Professional Documents
Culture Documents
JGCD 2021 TC
JGCD 2021 TC
Engineering Notes
Cooperative Transportation of a Flexible method for the cooperative manipulation of a cable-suspended pay-
load using multiple quadrotors to drive the payload to track the
Payload Using Two Quadrotors desired trajectory [2]. Masone et al. considered the cooperative trans-
portation system of a cable-suspended payload as a reconfigurable
cable-driven parallel robot and designed a controller based on the
Ti Chen∗ internal motion between payload and quadrotors [11]. Another
Nanjing University of Aeronautics and Astronautics, 210016 method is to realize the transportation of cable-suspended payload
Nanjing, People’s Republic of China based on the formation flying of a group of quadrotors. For instance,
Jinjun Shan† Qian and Liu designed a hierarchical controller based on a velocity
York University, Toronto, Ontario M3J 1P3, Canada stabilization control to stabilize the payload and drive the quadrotors
along a given trajectory [12]. Tagliabue et al. presented an admittance
and control method under a master–slave architecture for quadrotors
Hugh H. T. Liu‡ carrying a heavy payload via spherical joints or short cables, which
University of Toronto, Toronto, Ontario M3H 5T6, Canada implies that the quadrotor attitude motions are decoupled from the
https://doi.org/10.2514/1.G005914
payload motion [10]. Furthermore, some researchers have considered
Downloaded by Hugh Liu on July 8, 2021 | http://arc.aiaa.org | DOI: 10.2514/1.G005914
aerial transportation of a flexible payload. This study attempts to rq2 X 2 ; Y 2 ; Z2 T X 1 ; Y 1 ; Z1 T R1 0; l; wl; tT , where rp
present a solution to the control problem for the aerial transportation is the position vector of point P, X 1 ; Y 1 ; Z1 T is the coordinate of
of a flexible payload, i.e., a flexible beam, based on the dynamic model rq1 in the frame OXYZ, and wl; t represents the lateral deformation
of the whole system. Numerical and experimental results are presented of the point on the beam under quadrotor 2.
to validate the feasibility of such a transportation system. The main Let rc 1∕2rq1 rq2 Xc ; Y c ; Zc T . Essentially, rc
contributions of this paper can be summarized as follows. describes the position of the mass center of the entire system with
1) Compared with the studies on multiple quadrotors grasping a the undeformed beam. Therefore, rq1 , rp , and rq2 can be expressed as
payload rigidly in [20,21], the payload of concern is flexible, which
rq1 rc − 1∕2R1 0; l; wl; tT , rp rc − 1∕2R1 0; l; wl; tT
results in a time-varying relative pose between quadrotors.
2) Different from the works on flexible payload transportation in R1 0; y; wy; tT , and rq2 rc 1∕2R1 0;l;wl;tT , respectively.
[22,23], the assumed modes method is adopted to discretize the Let ω1 ω1;1 ; ω1;2 ; ω1;3 T and ω2 ω2;1 ; ω2;2 ; ω2;3 T be the
beam, then a nonlinear dynamics model is derived to describe the angular velocities of two quadrotors expressed in their body frames.
motion of the entire transportation system. A fictitious output is Based on the configuration shown in Fig. 1, one can find that
defined to indicate feasible trajectories for such a transportation task. R2 R1 R12 , where
LQR is used to design state feedback controllers with integral terms 2 3
based on the linearized model. 1 0 0
6 ∂w0;t 7
6 0 cos ∂wl;t ∂w0;t
∂y − ∂y − sin ∂wl;t
∂y − ∂y 7
R12 6
7
(1)
4 5
II. System Model 0 sin ∂wl;t ∂w0;t
cos ∂wl;t ∂w0;t
∂y − ∂y ∂y − ∂y
A. Notations
In ∈ Rn×n and 0m×n ∈ Rm×n represent the identity matrix and
zero matrix, respectively. The 3D rotation group SO3 Taking the time derivative of both sides of R2 R1 R12 yields
R2 ω×2 R1 ω×1 R12 R1 R12 ω×12 , i.e., ω2 RT12 ω×1 R12 ∨ ω12 ,
Downloaded by Hugh Liu on July 8, 2021 | http://arc.aiaa.org | DOI: 10.2514/1.G005914
chλj l cosλj l Note that in Eq. (5), the time derivatives of Euler angles, i.e., ϕ1 , θ1 ,
c1 ψ 1 , are included. Considering the fact that the angular velocity ωi is
shλj l sinλj l
usually measurable, let η X_ c ; Y_ c ; Z_ c ; ωT1 ; p_ 1 T . It is clear that
λ1 1.875∕l, λ2 4.694∕l, λ3 7.855∕l, · · · , and the mappings η A 1 q, _ where A 1 diagfI3 ; A1 ; 1g is a block diagonal matrix.
