IEEE Robotics & Automation Magazine - September 2017 - 28

Throttle
< 50%

Standby

Standby

(b)
Surge

Stopping

Throttle (%)

∆t > 2 s

Standby
Gliding
Surge
Stopping

100

50
Input
Output

φm ≤ φ ≤ φM

Gliding

Throttle
> 50%

0

(a)

(c)

15

20

25
Time (s)

30

35

(d)

Figure 9. The gliding mode controller automatically engages the mechanism's passive latch when the throttle drops below a threshold.
(a) A finite state machine displaying controller states and transition guards. The surge state is required to pull the bird out of gliding
mode, by supplying a high throttle for a short time. (b) The range wherein the lift force can engage the gliding lock; the dihedral angle
range. (c) The drive shaft/latch range. (d) A demonstration of the gliding controller. The throttle threshold is 50%. The input is the
manual RC signal received from the pilot and output goes to the ESC.

zz(((m)
m
m)

z (m)

could fly the Robird or program its flight path. When in
gliding mode, the bird closely resembles a normal glider
plane, and the standard autopilot of the Pixhawk/APM system suffices. The challenge lies in the flapping flight and in
switching between modes.
50
0

80
60
40
20
0

100
−100 −50

−− 100
100 −−50
50

0
50 100
150
x (m)
(a)

0

00
5
00 1100
50
50
000150
550
00
150−−50
50
xx(m
(m) )
(b)

00

y (m)

50
50

100
100

(mm))
yy((m)

Figure 10. A 3-D flight path log. The bird was brought to
altitude in fly-by-wire mode, then autonomously loitered around
(0, 0, 80) and (0, 0, 60) [(x, y, z) setpoint in meters] before
being brought down to land by the pilot. (a) Flying from blue to
red. (b) An anaglyphic 3-D flight path [same as 10(a); view with
red/cyan glasses].

Control
Although there are a few very skilled pilots who can manually fly the Robirds, the cognitive load is too extensive when
flying close to runways for an entire day. An autopilot system should make flying easier, e.g., in the end, anybody
28

*

IEEE ROBOTICS & AUTOMATION MAGAZINE

*

SEPTEMBER 2017

Flight Modes
As shown in the "System" section and Figure 2, the flapping
mechanism features a latch that acts as a lock for a passive
gliding mode. For the latch to engage-due to the aerodynamic lift pushing the notch wheel backward-the wings
must be stopped in the right position, at a dihedral angle of
approximately 20° [Figure 9(b)]. A gliding controller takes
care of automatically engaging the lock when the throttle
drops below a threshold, as depicted in Figure 9. The main
drive shaft has an angular encoder measuring the shaft angle
z ; the latch can engage when z ! [z m, z M].
Autopilot
While research on a workable aerodynamic model of the bird
in flapping flight is in progress, for stabilizing control, we
must rely on existing autopilot controllers offered by the Pixhawk/APM board, tuning parameters by observations of
many failed flights, pilot experience, and logged data analysis.
We are now at a point where the autopilot can successfully
stabilize the system.
There are some challenges left, most of which relate to
flapping flight-with vertical accelerations of 3 g in amplitude and heavy pitching, even state estimation is not
straightforward. The reason that the autopilot, which was
designed for fixed-wing aircraft, only has to be slightly
adapted for our flapping-wing system is that the flapping
frequency is relatively low compared to smaller-scale ornithopters: 6 Hz is well within the range of the aerodynamics
of fixed-wing aircraft, and the autopilot can handle it.
Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - September 2017
