Aerial manipulation (AM) expands UAV capabilities beyond passive observation to contact-based operations at high altitudes and in otherwise inaccessible environments. Although recent advances show promise, most AM systems are developed in controlled settings that overlook key aerodynamic effects. Simplified thrust models are often insufficient to capture the nonlinear wind disturbances and proximity-induced flow variations present in real-world environments near infrastructure, while high-fidelity CFD methods remain impractical for real-time use. Learning-based models are computationally efficient at inference, but often struggle to generalize to unseen condition. This paper combines both approaches by integrating a physics-based blade-element model with a learning-based residual force estimator, along with a rotor-speed allocation strategy for disturbance compensation, resulting in a unified control framework. The blade-element model computes per-rotor aerodynamic forces under wind and provides a refined feedforward disturbance estimate. A learning-based estimator then predicts the residual forces not captured by the model, enabling compensation for unmodeled aerodynamic effects. An online adaptation mechanism further updates the residual-force prediction and rotor-speed allocation jointly to reduce the mismatch between desired and realized thrust. We evaluate this framework in both free-flight and wall-contact tracking tasks in a simulated near-wall wind environment. Results demonstrate improved disturbance estimation and trajectory-tracking accuracy over conventional approaches, enabling robust wall-contact execution under challenging aerodynamic conditions.
Figures & tables
M∈R6×6
Generalized inertia matrix in the body frame
C∈R6×6
Coriolis and centrifugal matrix in the rigid-body model
V∈R6
Body-frame twist: linear and angular velocity [v⊤ω⊤]⊤
ω∈R3
Angular velocity of the vehicle in the body frame
v∈R3
Linear velocity of the vehicle in the body frame
R∈SO(3)
Rotation matrix from inertial frame to body frame
m∈R
Mass of the vehicle
Table 1
Figure 1: Rotor thrust and drag characteristics of the APC 8 × 6 propeller [ 2 ] under varying pitch angles β at a wind speed of 5 m/s, evaluated across different rotational speeds.
Component
Value / Description
Shared network architecture
26 → 64 → 32 (ReLU)
Each head architecture
32 → 32 → 3 (ReLU)
Batch size ( B )
256
Learning rate
0.0005
Table 2: Hyperparameters of the residual network
Parameter
Specification
Horizontal wind set
{−3,−1,0} m/s
Vertical wind set
{−5,−3,0,3,5} m/s
Trajectory profile
Randomized 3D trajectory in open space
Flight duration per trial
60 s
Table 3: Training data generation configuration
Parameter
Value / Description
Total mass m
2.1395 kg (drone + battery)
Rotor horizontal distance dh
0.28 m (center to rotor center)
Rotor vertical offset dv
0.095 m (above CoG)
Inertia matrix J
diag(0.0820, 0.0845, 0.1377) kg·m 2
Number of rotors
4
Rotor configuration
Symmetric cross, 90° apart
Table 4: Quadrotor simulation parameters used in the proximity-to-wall wind disturbance scenario.
Figure 2: Simulated 3D velocity field in the x – y plane at z=0 for an oblique freestream wind
Wall Distance (m)
Fx (N)
Fz (N)
Ty (Nm)
0.5
-0.396
0.429
-0.006
1.0
-0.859
0.763
0.019
5.0
-2.778
1.249
0.110
Table 5: Disturbance forces and torques at different wall distances during hover under a −10m/s wind applied in the x -direction. Due to the relative positions of the rotors, lateral and non-uniform axial disturbance forces can induce a net disturbance torque on the quadrotor, even though BEMT does not account for lateral moments generated by individual rotors.
Parameter
Value
Rod length
0.45 m
Tip position in body frame
[−0.45,0,0]⊤ m
Tip sponge radius
0.1 m
Linear spring constant
10 N/m
Coulomb friction coefficient
0.4
Table 6: End-effector geometry and contact model parameters
Figure 3: Noisified IMU acceleration and angular velocity measurements during a steady hover
Figure 4: Simulation reference trajectories.
Parameter
Learned Value
Ground Truth
cℓ,1
5.460
5.300
cℓ,2
2.060
1.700
cd
3.007
1.800
α0 (rad)
0.352
0.360
Table 7: Comparison of learned versus ground-truth aerodynamic coefficients.
Figure 5: Rotor thrust and drag characteristics of the APC 8×6 propeller comparison among simulation groundtruth, parameter-learned BEMT model and zero-wind approximation.