ch: R → R and sh: R → R are defined as chy ey e−y ∕2 and Note that A1 is defined at the end of the paragraph after Eq. (3).
shy ey − e−y ∕2, respectively; pj t represents the vibration Equation (5) can be rewritten as
amplitude of the corresponding mode shape. Hence, ∂2 wy; t∕∂y2
can be written as M_η Cη Kq
gQ (6)
Xn
d2 where M A −T −1 −T −1 −T
1 MA 1 , C A 1 CA 1 , g A 1 g, and
Φj ypj t −T
Q A1 Q.
j1
dy2
C. Model Linearization and Equilibrium at Hover
Without loss of generality, only the fundamental frequency is consid-
ered, i.e., n 1 in this study, and p1 t can be called the fundamental To derive the hover state of the transportation system, the follow-
modal coordinate. Based on the 3-2-1 Euler angles, the rotation matrix ing assumptions are made.
Ri can be written as Assumption 1: The deformation of the beam due to gravity is
negligible.
2 3 Assumption 2: Both quadrotors work in near-hover states; hence, it
cψ i cθi −sψ i cϕi cψ i sθi sϕi sψ i sϕi cψ i sθi cϕi
is assumed that ϕi ≪ 1, θi ≪ 1 and ψ i ≪ 1.
Ri 4 sψ i cθi cψ i cϕi sψ i sθi sϕi −cψ i sϕi sψ i sθi cϕi 5 (4)
Assumption 3: The flexible payload is subject to small deforma-
−sθi cθi sϕi cθi cϕi tion, i.e., p1 ≪ 1.
Downloaded by Hugh Liu on July 8, 2021 | http://arc.aiaa.org | DOI: 10.2514/1.G005914
ρlg dΦ1
M
q C q_ Kq
g Q Q0 2mq g ρlgθ1 ; − mq g p
(5) 2 dy
yl 1
where M,
C,
and 1 1
− 2mq g ρlgϕ1 ; f1 f2 ; τ1 τ4 lf2 − lf1 ; τ2
2 2
Z l 2
T
d Φ1 y 2 1 dΦ1
1
K diag 01×6 ;
dy τ5 ; τ 3 τ6 ; f2 τ − f
0 dy2 2 dy
yl 4 2 1
are the inertia matrix, the Coriolis and centrifugal matrix, and the The dynamic model of the transportation system of concern can be
stiffness matrix. g 2mg geT3 ρlgeT3 ; −ρgcθ1 sϕ1 ∫ l0 Φ1 y dy − linearized as Eq. (7) and
l∕2p1 ; −ρgsθ1 cϕ1 ∫ l0 Φ1 y dy − l∕2p1 ; 0; ρgcθ1 cϕ1 ×
∫ l0 Φ1 y dyT . M0 η_ Kq B1 f (10)
4 Article in Advance / ENGINEERING NOTES
−2mq g − ρlg 0 0 − mq g ρlg from last step, τ4 , τ6 , and −1∕2lf2 1∕2lf1 τ5 can be
2 dy yl
obtained. Similarly, from Eq. (14), one can reach
1 dΦ1
1
It should be noted that in Eq. (10), f ∈ R8 and B1 ∈ R7×8 are defined f2 τ − f
as 2 dx
xl 5 2 1
h iT Since
f f1 ; τ1 ; τ2 ; τ3 ; f2 ; τ4 ; τ5 ; τ6
2 3 dΦ1
1
0 0 0 0 0 0 0 0 ≠
6 7 dx
xl l
6 0 0 0 0 0 0 0 07
6 7
6 7 τ5 , f1 , and f2 can be computed from the values of
6 1 0 0 0 1 0 0 07
6 7 −1∕2lf2 1∕2lf1 τ5 ,
6 7
Downloaded by Hugh Liu on July 8, 2021 | http://arc.aiaa.org | DOI: 10.2514/1.G005914
B1 6
6 −l∕2 1 0 0 l∕2 1 0 077
6 7 1 dΦ1
1
6 0 0 1 0 0 0 1 07 f2 τ − f
6
6
7
7 2 dx
xl 5 2 1
6 0 0 0 1 0 0 0 17
4
5
dΦ1
and f1 f2 .
