Source-linked AI summary
Nonlinear Control of Aerial Manipulator
Sandesh Thapa
TL;DR
An aerial manipulator’s arm changes the coupled system dynamics, making hovering control under bounded disturbances a central problem. The paper develops a full dynamic model and compares Cartesian impedance with adaptive sliding mode control. In the reported comparison, adaptive control avoids the impedance controller’s significant z-position overshoot, while the evaluation remains limited to hovering and perturbations.
Problem
A robotic arm changes a UAV’s center of mass and inertia, making coupled-system hovering control under bounded disturbances a relevant problem.
Source: 1) Control Problem · p. 1Method
The paper develops a full dynamic model for the UAV and arm as one system and implements Cartesian impedance and adaptive sliding mode controllers for hovering.
Source: B. Literature Review · p. 2, A. Methodology · p. 2Results
The adaptive controller has no z-position overshoot, unlike the significant overshoot reported in all three Cartesian impedance cases.
Source: B. Adaptive Sliding Mode Controller · p. 4Takeaways & Limitations
The reported comparison indicates different controller performance across task-space and UAV-position measures, with the evaluation focused on hovering and perturbations.
Source: D. Simulation Results · p. 2, B. Adaptive Sliding Mode Controller · p. 4The controller evaluation is limited to hovering and perturbations, with no experimental data or tracking results reported.
Source: D. Simulation Results · p. 2, A. Methodology · p. 2
Abstract
from arXiv · showhide
This paper presents two controllers for a quadrotor equipped with a two-DOF robot manipulator that physically interacts with the environment. Cartesian impedance and adaptive sliding mode controllers are implemented and compared for a simple hovering task in the presence of some bounded disturbance. A full dynamic model is developed, and the two above controllers are implemented with the same initial conditions and desired position.
I. SUMMARY
The paper motivates integrated aerial-manipulator control by the arm’s effects on system dynamics, and describes Cartesian impedance control in end-effector coordinates. Its control law transforms joint-space dynamics and specifies a closed-loop error model driven by external forces.
A. Cartesian Impedance Control
- Cartesian impedance control relates end-effector motion in Cartesian coordinates to generalized external forces and moments.The cited approach modifies classical impedance control because measuring end-effector force is infeasible in aerial robotics.Source: A. Cartesian Impedance Control · p. 1
1) Control Problem:
- Coupling from the arm changes the system’s center of mass and moment of inertia, motivating a hovering-control problem with bounded disturbances.The paper notes potential uses including dexterous manipulation, rescue, construction, and power-line inspection.Source: 1) Control Problem · p. 1
2) Control Law:
- The system Jacobian transforms joint-space dynamics into manipulator end-effector Cartesian coordinates, with generalized torque related to the four motor thrusts.The passage introduces the Cartesian dynamics and the control law using symmetric positive-definite gain matrices K_P and K_D.Source: 2) Control Law · p. 1
- The stated closed-loop error dynamics include inertia, Coriolis terms, gain-weighted error and external force.The error dynamics are given as M_x ẍ̃ + (C_x + K_D) ẋ̃ + K_P x̃ = f_ext.Source: 2) Control Law · p. 1
3) Stability Analysis:
The stability analysis introduces a positive-definite Lyapunov candidate and differentiates it, but its guarantees depend on the motion setting and Jacobian condition.
3) Stability Analysis:
- A positive-definite Lyapunov candidate is defined, and its time derivative is taken before substituting the closed-loop dynamics.The supplied passages identify the candidate and derivative steps but do not provide a complete readable expression.Source: 3) Stability Analysis · p. 1
- For constrained motion, stability is ensured only in unconstrained Cartesian directions; for free motion, stability is ensured, while asymptotic stability is tied to an invertible Jacobian.The passage says the time-varying Cartesian inertia matrix prevents use of the invariance principle in the free-motion case.Source: 3) Stability Analysis · p. 1
A. Methodology
The paper models the UAV and arm as a coupled system and applies model-based Cartesian impedance control, while the reviewed approach and evaluation are described as limited in scope and detail.
A. Methodology
- The Cartesian impedance model requires exact UAV and manipulator dynamic parameters, focuses on hovering, and reports no tracking results.Some impedance parameters also depend on the system dynamics rather than being freely specifiable.Source: A. Methodology · p. 2
B. Literature Review
- Earlier reports controlled the UAV and arm separately, whereas this paper’s motivation is a full dynamic model of the combined system.The passage links this motivation to coupling effects and changes in center of mass and inertia.Source: B. Literature Review · p. 2
C. Dynamic Model/Mathematics
- The dynamic-model discussion flags potentially invalid assumptions, possible Jacobian non-invertibility, complicated C-matrix computation, and an underdeveloped stability argument.It questions whether a ground-manipulator stability argument establishes global asymptotic stability for the UAV-manipulator system.Source: C. Dynamic Model/Mathematics · p. 2
- The derivation omits a Jacobian formula and does not clearly distinguish manipulator velocity and angular Jacobians.The reviewer reports resolving the distinction using another paper.Source: C. Dynamic Model/Mathematics · p. 2
D. Simulation Results
- Simulation results are said to agree with theoretical analysis, but torque and generalized-force details, experimental data, and hovering tracking-error plots are absent.The evaluation is limited to hovering and perturbations, which the passage describes as insufficiently comprehensive for assessing practical aspects.Source: D. Simulation Results · p. 2
E. Suggestions
The suggestions focus on improving the presentation of the model and figures, strengthening stability analysis, and adding experimental evidence.
