Sie sind auf Seite 1von 2

ME 102B Report

ME 102B Report
Prithvi Akella, Carl Canteenwala, Phillip Downey, and Michael White
(Dated: 7 May 2018)
This report details Group Thirteen’s efforts to realize a self-stable monowheel. Briefly speaking, we exploited
the effects of Gyroscopic Precession to inform our system design, and controlled it with saturated PI control.

I. THEORETICAL FOUNDATIONS in that predefined rotation. Assuming frictional contact


with the ground will not retard the rotation of the wheel,
Gyrosopic Precession denotes the precession of the ro- gravitational torque will ensure that every subsequent
tation axis of a system due to outside torques (usually H rotates slightly more towards the −i direction (with
imposed by gravity). Without loss of generality, we can respect to the wheel), the proof for which is as follows:
define the rotation of our outer wheel to be in the +j
direction, translation in the +i direction, and gravity in Mg = ro sin θj × − mgk = Ḣ (2)
the −k direction as in Figure 1. Furthermore, we can Ho + Ḣdt = Iyy ωj − mgro sin θdti. (3)

The latter half of Euler’s equations implies that the nor-


mal force at the contact point with the ground is equal
and opposite to the gravitational force. In other words,
gravitational torques acting on a rotating and translating
system only serves to rotate the system in the direction
of its tilt - very much like a top precessing when it tips
in a direction. We exploited this phenomena to help us
turn in the direction we would like, and attempted to
upright our system by providing a torque in the +k di-
rection so as to shift the angular momentum vector in
the +k direction as well.

II. MECHANICAL DESIGN

At a high level, the monowheel consists of an in-


ternal ring gear and brushless DC motor drive system
FIG. 1. Orientation of our system with servo-actuated stabilizing wheels and a brushed DC
motor actuated flywheel. Three primary considerations
also define the moment of inertia of our system about drove the mechanical design of the monowheel: manufac-
it’s geometric center to be Iyy , it’s mass to be m, and it’s turabilty, location of center of mass and overall aesthet-
radius to be ro . With this setup, if our system rotates ics.
about an axis that coincides with the point of contact To ensure that each component could be readily pro-
at the ground and is parallel to the +i direction, it fol- duced, we kept the intended manufacturing process in
lows that the center of mass of the system will cease to mind throughout the design cycle. We considered the
remain above the contact point. For future purposes we limitations and capabilities of the resources and work
will call this phenomena destabilization and account for spaces available to us. For parts requiring tight toler-
it through the destabilization angle, θ, as measured from ances, we utilized machined aluminum, while for those
an axis perpendicular to the ground plane that coincides not requiring tight tolerances, we leveraged the rapidity
with the ground contact point. Once the system is un- of 3D printing and laser cutting.
stable, gravity’s action at the center of mass, displaced Given that the overarching fabric of our project re-
by ro sin θj from the contact point, will apply a torque to volved around stabilizing an inherently unstable system
the system in the −i direction about the contact point in the vertical position, we designed and placed parts
at the ground. From this point, Euler Balance equations in the system such that the center of mass of the en-
apply wherein, tire wheel was centered both laterally, and in the fore
and aft directions. We achieved lateral balance by sim-
M = Ḣ, F = ma. (1) ply designing each component to be symmetrical about
the ik plane of the wheel. For the fore and aft stabil-
Noting that our system originally had H = Iyy ω where ω ity, we located each part within the wheel in such a way
denotes the rotation rate of the wheel in the +j direction, that both allowed the individual elements to function as
the resulting gravitational torque identifies the change needed and evenly distributed the mass throughout. Due
ME 102B Report 2

to the presence of the stabilizing wheels when the system for given duty cycles, choosing an initial Kp as a factor
is at low speed, we actually wanted the center of mass to of 10, and iterating from there. Eventually we settled on
be slightly behind the vertical diameter of the wheel so Kp = 2 × 10−5 which yielded a startup time of roughly
that the stabilizing wheels would engage with the driving one second, with minimal overshoot and oscillation.
surface. Flywheel stabilization hinged on simple bang-bang
As with any product, we wanted to fuse aesthet- control wherein if we measured a destabilization by ro-
ics with functionality. To this end, we chose component tation of the system in the −i direction, we provided the
designs and materials with the overall look of the final max torque possible to the flywheel in the −k so as to
product in mind, making sure to round off any harsh provide a torque to our system in the +k direction to sta-
edges and maintain design uniformity. As with any good bilize our system. Automatic training wheel retraction
project, form followed function, and to a certain extent, mechanisms hinged on a case structure associated with
the aesthetics of the project fell in to place simply by whether or not the system was translating fast enough.
ensuring that each component functioned properly both If the system surpassed a predefined minimal speed, then
on its own and within the context of the whole. the training wheels retracted automatically. If not, they
remained on the ground in perpetuity. Finally, for each
control system (speed, flywheel stabilization, and train-
III. CONTROL THEORY ing wheel retraction) we implemented an override button
so as to ensure safety of our system through repeated
tests.
In order to meet the constant speed assumption pre-
sented in the Theoretical Foundations section, we re-
quired that our system always translate at a consistent IV. ELECTRICAL SYSTEMS
velocity. Given that our resulting system only had one
brushless motor coupled to a shaft, we only required one
Our final system consisted of one 12V battery which
speed control system. We estimated that the since the
powered all the motors. Regulating the supply down to
encoder speed measurement would be highly noisy, that
5V provided power for the ESC computation systems,
we would proceed with a simple PI control system. We
servo motors for the stabilization wheels, encoder, and
also noted that for safety reasons, we would like to control
IMU. The myRio is powered by its own battery pack and
max speed (max duty cycle) and max acceleration (max
shared a ground with the actuator 12V battery so as to
change in duty cycle). To achieve this, we constructed
shield the Rio from any unwanted current spikes. Having
a simple control system wherein our input at every time
two battery packs was also beneficial from a mechanical
step k was constructed as,
design standpoint as we were able to shift significant mass
fore and aft to correctly balance the tilt about the j axis.
uk = Kp (r − y) + uk−1 . (4)

If we note that at k = 0, our input, u = 0, then at every V. RESULTS


successive time step, the input could be reconstrued as,

k The project succeeded in creating a self-stable


monowheel, capable of driving up to about 2 m/s. The
X
6 0
uk = Kpn (r − y[n]) Kp0 =0 Kp= = Kp . (5)
n=0
system design came together well as a proper and thor-
ough 3D model was developed. Upon assembly, it became
In continuous time, our Open Loop Transfer Function, evident that the brushed motors initially selected to drive
L, becomes, the system were insufficient to overcome the friction in
the ring gear. As a result, a new brushless motor with a
P Kp G significantly better ability to deliver torque to our system
G= L= , (6) was selected. The new motor had little trouble overcom-
1 − e−sτ 1+G
ing the friction in the gears and was never operated at
where τ is our sample time. Note that this setup achieves more than 20 percent of its operating capability.
integral control while only requiring one gain, and the Another realization upon assembly was that the motor
choice of that gain affects rise time, settling time, and selected to turn the flywheel could not generate enough
overshoot. Noting that our plant, the brushless motor, torque to meaningfully affect our system once it was at
has inherent high-frequency rolloff, we noted that having speed. A future iteration of the project would attempt
a low sample time, thereby funneling values to the motor to reduce mass of the entire system, and select a motor
at high frequency, would counteract the effect of sensor with a larger torque output.
noise on our output. Finally, we chose the appropriate Kp For more text, images and videos, please visit
for speed control by noting nominal encoder speed values www.wheeltime2018.weebly.com.

Das könnte Ihnen auch gefallen