−1∕2 0 0 0 1∕2 dx xl 0 0
So far, all states and inputs of the transportation system have
been determined by Y, Y, _ and Y. This completes the proof of
where f1 f1 − mq g − ρgl∕2 and f2 f2 − mq g − ρgl∕2. Lemma 1. □
Essentially, the terms mq g ρgl∕2 in f1 and f2 aim to compensate
for the quadrotors and payload gravities. Let χ qT ; ηT T . Equa-
III. Control System Design
tion (10) can be written as
For a quadrotor or the cooperative transportation of a rigid payload
χ_ Aχ Bf (11) in [20], the popular hierarchical control methods consist of an inner
loop and an outer loop, where the outer loop aims to realize the
position control by providing the total thrust and the reference roll
07×7 I7 07×8
where A and B . and pitch angles and the inner loop generates the control torques to
−M−1
0 K 07×7 M−1
0 B1 drive the system to track the desired attitude. However, in this study,
as shown in Eqs. (12–14), the inner-outer loop control methods are
D. Fictitious Outputs not applicable anymore because one cannot extract the reference roll
An open question for such a transportation system is what varia- and pitch angles from the position control due to the coupling
bles we can control directly; i.e., how to define feasible trajectories. between the quadrotor motion and payload vibration. The funda-
This section aims to discuss this in the near-hover configuration; mental reason is that the payload deformation changes the relative
i.e., Eq. (10) is used. Here the fictitious outputs mean the variables position and orientation of the quadrotors. Instead, this section aims
that can be used to explicitly express all states and inputs based on the to design a feedback controller based on the linearized model (11)
fictitious outputs and a finite number of its derivatives. The following using the LQR algorithm for the transportation system.
lemma gives the fictitious output of the transportation system in
this study. A. Measurement of Modal Coordinate and Velocity
Lemma 1: Y rC ; ψ 1 ; p1 ; τ 1 ∈ R3 × R × R × R3 is the set of This section presents a method based on the kinematic relationship
fictitious outputs for the system governed by Eq. (10) in the near- between two quadrotors to obtain the modal coordinate and velocity
hover configuration. of the flexible beam. It is assumed that the states of both quadrotors
Proof: Equation (10) can be rewritten as can be measured or estimated. The first three elements of the gener-
alized coordinate vector represent the position vector of the middle
ρlg dΦ1
1
f2 τ − f (14) In the derivation of the linearized model (11), the gravity of the
2 dx
xl 5 2 1 beam is neglected according to Assumption 1. As shown in Sec. II.C,
Article in Advance / ENGINEERING NOTES 5
the equilibrium configuration agrees with the undeformed situation Table 1 Parameters of quadrotors and
shown in Fig. 1 without the consideration of the deformation of the payload
beam caused by gravity. However, for a practical payload, its gravity Parameter Value Unit
will result in a stable configuration with nonzero beam deformation
mq 1.121 kg
only with the supports at tips; i.e., there would exist a steady-state error
Jq diagf0.01; 0.0082; 0.0148g kg ⋅ m2
if the classical feedback controller based on Eq. (11) is applied to
control the nonlinear system in Eq. (6). Furthermore, for the trans- g 9.81 m∕s2
portation system in this study, there may exist some external disturb- ρ 0.138 kg/m
ances and uncertainties. For example, the uncertain payload or l 0.83 m
quadrotor mass may cause that the compensation for gravity is not EI 6.46 N ⋅ m2
exactly equal to the real quadrotor and payload gravities; i.e., some
constant forces in the vertical direction may act on the closed-loop
system. These constant forces will also cause some steady-state errors
that may prevent the payload from moving to the desired position and P is the solution to the Riccati equation A T P PA −
−1
attitude. One method to provide zero steady-state error is to include an PB
R P Q 0.