E. Suggestions
- Use clearer notation when deriving the complete dynamics.Source: E. Suggestions · p. 2
- Provide clearer equations for the system and manipulator Jacobians.Source: E. Suggestions · p. 2
- Label the UAV inertial and body frames more clearly in the figure.Source: E. Suggestions · p. 2
- Improve the stability analysis.Source: E. Suggestions · p. 2
- Include experimental data.Source: E. Suggestions · p. 2
F. Sliding Mode Control
The adaptive sliding mode controller defines an error and sliding surface, then uses estimated dynamics and positive gain matrices in its control law. Its stated convergence guarantees depend on the disturbance condition and a slowly varying disturbance estimate.
F. Sliding Mode Control
- The adaptive sliding mode controller is presented as a comparison controller from.Source: F. Sliding Mode Control · p. 2
- The controller is introduced by defining an error and sliding surface, with a Lyapunov candidate used for stability analysis.The passage also introduces taking the candidate's time derivative.Source: F. Sliding Mode Control · p. 2
- The control law uses estimated dynamics, desired motion, the sliding surface, and positive gain matrices.The estimates represent system matrices, while A and K are positive gain matrices.Source: F. Sliding Mode Control · p. 2
- With τext = 0, the sliding variable tends to zero and the error converges exponentially; with τext ≠ 0, both are bounded.The stability statement assumes the disturbance changes slowly compared with the adaptation rate.Source: F. Sliding Mode Control · p. 2
III. SIMULATION EXPERIMENT
The simulation uses an AscTec Pelican and a two-DOF robotic arm, with a specified desired position and bounded forces modeling a windy situation. The cases are compared with an adaptive sliding mode controller.
III. SIMULATION EXPERIMENT
- The simulation models an AscTec Pelican with mass 2 kg and inertia diag(1.24, 1.24, 2.48).Source: §III · p. 3
- The modeled manipulator is a 2-DOF robotic arm whose first link is 15 cm; the passage reports masses of 0.05 kg for each of its two links.Source: §III · p. 3
- The desired position is qd = [0, 0, 6, 0, 0, 0, 0, 0, 0.5, 0.5], with 1 N and 0.5 N forces modeling a windy situation.The 1 N force is sinusoidal at π/4 rad/s along xb; the 0.5 N force acts along the end-effector x-direction for 20 s, and cases are compared with an adaptive sliding mode controller.Source: §III · p. 3
A. Cartesian Impedance Controller
The section identifies time-history simulations for rigid, compliant, and compliant-rigid cases, and gives gain matrices for the rigid UAV and compliant manipulator.
A. Cartesian Impedance Controller
- The rigid case is identified as a time-history simulation.Source: Figure 1 · p. 3
- The compliant case is identified as a time-history simulation.Source: Figure 2 · p. 3
- For the rigid UAV and compliant manipulator, KP = diag(10I6, 100I2) and KD = diag(3I6, 50I2).Source: A. Cartesian Impedance Controller · p. 3
- The compliant-rigid case is identified as a time-history simulation.Source: Figure 3 · p. 3
B. Adaptive Sliding Mode Controller:
The adaptive sliding mode controller keeps errors bounded under the simulated conditions, while comparisons with Cartesian impedance show different strengths in UAV motion and end-effector accuracy.
B. Adaptive Sliding Mode Controller:
- The reported adaptive-controller design parameters are K = 8I8×8, Λ = 2I8×8, and A = 10I8×8.Source: B. Adaptive Sliding Mode Controller · p. 3
- The adaptive controller has no z-position overshoot and lower error than impedance except at the end-effector, where its position error is higher.The z-position comparison favors adaptive control, while Figure 9 shows significantly less error except for end-effector position.Source: Figure 5 · p. 3, Figure 9 · p. 4, B. Adaptive Sliding Mode Controller · p. 4
- The z-position converges from 0 to 6 in less than 5 seconds in all three cases, and adaptive-controller errors remain bounded despite nonzero orientation error under disturbance.The paper attributes the orientation error not reaching zero to external disturbance.Source: Figure 4 · p. 3, Figure 5 · p. 3, B. Adaptive Sliding Mode Controller · p. 4
IV. CONCLUSION
The conclusion compares Cartesian impedance and adaptive sliding mode control for the coupled aerial-manipulator system. It identifies a possible combination of the two controllers as a solution.
IV. CONCLUSION
- Cartesian impedance performs better in task space, while adaptive control performs better for UAV position and orientation.Source: §IV · pp. 4–5
- The coupled dynamics of an aerial manipulator make its control problem challenging.Source: §IV · p. 4
- The conclusion proposes combining the controllers in an adaptive Cartesian impedance controller.Source: §IV · p. 5