System identification of a quadrotor UAV Turgay Kale / Güleser Kalaycı Demir
Oct 26, 2014
System identification of a quadrotor UAV
Turgay Kale / Güleser Kalaycı Demir
Overview
Scope
Literature
My studies and goal
Basic concepts
Kinematics and dynamics
Simulink model
Things to do
Evaluation
Scope
Understanding the law and facts behind the flight of an UAV
Dynamic modelling of a quadrotor
Controlling of a rigid quadrotor plant
Building a Simulink model of the entire flight system
Evaluation of simulation test benchmark
Literature
Project owner University Control method Picture
Bauabdallah
ÉCOLE POLYTECHNIQUE
FÉDÉRALE DE LAUSANNE
Backstepping, PID,
Optimal control,
Sliding mode
Fower
Brigham Young University
Visual feedback
MD4-200R
Micro Drones
Gmbh
PI
Weng & Shukri
Universiti Teknologi
Malaysia
PID
Literature
Project owner University Control method Picture
Starmac
Stanford
University
Reinforcement
learning
Hanford
Pennysylvania
University
PI
HMX4
Micro Drones
Gmbh
Feedback
linearization
My studies & goal
Efforted in data correction of Sparkfun’s ultimate imu
Data acquisition & filtering effort with 9-dof Imu
Arduino platform
Open source libraries : SD card
Proper peripheral units for control
Worldwide support
Pololu IMU-9
9 degree-of-freedom
I²C support
Cheap cost & good feedbacks
Dynamic modelling effort in Matlab/simulink
To complete accurate simulink model (goal)
Basic concepts
Throttle (N) Roll (N.m)
Pitch (N.m) Yaw (N.m)
Quadrotor movements are ensured by controlling the rotational speeds of 4-rotors
Kinematics
Position and attitude vector with respect to E-frame
Linear position vector :
Angular attitude vector :
Generalized position vector w.r.t. E-frame
Kinematics
Rotations between E-frame and B-frame
Kinematics
Hence overall rotating matrice is
Velocity vectors with respect to B-frame
Linear velocity vector :
Angular velocity vector :
Generalized velocity vector w.r.t. B-frame
Kinematics
Thus it is possible that and where
As a result
and
Hence the equivalent form of statements becomes,
Dynamics
Newton-Euler formulation is adopted
Three assumptions are made to decrease the computation complexity
Inertia matrix is time-invariant (fundamental)
Origin of the body-fixed frame is assumed as the center of mass (COM)
In hovering condition, moment of inertia matrix I is approximately diagonal
From the Euler’s first axiom of the Newton’s second law ;
Dynamics
where
From Euler’s second axiom of the Newton’s second law ;
is body inertia matrix (in the body-fixed frame)
is torque vector w.r.t. E-frame
is quadrotor total mass
is rotation matrix derivative
is force vector w.r.t. E-frame
Dynamics
By associating two axioms, it is possible to define motion dynamics of rigid
body. Below equation shows the matrix formulation of dynamics.
(1)
Let’s define a generalized force (including torque) vector as;
where , means an identity matrix with dimension 3 times 3. First matrix in
equation is diagonal and constant.
Dynamics
Therefore it is possible to rewrite equation (1) in matrix form
where
is generalized acceleration vector w.r.t. B-frame
is system inertia matrix w.r.t. B-frame
is Coriolis-centripetal matrix w.r.t. B-frame
Dynamics
generalized force vector can be splitted into three components according
to the nature of quadrotor contributions.
First contribution is gravitational vector
Second contribution is gyroscopic effects produced by the propeller rotation.
Dynamics
where is total rotational moment of inertia.
Third contribution is aerodynamic effects produced by main movements. In this
phenomenon, forces and torques are assumed to be proportional to squared propeller’s
speed.
where is movement matrix
Dynamics
Hence, it is possible to describe the dynamics considering these last three
contributions according to equation (1) .
By re-arranging , it is possible to write equation (23) ;
Dynamics
Hybrid model
Generalized velocity vector in hybrid notation called H-frame ;
By rewriting in a matrix form based on H-frame ;
Dynamics
Dynamics
By re-arranging , it is possible to write ;
Hence it is possible to write system equations in hybrid system as below ;
Simulink model
task control
sensors
dynamics +
-
IMU 9-dof
IR
PI/PID
control
commands
quadrotor
plant
position
tracking states
Things to do
Complete simulink model
Benchmark test results & restorate model
Hardware implementation & tests (partialy optional)
Design 3D computer simulation model based on openGL (optional)
Port simulink model blocks to microsoft .net framework (optional)
End of defence