Control of Aerial Manipulator

Control of Aerial Manipulator
MAE 5463 Nonlinear Systems
Sandesh Thapa
Department of Mechanical and Aerospace Engineering

Oklahoma State University
December 17, 2019
Sandesh Thapa (MAE, OSU) Control of Aerial Manipulator December 17, 2019 1 / 22
Outline
Introduction
Dynamics of the Combined System
Control Problem
Control Approach
I Cartesian Impedance Control
I Adaptive Sliding Mode Controller
I Adaptive Cartesian Impedance Control
Stability Analysis
Simulation Results
Conclusion and Future Work
Introduction
Aerial Manipulator: Rotorcraft Equipped with a robotic arm

Current research in Aerial Manipulator:
I Installing a gripper at the bottom of UAV to hold a payload
I Suspending payload with cables
I Autonomous navigation
Use:
I Inspection and maintenance of power lines
I Rescue operations
I Construction in inaccessible sites
Dynamics of Combined System
Based on Euler-Lagrangian formulation

M(q)q̈ + C(q, q̇)q̇ + G(q) = τ + τext (1)
Where, q is the generalized joints vector of 6+ n components. In
this case, the arm has 2 DOF and n = 2.
M is the inertia matrix, C is the centripetal/Coriolis matrix and G is
the gravity.
Figure 1: UAV/Arm system with corresponding reference frame [1].

Control Problem
Complicated coupled dynamics

Displacement of center of mass
Variation of mass distribution, moment of inertia
Cartesian Impedance Control
Control Law:
ẋ = Ja q̇ (2)
ẍ = Ja q̈ + J˙a q̇ (3)
x̃ = xd − x (4)
τ = JaT f (5)
From (1),(2),(3),(3),(4) & (5),the joint space dynamics can be
transformed into manipulator end-effector Cartesian coordinates(x).
Here, Ja is the Jacobian of the system. τ is the generalized torque
related to four motor thrust and joint actuator.
Mx ẍ + Cx (x, ẋ)ẋ + Gx (x) = f + fext (6)

where, Mx = Ja−T Mq Ja−1 , Cx = Ja−T (Cq − Mq Ja−1 J˙a )Ja−1 ,
Gx = Ja−T Gq ,f = Ja−T τ , fe xt = Ja−T τext
Following control law is defined where KP and KD are nq × nq

symmetric and positive definite matrices.
τ = G + JaT (Mx x¨d + Cx x˙d + KD x̃˙ + KP ẋ) (7)
Closed loop dynamics of the system is:
Mx x̃¨ + (Cx + KD )x̃˙ + KP x̃ = fext (8)
Stability Analysis:
Positive definite candidate Lypanuv function
1 ˙T 1
V (x, t) = x̃ Mx x̃˙ + x̃ T KP x̃ (9)
2 2
Time derivative of V:
T
V̇ = −x̃˙ T KD x̃˙ + x̃˙ fext (10)
If fext = 0 for free motion case, Using Theorem 8.5, from Khalil,we
can conclude that x̃˙ → 0 as t → ∞
In case of constraint motion , fext 6= 0, only stability can be
ensured.
Simulation Results:
ASCTEC PELLICAN, mb = 2 kg, Ib = diag(1.24, 1.24, 2.48)
2-DOF robotic arm, l1 =15 cm l2 5 cm with 3 revolute joints,
m1 = 0.05 kg, I1 = 0.0019 m2 kg , m2 = 0.05 kg, I1 = 0.0011 m2 kg
Focus on hovering control
T
initial conditions, q = 0, 0, 0, 0, 0, 0, 0, 0, π2