integral term. Considering that the control of the payload’s position and Figure 2 shows the control diagram of the designed controller (16).
yaw angle is more important in a transportation task, some new Based on the feedback of the system state and the reference state, the
variables representing the integral of the tracking errors in XC , Y C , control commands for both quadrotors can be obtained with the gain
ZC , and ψ 1 directions are introduced to augment the linear system in designed in Eq. (16). Note that the calculation can be done on the
Eq. (11). Denote Ξ1 ∫ t0 rc − rref c ds, where rc is the desired
ref onboard computer of one of these two quadrotors; for example, the
transportation position in the inertial frame. Note that for a vector first quadrotor is adopted to calculate the control commands. Then,
x x1 ; · · · ; xn T , ∫ t0 x ds is defined as ∫ t0 x1 ds; · · · ; ∫ t0 xn dsT with the last four elements of f are sent to the second quadrotor. What the
Downloaded by Hugh Liu on July 8, 2021 | http://arc.aiaa.org | DOI: 10.2514/1.G005914
1
Q diagf3; 6.5; 20; 0.01; 0.01; 1; 1; 0.015; 0.0525; 0.015; 0.01;
XC (m)
0.001; 0.01; 0.1; 0.2; 0.2; 0.4; 0.2g 0.5
2 3
diagf0.3; 2; 1; 0.5g −0.2; 01;3 T 04;3 0
6 7 0 10 20 30 40 50
R 6
4 −0.2; 01;3 0.3 01;3 7
5 Time (s)
03×4 03;1 diagf2; 1; 0.5g 1
0.52
YC (m)
0.5 0.5
As shown in Fig. 3, there are no steady-state errors in XC , Y C , and 30 40 50
ZC directions. As shown in Fig. 4, in the stable configuration, the 0
modal coordinate of the beam deformation caused by the beam 0 10 20 30 40 50
gravity is less than 0.0035; hence, it is reasonable to make Time (s)
Assumption 1 in the system analysis and controller design. Further- 1.5
more, Fig. 4 indicates that the roll angle is not zero with the small
ZC (m)
1
beam deformation at the stable configuration. The main reason is that Desired trajectory
in the controller (16), only the integrals of system position and yaw 0.5 Response
angle tracking errors are included. Because of the nonzero roll angle, 0
0 10 20 30 40 50
the thrust forces provided by the quadrotors are not along the vertical
axis in the inertial frame. Time (s)
Fig. 5 The rC with stair-step signal as reference trajectory.
Downloaded by Hugh Liu on July 8, 2021 | http://arc.aiaa.org | DOI: 10.2514/1.G005914
1
XC (m)
0.5 0.5 Y direction from 20 to 25 s. All system and control parameters are
kept the same as those in the previous case, and the controller in
30 40 50
0 Eq. (16) is used. As shown in Fig. 5, the system can be driven to the
0 10 20 30 40 50 desired position with zero steady-state error. The time histories of
Time (s)
the attitude angles of the first quadrotor and the modal coordinate of
1.5 the beam are given in Fig. 6. As indicated in Fig. 6, the maximum roll
and pitch angles are much smaller than those in Fig. 4, and the
ZC (m)
1
Desired trajectory maximum magnitude of the excited deformation is also smaller.
0.5 The main reason is that the control inputs in X and Y directions are
Response
0 larger at the step times in Fig. 4 than those at stair-step times in Fig. 6;
0 10 20 30 40 50
i.e., the introduction of stair-step reference trajectories can avoid
Time (s)
large control inputs.
Fig. 3 The rC with the designed controller (16).
