveraging geometric robust integral-of-tanh-error control. Robust integral terms and Lipschitz nonlinearities guarantee disturbance rejection during aggressive arm motions and contact (Lee et al., 27 Aug 2025).
• Prescribed Performance Control: Variable-gain extended state observers estimate fast time-varying dynamic coupling. A funnel-based prescribed performance controller ensures error stays strictly within a predefined envelope for both position and attitude tracking—even during aggressive manipulation, such as aerial staff twirling and cart-pulling—yielding sub-centimeter precision (Ji et al., 28 Dec 2025).
• Passivity-Based Compliance Control: Underactuated UAV platforms achieve stable environmental interaction via strictly passive impedance control in end-effector coordinates plus time-domain passivity observation and controller for the UAV fuselage. The interconnection ensures asymptotic stability with minimal model dependence (Kim et al., 2018).
• Proxy-Based Super-Twisting Control: Robust sliding-mode control algorithms mitigate manipulator-induced disturbance. Using proxy variables and super-twisting terms ensures finite-time convergence and substantial chattering elimination, outperforming classical sliding-mode controllers under high coupling (Hua et al., 2023).
• Hybrid Physical Interaction Control: Three-stage controllers (approach, impact feed-forward, constrained force control) orchestrate reliable contact and smooth transition to force-control, leveraging operational-space inertia modeling (Praveen et al., 2020). Such paradigms permit on-the-fly estimation of surface stiffness and coefficient of restitution during contact, integrating inspection and manipulation.

4. Whole-Body Motion Planning and Optimization

Recent frameworks enable integrated trajectory planning for coupled UAV+manipulator systems and guarantee safety, feasibility, and task execution.

• SE(3) × ℝⁿ Planning: High-dimensional optimization leverages whole-body differential flatness, encoding waypoints in SE(3) × ℝ³ (for base pose and end-effector states) (Deng et al., 11 Jan 2025). Variable-geometry approximations of the manipulators’ collision volume (dynamic ellipsoid models) and Safe Flight Corridor (SFC) polyhedra ensure collision avoidance as manipulator configurations change.
• Dynamic Feasibility Constraints: Constraints on actuator, joint limits, instantaneous thrust vector, and kinematic reach, are enforced as smooth penalties within spatio-temporal minimum-control optimization (MINCO). Waypoint constraints for perching or interaction enforce base attitude and end-effector states at critical time points.
• Experimental Validation: Real-world experiments demonstrate flight and motion planning through complex environments (multigate corridors, rapid perching), robust waypoint tracking (<5 cm error), and fast replanning rates (4–10 Hz on NVIDIA Jetson platforms) (Deng et al., 11 Jan 2025).

5. Contact Interaction, Force Estimation, and Sensing Integration

Safe and effective aerial manipulation requires compliant interaction strategies and lightweight force-sensing.

• IMU-Based Force Estimation: Modular continuum manipulators achieve real-time tip force estimation using low-cost IMUs mounted on the end disk. Closed-form mapping from IMU-measured deformation to configuration changes, combined with analytical stiffness matrices, enables estimation of external tip forces without dedicated F/T sensors. Calibration and real-time filtering yield sub-0.1 N accuracy across a range of loads and configurations (Zhang et al., 2022).
• Admittance and Impedance Control: Compliance can be modulated via configuration-dependent stiffness and explicit admittance control (e.g., switching to high-damping, high-stiffness after CoM preload in contact tasks) (Hui et al., 2024). Passive compliance in continuum designs inherently limits impact forces during perching and grasping tasks (Zhao et al., 2022).
• Inspection-on-the-Fly: Hybrid controllers utilize contact phase dynamics not only for manipulation but also for in-situ estimation of environmental properties, such as surface stiffness and elasticity, without additional sensing or inspection phases (Praveen et al., 2020).

6. Integration, Validation, Simulation, and Open Challenges

• Platform Integration: Mechanical interfaces for modular arms, centralized undercarriage mounting, and power/control via onboard microcontrollers are now standard (Zhao et al., 2022, Peng et al., 2021). Base-actuated arms (all motors on UAV body) minimize coupling disturbances and maximize stiffness-to-weight ratio (Xu et al., 2024).
• Simulation and Planning Tools: Modular ROS/Gazebo-based simulation platforms test perception, planning, and control, supporting algorithm validation prior to deployment and enabling extensible workflows for autonomous grasping in cluttered dynamic environments (Quan et al., 2021).
• Experimental Benchmarks: Empirical validation across physical prototypes and simulations confirm advantages in disturbance rejection, precision (millimeter-to-centimeter), compliance, and payload efficiency. Tasks validated include perching, cart-pulling, manipulation in cluttered spaces, aggressive arm motion, contact-based inspection, and non-destructive testing (Zhao et al., 2022, Xu et al., 2024, Ji et al., 28 Dec 2025).
• Open Challenges: Remaining limitations include managing aerodynamic coupling at high speeds, extending frameworks to multi-arm or cooperative UAV teams, real-time online adaptation of compliance and observer thresholds, and complete in-flight control integration for continuum designs (Peng et al., 2021, Xu et al., 2024).

7. Future Directions

• Advanced Sensing: Integration of distributed strain gauges or fiber Bragg gratings for shape and force sensing along continuum backbones (Peng et al., 2021).
• Multi-Quadrotor and Multi-Arm Integration: Coordinated manipulation and collaborative transport with multiple aerial manipulators.
• Nonlinear and Data-Driven Modeling: Cosserat-rod or Koopman operator-based models for continuum arm dynamics, supporting high-speed and high-load tasks.
• Closed-Loop Force Control: Full flight validation of tension-regulated continuum arms and real-time force observers for compliance regulation in contact-rich aerial manipulation.

Aerial manipulators have transitioned from rigid-link arms on multirotors to sophisticated tendon-driven continuum and anthropomorphic designs whose compliance and sensing capabilities are actively exploited for safe, dexterous, and robust interaction. The fusion of high-fidelity modeling, robust nonlinear control, whole-body planning, compliance design, and integrated sensing now enables millimeter-level precision contact tasks and dynamic adaptation to challenging environments, with ongoing research targeting cooperative, multi-arm, and perching-enabled aerial manipulation (Zhao et al., 2022, Xu et al., 2024, Zhang et al., 2022, Peng et al., 2021, Deng et al., 11 Jan 2025, Ji et al., 28 Dec 2025, Deshmukh et al., 24 Dec 2025, Praveen et al., 2020).