I
I desired position, qd = [0, 0, 6, 0, 0, 0, 0, 0, 0.5, 0.5]
A force of 1 N is modelled through of sine wave of π/4 rad/s which
acts along the xb axis and another force of 0.5 N magnitude acts
along the x-direction of end-effector of 20 s to model the windy
situation.
Simulation Results: KP = 80I8 ,, KD = 16I8
Figure 2: Time histories simulation for Rigid Case.
Simulation Results: KP = 5I8 ,, KD = 2I8
Figure 3: Time histories simulation for Compliant Case.
Simulation Results:KP = diag(10I6 , 100I2 ), KD = diag(3I6 , 50I2 )
Figure 4: Time histories simulation for Compliant Rigid Case.
Adaptive Sliding Mode Control
Control Law:
e = q − qd (11)
s = ė − Λe (12)
qr = q̇ − s = q˙d − Λe (13)
To design the controller, we define error e and sliding surface s. Now
the control law becomes:
ˆ − As − Ksgn(s)
τ = M̂ q¨r + Ĉ q˙r + ∆ (14)
here, M̂, Ĉ, Ĝ represents estimation of each matrix. A and K are
positive gain matrices.
ˆ − Ksgn(s)
M ṡ + Cs + Λs = −∆ + ∆ (15)
where,
∆ = M̃ q¨r − C̃ q˙r − G̃ (16)
Stability Analysis: Let a Lyapunov candidate function be
1 T 1 ˜T ˜
V = s Ms + ∆ ∆>0 (17)
2 2
It time derivative is :
1 T ˜ ˜ T ˜˙ T T
V̇ = s (−Λs + ∆−Ksgn(s)+τext )+ ∆ ∆ = −s Λs −S (Ksgn(s)−τext ) ≤ 0 (18)
2
Adaptation law:
ˆ˙ − ∆
∆ ˙ = −s (19)
with the assumption that ∆ changes very slow compared with the adaptation rate,
we get
ˆ˙ = −s
∆ (20)
Free motion: If we assume that τext = 0, from (19) and (20) , we can conclude that
s → 0 as t → 0, so e → 0 exponentially.
If τext 6= 0, we can conclude that s is bounded , so =⇒ e is bounded.
Simulation Results:
Design Parameters: K = 8I8×8 , Λ = 2I8×8 , A = 10I8×8
Figure 5: Time histories simulation for adaptive sliding mode controller.

Simulation Results:Position History
Figure 6: Position history of the quadrotor.
Simulation Results:Attitude History
Figure 7: Attitude history of the quadrotor.
Simulation Results:Torque Plots
(a) Torque1 History. (b) Torque2 History

Figure 8: Torque History for both links
Conclusion and Future Work
Coupled dynamics of aerial manipulator makes the control

problem challenging
Computation of C matrix
Can implement a hierarchical control separately for both arm and
the rotor.
Future work on other applicable control strategies and platform for
testing
References
[1] V. Lippiello and F. Ruggiero,Cartesian impedance control of a UAV
with a robotic arm. In 10th International IFAC Symposium on Robot
Control, Dubrovnik, Croatia, Sep 2012
[2] Kim S. et al. Aerial Manipulation Using a Quadrotor with a Two DOF
Robotic Arm, 2013 IEEE/RSJ International Conference on Intelligent
Robots and Systems (IROS 2013) November 3-7, 2013, Tokyo, Japan.
Questions?

Control of Aerial Manipulator

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Control of Aerial Manipulator

Uploaded by

Copyright:

Available Formats

Control of Aerial Manipulator

MAE 5463 Nonlinear Systems

Department of Mechanical and Aerospace Engineering

December 17, 2019

Aerial Manipulator: Rotorcraft Equipped with a robotic arm

Based on Euler-Lagrangian formulation

Figure 1: UAV/Arm system with corresponding reference frame [1].

Complicated coupled dynamics

Mx ẍ + Cx (x, ẋ)ẋ + Gx (x) = f + fext (6)

Following control law is defined where KP and KD are nq × nq

τ = G + JaT (Mx x¨d + Cx x˙d + KD x̃˙ + KP ẋ) (7)

Closed loop dynamics of the system is:

Mx x̃¨ + (Cx + KD )x̃˙ + KP x̃ = fext (8)

Figure 2: Time histories simulation for Rigid Case.

Figure 3: Time histories simulation for Compliant Case.

Figure 4: Time histories simulation for Compliant Rigid Case.

Figure 5: Time histories simulation for adaptive sliding mode controller.

Figure 6: Position history of the quadrotor.

Figure 7: Attitude history of the quadrotor.

(a) Torque1 History. (b) Torque2 History

Coupled dynamics of aerial manipulator makes the control

You might also like