0.6 0.2
Euler angles (rad)
Euler angles (rad)
0.4 1 1
1
0.1 1
0.2
1 1
0
0
-0.2
-0.4 -0.1
0 10 20 30 40 50 0 10 20 30 40 50
Time (s) Time (s)
0.04 0.01
Modal coordinate
Modal coordinate
0.02
0 0
-0.02
-0.04 -0.01
0 10 20 30 40 50 0 10 20 30 40 50
Time (s) Time (s)
Fig. 4 Euler angles and modal coordinate with the designed controller Fig. 6 Euler angles and modal coordinate with stair-step signal as
(16). reference trajectory.
Article in Advance / ENGINEERING NOTES 7
V. Experimental Verification payload. The system parameters in experiments are the same as those
This section presents the experiments conducted using the York in Table 1. The weight matrices in the cost function of LQR are
University Autonomous Unmanned Vehicle (YU-AUV) Facility at chosen the same as those in Sec. IV.A.
York University. As shown in Fig. 7, the experimental system mainly Two tests with different reference trajectories are performed.
consists of an OptiTrack motion capture system, a workstation, a According to the simulation results in Sec. IV, the reference trajecto-
router, two QDrone quadrotors [26] from Quanser Consulting, Inc., ries are chosen as stair-step signals. Also, to show the difference
and a wooden ruler as the payload to be transported. The workstation between experimental and numerical results, two numerical simula-
can calculate the positions and orientations of quadrotors in the tions with the same setup as that in experimental tests are performed.
inertial frame with a sampling rate of 100 Hz based on the measure- In the first test, the system is expected to move 0.13 m∕s for 10 s
ments from 16 Flex 13 cameras in the motion capture system. The along X axis with a height of 0.5 m. The desired values of all other
mean measurement error of the motion capture system is about generalized coordinates are zero. Figure 8 presents the time histories
0.6 mm. Then rC and p1 obtained based on these two quadrotors’ of the position vector of the middle point of these two quadrotors,
pose, and the yaw angle of the first quadrotor are sent to the onboard Euler angles of the first quadrotor, and the modal coordinate of
computer of the first quadrotor via Wi-Fi. Based on the optimal the beam in both experiment and simulation. fϕ1 ; θ1 ; ψ 1 g and
feedback gain from the LQR algorithm, the eight-dimensional con- fϕ1s ; θ1s ; ψ 1s g represent the sets of the Euler angles of the first
trol command for both quadrotors can be calculated in the first quadrotor in experiment and simulation, respectively. As shown in
quadrotor with the attitude and angular velocity information from Fig. 8a, the system starts taking off at t 0. The final tracking errors
Inertial Measurement Units (IMU) at 1 kHz. The sensitivity of gyro is in X and Y directions are within 3.5 and 10 cm in the experiment,
16.4 LSB/°/s with the range 2000°∕s, and the zero-rate offset of the respectively. Also, there exists a nonzero steady-state error in the first
gyro in IMU is less than 3 °∕s. Quadrotor 1 will adopt the first four 15 s in ZC direction, and the introduction of integral terms in the
elements as its own control command and then send the remaining controller (16) eliminates this steady-state error finally. As shown in
Fig. 8b, the maximum pitch angle to achieve the movement in X
Downloaded by Hugh Liu on July 8, 2021 | http://arc.aiaa.org | DOI: 10.2514/1.G005914
four elements to the second quadrotor. The red dashed curve in Fig. 7
shows such data flow from the first quadrotor to the second one via direction is less than 0.3 rad. Because the normalized mode shape
Wi-Fi; i.e., quadrotor 2 does not need to calculate the control com- function of a cantilever beam is used, the deformation of the point
mand by itself, and it can implement the received command directly. under the second quadrotor on the beam is less than 0.025 m with the
The real-time control of quadrotors is realized using QUARC soft- modal coordinate less than 0.025.
