Information Technology Reference
In-Depth Information
Fig. 1. QA3 Mini, quadrotor under development at IT Research Center
There are different orientation representation methods including Euler axis,
Euler angles, direct cosine matrix and quaternion [6]. Euler angles orientation
representation exceeds other methods in clarity and with some constraints can
be used for quadrotor flight control in hovering mode. Orientation of a rigid body
referred to a global reference can be represented by Euler angles by a sequence of
three elemental rotations (yaw, pitch and roll) about the axes of the coordinate
system of the body ( z , y and x , respectively). Usually, in the body reference
system the x-axis points forward and y-axis points right and to complete the
right-hand rule, z-axis points down.
The quadrotor can carry large number of sensors on board, accelerometers, gy-
roscopes and magnetometers among them. This sensors are generally mounted
in an arrangement to measure vehicle attitude and motion. It is well known
that MEMS (Micro Electro Mechanical Systems) gyroscopes are cheap and
lightweight but have large bias due to its operating principle. As a consequence
of this bias, the estimation of attitude angles has an important drift and grows
without limits. To eliminate this bias, accelerometers complement the gyroscopes
to estimate roll and pitch, and magnetometers are used to correct yaw angle.
Nevertheless, magnetometers suffer high distortion in indoor navigation due to
power wires and electric and electronic devices [7]. Thus, in this cases, another
sensor must be used to assist gyroscopes. Cameras are lightweight and cheap
sensors which are able to provide big amount of information. There are numer-
ous methods that can be used to recover camera rotations from two successive
images like feature tracking and optical flow [8]. In hovering flight, given the
slow variation of the yaw angle respect to the frame rate of the camera, it is
possible estimate this angle using images taken by a downward looking camera
attached to the quadrotor chassis. Gyroscopes can be aided with this informa-
tion to obtain drift-free yaw angle estimation. For the case of low altitude indoor
flight, it is hard to find features to track due to the regularity of ground (tiling or
carpet). In this cases, frequency domain representation of the images can detect
displacement of the camera using the Fourier Transform and the so called Cross
Power Spectrum [9]. This method can be used on small portions of the images
to obtain the displacement the so called spectral features to aid gyroscope read-
ings [10]. However, this approach can be computationally expensive and may not
run eciently on a single computer.
 
Search WWH ::




Custom Search