given environment. This cannot be attributed to its structural robustness; the Y6 is definitely the least “sturdy” of
the three. We conjecture that the different control logic of
the Y6 offers additional opportunities to reactive control.
A similar reasoning applies to Cleanflight, as shown in
Figure 7(a). Being the youngest of the autopilot we test,
it is fair to expect the control logic to be the least refined.
Reactive control is still able to drastically improve the pitch
error, by a 32% (37%) factor with the quadcopter (
hexacopter) in Arch.
The improvements in attitude error translate into
more accurate motion control and fewer attitude corrections. As a result, energy utilization improves. Figure 7(b)
shows the results we obtain in this respect. Reactive control reaches up to a 24% improvement. This means flying more than 27min instead of 22min with OpenPilot in
Arch. This figure is crucial for aerial drones; the improvements reactive control enables are thus extremely valuable. Most importantly, these improvements are higher
in the more demanding settings. Figure 7(b) shows
that the better resource utilization of reactive control
becomes more important as the environment is harsher.
Similarly, the quadcopter shows higher improvements
than the hexacopter. The mechanical design of the latter already makes it physically resilient. Differently, the
quadcopter offers more ample margin to cope with the
environment influence in software.
5. END-USER APPLICATIONS
The performance improvements of reactive control reflect
in more efficient operation of end-user drone applications
ranging from 3D reconstruction to search-and-rescue. 18
The latter is a paradigmatic example of active sensing
functionality, whereby data gathered by application-spe-cific sensors guides the execution of the application logic,
which includes here the drone movements. We build a
prototype system to investigate the impact of reactive control in this kind of applications.
System. Professional alpine skiers are used to carry
a device called Appareil de Recherche de Victimes en
Avalanche (ARVA) 20 during their excursions. ARVA is noth-
ing but a 457KHz radio transmitter expressly designed
safety, most GCS implementations instruct the drone
to return to the launch point upon reaching this thresh-
old. In general, the lifetime of aerial drones is currently
extremely limited. State of the art technology usually pro-
vides at most half an hour of operation. This aspect is
thus widely perceived as a major hampering factor.
In the following, we describe an excerpt of the results
we collect based on 260+ hours of test flights performing
way-point navigation in the three environments. 7
Results. As an example, Figure 7(a) shows the average
improvements in pitch error; these are significant, ranging from a 41% reduction with Cleanflight in Lab to a
27% reduction with Ardupilot in Arch. We obtain similar
results, sometimes better, for yaw and roll. 7 Comparing
this performance with earlier experiments, we confirm
that it is the ability to shift processing resources in time
that enables more accurate control decisions. 7 Not running the control loop unnecessarily frees resources,
increasing their availability whenever there is actually the
need to use them. In these circumstances, reactive control
dynamically increases the rate of control, possibly beyond
the pre-set rate.
Evidence of this is shown in Figure 8, showing an example trace that indicates the average control rate at second
scale using Ardupilot and the hexacopter. In Arch, reactive control results in rapid adaptations of the control rate
in response to the environment influence, for example,
wind gusts. On average, the control rate starts slightly
below the 400Hz used in time-triggered control and slowly
increases. An anemometer we deploy in the middle of the
field confirms that the average wind speed is growing during this experiment.
In contrast, Figure 8 shows reactive control in Lab
exhibiting more limited short-term adaptations. The average control rate stays below the rate of time-triggered
control, with occasional bursts whenever corrections are
needed to respond to environmental events, for example,
when passing close to a ventilation duct. The trends in
Figure 8 demonstrate reactive control’s adaptation abilities both in the short and long term.
Still in Figure 7(a), the improvements of reactive control apply to the Y6 as well; in fact, these are highest in a
Quadcopter - Ardupilot
Hexacopter - Ardupilot
3DR Y6 - Ardupilot
Quadcopter - Cleanflight
Hexacopter - Cleanflight
Quadcopter - OpenPilot
Hexacopter - OpenPilot
Lab Rugby Arch P i t c
(a) Pitch error improvement.
Lab Rugby Arch F l i g
(b) Flight time improvement.
Figure 7. Performance improvements with reactive control.