ware from Quanser, Inc. Furthermore, as shown in Fig. 7, neodym- In the second test, the system is expected to move 1.9 m along
ium bar magnets are used to connect the quadrotor bodies and the Y axis at the height of 0.5 m and keep other generalized coordinates
zero. The desired trajectory in Y direction is the stair-step signal given
as the blue dashed curve in the second subfigure of Fig. 9a. Both the
experimental and simulation results are given in Fig. 9. As shown in
Fig. 9a, the entire system takes off at t 0 and the transportation
mission starts at t 2.5 s. In the transportation mission, X C is
controlled within 3.5 cm. In Y C direction, the tracking error is less
than 10 cm, which implies that the designed controller can complete
the payload transportation mission successfully. Also, as shown in
Fig. 9b, the modal coordinate of the excited vibration of the flexible
beam is less than 0.025; i.e., only a small deformation happens for the
flexible payload during the transportation mission. As indicated in
Fig. 9c, the movement from right to left is achieved by a negative
attitude maneuver in the roll direction. The transportation task in
X direction in Fig. 8b needs larger pitch angle than that in Fig. 9b
because the movement in X direction is mainly coupled with the
attitude maneuver in pitch direction.
In conclusion, in both tests, the flexible payload transportation task
is completed with satisfactory tracking performance and only small
Fig. 7 Experimental quadrotor transportation system. dynamic deformations of the beam are excited.
1.5 0.6
Euler angles (rad)
XC (m)
1 1.35 1
0.5 1.3 0.4
1.25 1
0 16 18 20 0.2
-0.5 1
0 5 10 15 20 0 1s
Time (s) -0.2 1s
0.2 -0.4 1s
YC (m)
0 10 20
0 Time (s)
-0.2 0.04
0 5 10 15 20
Modal coordinate
0.6 0
ZC (m)
0.1
XC (m)
0.05
0
-0.05
0 5 10 15 20
Time (s)
3
YC (m)
2
1 2
1.9
0 1.815 20 25
-1
0 5 10 15 20
Time (s)
0.6
ZC (m)
0.4 Desired
0.2 Measured
0 Simulated
0 5 10 15 20
a) rC Time (s)
0.4
Euler angles (rad)
0.2 1
Downloaded by Hugh Liu on July 8, 2021 | http://arc.aiaa.org | DOI: 10.2514/1.G005914
0 1s
1s
-0.2 1s
0 10 20
0.04
Modal coordinate
0.02
0 Time (s)
-0.02 Measured
Simulated
-0.04
0 5 10 15 20
Time (s)
b) Euler angles and modal coordinate c) The snapshots
Fig. 9 The second experiment with the desired movement in Y direction.
2Jq3 g, M04 Jq1 dΦ1 ∕dyjyl − 1∕2ρl∫ l0 Φ1 dy ρ∫ l0 yΦ1 dt Jq3 dΦ1 ∕dyjyl 2ρ∫ l0 yΦ1 y dy mql, C0;13 −1∕2ψ_ 1 ×
1∕2mq l; 0; 0T , and M05 mq ∕2 Jq1 dΦ1 ∕dyjyl 2 dΦ1 ∕dyjyl −Jq3 J q2 Jq1 , and C0;14 −1∕2θ_ 1 ×
1∕4ρl − ρ∫ l0 Φ1 dy ρ∫ l0 Φ21 dy, respectively. dΦ1 ∕dyjyl −Jq3 J q2 Jq1 .
Article in Advance / ENGINEERING NOTES 9
rotor UAV: Hybrid Model, Trajectory Generation, and Control,” “Cooperation in the Air: A Learning-Based Approach for the Efficient
Autonomous Robots, Vol. 41, No. 8, 2017, pp. 1629–1643. Motion Planning of Aerial Manipulators,” IEEE Robotics and Automa-
https://doi.org/10.1007/s10514-017-9632-2 tion Magazine, Vol. 25, No. 4, 2018, pp. 76–85.
[5] Chen, T., and Shan, J., “A Novel Cable-Suspended Quadrotor Trans- https://doi.org/10.1109/MRA.2018.2866766
portation System: From Theory to Experiment,” Aerospace Science and [19] Lee, H., Kim, H., and Kim, H. J., “Planning and Control for Collision-
Technology, Vol. 104, Sept.2020, Paper 105974. Free Cooperative Aerial Transportation,” IEEE Transactions on
https://doi.org/10.1016/j.ast.2020.105974 Automation Science and Engineering, Vol. 15, No. 1, 2016,
[6] Qian, L., and Liu, H. H., “Path-Following Control of a Quadrotor UAV pp. 189–201.
with a Cable-Suspended Payload Under Wind Disturbances,” IEEE https://doi.org/10.1109/TASE.2016.2605707
Transactions on Industrial Electronics, Vol. 67, No. 3, 2020, [20] Mellinger, D., Shomin, M., Michael, N., and Kumar, V., “Cooperative
pp. 2021–2029. Grasping and Transport Using Multiple Quadrotors,” Distributed
https://doi.org/10.1109/TIE.2019.2905811 Autonomous Robotic Systems, Springer, New York, 2013, pp. 545–558.
