Professional Documents
Culture Documents
(ICISC-2017)
The mathematical modeling of the quadrotor is based Based on the above assumptions and simplifications we
on the vast amount of work that has been carried out derive our dynamic model which is a set of equations
by C. Balas[3], S. Bouabdallah[4] and using the Euler- of motion. These are highly coupled non-linear dynamic
Lagrangian and Newtonian approaches. We use the standard equations[3]; however we dramatically simplify them but
approaches and obtain our own derived dynamic model. keep them effective for governing the control over the
Standard assumptions have been taken into consideration and quadrotor. The standard model equations[4] are as follows
we make a few of our own assumptions that are realistic and which describe the moment of inertia along the axes of the
reasonable. The standard assumptions are as follows- The quadrotor.
structure of quadrotor is rigid in nature. It is symmetrical
and the origin of the frame and center of gravity coincide. Ixx φ̈ = θ̇ ψ̇(Iyy − Izz ) + Jr θ̇ Ωr + τx (3)
Few pre-requisites are needed for a clear understanding of the Iyy θ̈ = φ̇ ψ̇(Ixx − Izz ) + Jr Ω̇r + τy (4)
dynamics, these are - The dynamic modeling of the quadrotor
is always carried out with the aid of body frame and inertial Izz ψ̈ = θ̇ φ̇ (Ixx − Iyy ) + τz (5)
frame.
τx = bl(Ω24 − Ω22 ) (6)
The figure-2 shows a plus (+) and X type alignment. The
body frame is defined by the orientation of the quadrotor τy = bl(Ω23 − Ω21 ) (7)
whereas the inertial frame is governed by the earth/ground.
(negative z-axis pointing towards ground). The relation be- τz = d(Ω21 − Ω22 + Ω23 − Ω24 ) (8)
tween the body frame and inertial frame is given by the
rotation matrix R. The state vector is X = φ φ̇ θ θ̇ ψ ψ̇ zż yẏ xẋ
R= In the above equations θ̇ ψ̇(Iyy − Izz ) = body gyro effect
and Jr θ̇ Ωr = propeller gyro effect.
cθ cψ cθ sψ −sθ
−cθ sψ + sφ sθ cψ cφ cψ + sφ sθ sψ sφ cθ
B. Simplifying EOM using Matlab
sφ sψ + cφ sθ cψ −sφ cψ + cφ sθ sψ cφ cθ
The ability of the quadrotor to do acrobatic maneuvers
where ’c’ denotes cos and ’s’ denotes sin. Another important results in complex non-linear dynamic equations[1]. There
parameter to the flight dynamics is the Inertia matrix which are six complex equations of motion.This creates compli-
describes the mass moment of inertia across the assumed cations in analyzing the behavior of any system. In our
axes of the rotor. This matrix plays a major role in dynamics approach we aimed at simplifying these complex equations
of both translation and rotational motion. The way the mass using MATLAB-SIMULINK[2]. A far better insight to this
is distributed along the axes holds a remarkable significance. is presented in the section-V.
The parallel axis theorem is used to obtain the Inertia matrix. The model in figure-3 is a subsystem of the Equations of
Jb= ⎛ ⎞ motion model that demonstrates one of the six equations of
Jx 0 0
⎝ 0 Jy 0 ⎠ motion implemented in Simulink. The function block rep-
resents the instantaneous equation taken into consideration.
0 0 Jz
The equation needs to be integrated twice to obtain position
where Jx, Jy and Jz are the mass moments of inertia along and velocity data. We extract the state space model as it
X, Y and Z axis respectively. The translational motion is serves as an effective approach to not only simplify the
defined by higher order systems but also gives us realistic and equitable
∑F = m∗a (1) results.
P = (φ , θ , ψ − φ ∗ , θ ∗ , ψ ∗ ) ∗ Pgain (9) techniques used are PD and PID control. Various perfor-
mance metrics have been taken into consideration to ensure
a robust adaptive system.
I = I + (φ , θ , ψ − φ ∗ , θ ∗ , ψ ∗ ) ∗ Igain (10)
B. Performance evaluation
The performance analysis of the quadrotor is done by
D = (φ , θ , ψ) − (φ ∗ , θ ∗ , ψ ∗ ) − (m) − (n) ∗ Dgain (11) tuning the control parameters and observing their effects
on the system. Performance metrics like stability, peak
where m= previous value of φ , θ , ψ (from gyro) overshoot, rise time, settling time and steady state error are
n= previous value of φ ∗ , θ ∗ , ψ ∗ (from transmitter) key parameters in our evaluation methodology. We made a
IV. S IMULTAION AND IMPLEMENTATION RESULTS small comparative analysis of the effects of tweaking the
control parameters on these performance metrics. These are
A. Simulation in Matlab presented in a tabular format in table-II. Remarks have been
For simulating our quadrotor’s dynamics, the platform we made on the same in discussions section. The graphs ob-
have used is MATLAB-Simulink. Simulink proves to be an tained of the performance evaluation results of our quadrotor
appropriate choice as it offers ease of use and unparalleled system in figure-9 to figure-18 exhibit a good response and
processing power[8]. As the quadrotor is a highly non-linear
complex system we made two simulation models. The first
model in figure-6 is an implementation of the six equations
of motion.
This simulation serves as a ground zero for the user to
understand the mathematics and dynamics related to flight.
The user gets a glimpse of the aerodynamic aspects related Fig. 9. Stabilizing phi to reference input with PID
to the quadrotor.
This builds a concrete foundation for the second simu-
lation as shown in figure-7 which is the full-fledged archi-
tecture model of the quadrotor. The various system blocks
can be seen; however these further contain sub-systems (not
visible in the figure) and other parameters. The controlling Fig. 10. Stabilizing theta to reference input with PID
C. Discussions
A comparative analysis of PD and PID technique can be
done from the results obtained. We observed that the PD
Fig. 13. PD vs PID response for Phi(step input) technique was easier to implement but was not sufficient
to cope-up with the complex flight dynamics. With the
PID controller the problem of integral wind-up was faced
which is eliminated by optimal tuning values. We faced the
following occurrences during our testing- An increment in
Kp would many times lead to overshoot and reduce the
stability however it increased response speed for controlling
action. Increment in Ki lead to a slower stable system. Kd
would serve as the most important parameter as it would take
corrective actions on overshoot due to P and in such real
Fig. 14. PD vs PID response for Theta(step input) time applications like a flying vehicle tuning this is critical.
With a large amount of experimental tests carried out we
came up with PID values that proved best for our system and
also we tweaked them to make the quad do some complex
maneuvers.
TABLE II
P ERFORMANCE METRICS
Parameter Kp Ki Kd
Fig. 15. PD vs PID response for Psi (step input) Stability Reduces Reduces Increases
Accuracy Increases Increases Increases
Response speed Increases Reduces Increases
V. C ONCLUSION
Fig. 16. PID response for complex maneuvers (roll variations)
A robust control model is designed with constraint opti-
mization approach. An efficient controller has been achieved
by making an IMU using low cost MPU 6050 sensor with
use of PD and PID control techniques. The simulations and
hardware results show the system as a stable platform and no
trade-offs in performance of the quadrotor even while doing
Fig. 17. PID response for complex maneuvers (pitch variations)
complex maneuvers. Also an additional PID algorithm on
the Z-axis gives a self-leveling feature if there is no control
action from transmitter. Our Arduino based architecture can
be modified to implement other control techniques like IDA-
PBC and back-stepping control. Even though using low cost
components our controller design and algorithm results in
Fig. 18. PID response for complex maneuvers (yaw variations) stable flights and do some aggressive maneuvers. For future