IEEE Robotics & Automation Magazine - December 2018 - 41

nonlinearly coupled. Therefore, it is difficult to constrain them
separately. This means that this allocation does not account for
the slower dynamics of the tilting angles. An example of a situation in which this becomes critical is hovering at a roll angle
of 90c. In this configuration, one of the three rotor axes is
completely vertical. Therefore, the two thrust motors on this
axis can produce a force only in the horizontal direction.
Hence, they are not used for counteracting gravity but for disturbance rejection and horizontal position stabilization.
These two motors are required to constantly rotate 180c
back and forth at a high rate to provide the desired counteracting moments and forces. The dynamics of the tilting motors
are, however, much slower than those of the thrust motors
and, as a result, cannot keep up with the desired allocation
commands. This leads to forces and moments being produced
in directions that have not been commanded. To overcome
this, the allocation was modified to exclude these two motors
in this particular configuration. In this configuration, the system flies with the four remaining motors, which still allow it to
control all 6 DoF. This is further explained in [25].
Evaluation

Wall Interaction Experiments
To enable physical interaction with the wall, a three-sphere
module that can roll passively was mounted on the Voliro platform, as shown in Figure 11. The module was passively compliant to improve interaction stability and reduce oscillation
during contact. The phases to achieve physical interaction with
the wall can be summarized as follows:
1) transition from horizontal flight to vertical flight, which is
achieved at pitch 90c, as shown in Figure 12(a)
2) approaching the wall while maintaining a pitch angle of
90c until contact is established with the compliant threesphere module; this phase is shown in Figure 12(b)
3) driving on the wall is achieved by generating a force vector
using a simple proportional controller; tracking a circle on
the wall is shown in Figure 12(c).
More details about wall interaction control and experiments
are available in [28].
The experiments demonstrated the vehicle's omnidirectionality. It was able to obtain all orientations along one
rotation axis and perform translations at an inclined orientation. However, the controller was not able to counteract
all of the system's translational and rotational dynamics,

Simulation
The controller was tested in simulation, which used Gazebo
[26] as a physics environment and to model the tilting
motors. To simulate the modeled sensors and the thrust
motors, the RotorS plugin [27] was used. The sensor data
were sent via the MAVLink protocol (pixhawk.org/dev/mavlink) to the software-in-the-loop version of the PX4 software.
The model of the system was then loaded into this environment (Figure 7). Similar to the actual system, the commands
were sent via the robot operating system to the controller.
Experimental Results
In this section, we present our experimental results. The position and orientation of the vehicle were determined with an
external motion-capture system, sent to the flight controller at
10 Hz, and fused with the onboard IMU. To showcase the vehicle's capabilities, two different maneuvers were demonstrated.
Both maneuvers are not feasible with a standard multirotor.
Moreover, we show experimental results in which the Voliro
platform was able to interact with and move along a wall.
Free-Flight Experiments
The first experiments show the rotation of the vehicle around
the body's y axis from its initial horizontal orientation to
upside-down and back (Figure 8). During this maneuver, the
other desired rotations and positions were set to zero. The
results of this experiment are shown in Figure 9. This maneuver was conducted slowly to display the vehicle's capability to
stabilize in all of the rotations around the y axis.
The second experiment displays a horizontal translation in
the x and y directions. The desired rotation around the x axis
was set constant to 50c, while the other rotations were set to
zero. The results are shown in Figure 10.

Figure 7. The model of the system in the simulation environment.

Figure 8. Snapshots of the system's rotation from horizontal to
upside-down flight.

december 2018

IEEE ROBOTICS & AUTOMATION MAGAZINE

https://mavlink.io/en/ https://mavlink.io/en/

IEEE Robotics & Automation Magazine - December 2018

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