[7] Geng, J., and Langelaan, J. W., “Estimation of Inertial Properties for a [21] Loianno, G., and Kumar, V., “Cooperative Transportation Using Small
Multilift Slung Load,” Journal of Guidance, Control, and Dynamics, Quadrotors Using Monocular Vision and Inertial Sensing,” IEEE
Vol. 44, No. 2, 2021, pp. 220–237. Robotics and Automation Letters, Vol. 3, No. 2, 2017, pp. 680–687.
https://doi.org/10.1007/s10514-017-9632-2 https://doi.org/10.1109/LRA.2017.2778018
[8] Villa, D. K., Brandao, A. S., and Sarcinelli-Filho, M., “A Survey on [22] Ritz, R., and D’Andrea, R., “Carrying a Flexible Payload with Multiple
Load Transportation Using Multirotor UAVs,” Journal of Intelligent Flying Vehicles,” 2013 IEEE/RSJ International Conference on Intelli-
and Robotic Systems, Vol. 98, May 2020, pp. 267–296. gent Robots and Systems, Inst. of Electrical and Electronics Engineers,
https://doi.org/10.1007/s10846-019-01088-w New York, 2013, pp. 3465–3471.
[9] Sreenath, K., and Kumar, V., “Dynamics, Control and Planning for https://doi.org/10.1109/IROS.2013.6696850
Cooperative Manipulation of Payloads Suspended by Cables from [23] Kotaru, P., and Sreenath, K., “Multiple Quadrotors Carrying a Flexible
Multiple Quadrotor Robots,” Proceedings of Robotics: Science and Hose: Dynamics, Differential Flatness and Control,” IFAC-PapersOn-
Systems, Berlin, Germany, June 2013. Line, Vol. 53, No. 2, 2020, pp. 8832–8839.
https://doi.org/10.15607/RSS.2013.IX.011 https://doi.org/10.1016/j.ifacol.2020.12.1396
[10] Tagliabue, A., Kamel, M., Siegwart, R., and Nieto, J., “Robust Collabo- [24] Yang, H., Staub, N., Franchi, A., and Lee, D., “Modeling and Control of
rative Object Transportation Using Multiple MAVs,” International Multiple Aerial-Ground Manipulator System (MAGMaS) with Load
Journal of Robotics Research, Vol. 38, No. 9, 2019, pp. 1020–1044. Flexibility,” 2018 IEEE/RSJ International Conference on Intelligent
https://doi.org/10.1177/0278364919854131 Robots and Systems (IROS), Inst. of Electrical and Electronics Engi-
[11] Masone, C., Bülthoff, H. H., and Stegagno, P., “Cooperative Trans- neers, New York, 2018, pp. 1–8.
portation of a Payload Using Quadrotors: A Reconfigurable Cable- https://doi.org/10.1109/IROS.2018.8593834
Driven Parallel Robot,” 2016 IEEE/RSJ International Conference on [25] Dwivedy, S. K., and Eberhard, P., “Dynamic Analysis of Flexible
Intelligent Robots and Systems (IROS), Inst. of Electrical and Electron- Manipulators, a Literature Review,” Mechanism and Machine Theory,
ics Engineers, New York, 2016, pp. 1623–1630. Vol. 41, No. 7, 2006, pp. 749–777.
https://doi.org/10.1109/IROS.2016.7759262 https://doi.org/10.1016/j.mechmachtheory.2006.01.014
[12] Qian, L., and Liu, H. H., “Path Following Control of Multiple Quad- [26] “Qdrone—QUANSER,” 2021, https://www.quanser.com/products/
rotors Carrying a Rigid-Body Slung Payload,” AIAA Scitech 2019 qdrone [retrieved 10 May 2021].