IEEE Robotics & Automation Magazine - December 2018 - 64

1.2
1.15
1.1
z (m)

1.05
1
0.95
0.9
0.85
0.8

-0.1

-0.05

0
y (m)
(a)

0.05

0.1

Angle (°)

Orientation of the UAV

φb
θb
ψb

40
30
20
10
0
-10
-20
Position of the End Effector

Position (m)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18

xe
ye
ze

1.5
1
0.5
0
400

450

Trial 6

500

Trial 7

1,000

Trial 15
Time (s)

1,050 1,100
Trial 16

(b)
Figure 9. The results of the repeatability experiment in which
18 consecutive trials of the multimodal locomotion procedures
were performed. The surface-movement trajectories are shown
in (a). A trend can be noted in the point of attachment over
time, which is indicated by the circles. In (b), the orientation of
the multirotor and the position of the end effector is plotted for
trials 6 and 7 and for trials 15 and 16.

environment occurred during approach (see x e), which
resulted in oscillations in i b, causing a vertical offset in the
point of attachment (see z e) . From the video it appears
that the horizontal offset in the 16th trial was caused by a
bounce due to disturbances as well, in turn causing the
system to make contact a bit farther to the side. We notice
no remarkable behavior in the plots for trial 16.
Lessons Learned
From the experiments, several lessons were learned.
● Slipping of wheels occurred frequently, resulting in uncontrolled rotations of the end effector, which caused additional
disturbances to the multirotor. The ground-locomotion
system may benefit from a design with only two
64

*

IEEE ROBOTICS & AUTOMATION MAGAZINE

*

December 2018

●

●

●

perpendicular actuated omnidirectional wheels, aligned
such that slipping does not cause uncontrolled rotation. The
rotation of the end effector on the surface may be controlled
solely by the multirotor and the elastic element.
Under certain relative orientations between the end effector and the multirotor, the elastic decoupling element
applies undesired torque. This can cause disconnection of
certain wheels, even if the net interaction force applied by
the multirotor is properly aligned.
The contact controller demonstrates resilience to the disconnection of individual wheels from the surface but
becomes unstable whenever there is no contact with the
environment, even for a brief moment. Therefore, persistence of contact is crucial, but persistent contact of all three
wheels is not required.
The contact controller outputs desired torque and thrust
values. Given the importance of aligning the interaction
force to the approach, it is crucial to have an accurate mapping from rotor thrust to rotor velocities. This mapping
depends on voltage. Therefore, the use of batteries (e.g. in
outdoor applications) may pose additional challenges.

Conclusions and Future Work
The multimodal locomotion system evaluated in this article is
an approach to active tool handling in remote locations. The
system was designed as an aerial manipulator carrying an end
effector composed of three actuated omnidirectional wheels
and a tool. The aerial manipulator represented the aeriallocomotion system, while the end effector represented the
ground-locomotion system. The end effector was specifically
designed to allow the repositioning of a tool on the surface.
To deal with the issues related to the deployment this kind of
platform, as described in the article, a control strategy based
on the authors' previous work was modified and implemented on an experimental setup.
Experiments in which a 5 # 10 -cm area was successfully
cleaned with a 3-cm-diameter brush validated the approach.
The results prove that the applied control strategy successfully counteracted the disturbances on the aerial platform
introduced by the locomotion of the end effector, with the
angular errors remaining below 7.5°. The repeatability of the
multimodal locomotion approach was qualitatively demonstrated by an experiment in which 18 consecutive trials were
stably performed.
The success of these experiments naturally suggests an
extension of this approach to surfaces of any orientation and
curvature so that a broader range of scenarios might be examined. Furthermore, strategies for trajectory generation can be
investigated to seamlessly transition from free flight to contact
with the environment. Finally, the switching nature of the
proposed control strategy should be studied.
Acknowledgments
This work was funded by the European Commission's H2020
project AEROWORKS under Grant 644128. We thank G. te
Riet o/g Scholten, V. Groenhuis, and S. Smits for their advice
IEEE Robotics & Automation Magazine - December 2018

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - December 2018