Parameter
Specification
Horizontal wind set
{−5,−2,0} m/s
Vertical wind set
{−10,−4,−1,1,4,10} m/s
Trajectory profile
Figure-eight trajectory near wall
Flight duration per trial
40 s
Table 8: Test data generation configuration
Figure 6: Comparison of training and evaluation losses for the BEMT-residual model and the zero-wind-residual model. Incorporating the BEMT prior reduces the overall loss on both datasets, indicating improved alignment between the predicted and measured aerodynamic forces.
Autonomous Aerial Manipulators (AAMs) are inherently coupled, nonlinear systems that exhibit nonstationary and multiscale residual dynamics, particularly during manipulator reconfiguration and abrupt payload variations. Conventional analytical dynamic models rely on fixed parametric structures, while static data-driven model assume stationary dynamics and degrade under configuration changes and payload variations. Moreover, existing learning architectures do not explicitly factorize cross-variable coupling and multi-scale temporal effects, conflating instantaneous inertial dynamics with long-horizon regime evolution. We propose a predictive-adaptive framework for real-time residual modeling and compensation in AAMs. The core of this framework is the Factorized Dynamics Transformer (FDT), which treats physical variables as independent tokens. This design enables explicit cross-variable attention while structurally separating short-horizon inertial dependencies from long-horizon aerodynamic effects. To address deployment-time distribution shifts, a Latent Residual Adapter (LRA) performs rapid linear adaptation in the latent space via Recursive Least Squares, preserving the offline nonlinear representation without prohibitive computational overhead. The adapted residual forecast is directly integrated into a residual-compensated adaptive controller. Real-world experiments on an aerial manipulator subjected to unseen payloads demonstrate higher prediction fidelity, accelerated disturbance attenuation, and superior closed-loop tracking precision compared to state-of-the-art learning baselines, all while maintaining strict real-time feasibility.
Small multirotor aircraft are increasingly tasked with operations in the atmospheric boundary layer, where turbulent winds comparable to the vehicle's airspeed degrade trajectory tracking and can defeat conventional feedback control. This work illustrates a two-stage learning pipeline that first estimates the local wind from onboard kinematics and dynamics and then exploits that estimate inside a reinforcement learning (RL) flight controller. The wind estimator, an attention-augmented gated recurrent network trained on thousands of simulated flights through von Karman turbulence with power-law shear and veer, recovers the horizontal wind vector with a per-flight root-mean-square error of 0.40 m/s and a direction error of 3.2 degrees on unseen wind regimes, an accuracy near the floor imposed by unresolved turbulence, and generalizes to vertical ascent profiles with a skill score of 0.861 over a constant-wind reference. A proximal policy optimization controller receiving the frozen estimator's output reduces horizontal trajectory tracking error by 48% relative to a wind-blind proportional-derivative baseline across mean winds of 4 m/s to 12 m/s, winning on 100% of evaluation episodes. A three-way ablation decomposes this improvement into a kinematic component, available without wind information, and a wind-perception component; the perception share rises with wind speed, from small in light winds toward roughly half the total benefit in strong winds, consistent with the quadratic scaling of aerodynamic drag. The controller degrades gracefully on out-of-distribution winds of 13 m/s to 15 m/s, where the baseline fails catastrophically.
Aerial physical interaction represents a promising direction for next-generation unmanned aerial vehicles (UAVs), but it requires an aerial platform that can exert contact forces while maintaining stable flight. For close-proximity tasks, this translates into three coupled design requirements: multidimensional wrench generation for stable contact, compactness for maneuverability and safety in confined spaces, and reduced lateral airflow toward the target when generating horizontal force. This article presents FLOAT Drone, a fully actuated coaxial UAV with servo-driven control surfaces for close-proximity physical interaction. The coaxial dual-rotor layout provides a compact propulsion layout, while the control surfaces, immersed in the rotor downwash, generate lateral forces and moments for 6-DoF wrench generation. A force-matched computational fluid dynamics (CFD) comparison with a tilted-rotor alternative quantifies the reduction in target-facing lateral airflow. To account for nonlinear rotor--control-surface coupling in the rotor wake, a high-fidelity polynomial aerodynamic wrench model is identified from precision force measurements and embedded in a constrained nonlinear allocator for real-time wrench tracking. Comparative flight and interaction experiments show that the proposed framework improves control accuracy over linear allocation baselines, rejects ground-effect and payload disturbances, and enables close-proximity drawer push--pull manipulation through a 2cm handle clearance.
Junxiao Lin, Kehan Zhou, Shuhang Ji +4
Institute of Cyber-Systems and Control, College of Control Science and Engineering, Zhejiang University, Hangzhou 310027, China · Differential Robotics Technology Company, Hangzhou 311100, China