Abstract
—MARVIN is an autonomously ying robot based
on a model helicopter. It has been designed for participation
in the International Aerial Robotics Competition (IARC) Millennial
Event 1998 – 2000. The competition task consists of
a search operation in an unknown environment with possible
threats such as res, water fountains, and smoke. For autonomous
ight, a ight control algorithm runs on the on-board
computer that stabilizes the helicopter and enables it to reach
a way-point issued by the ground station. The ight control algorithm
consists of a hierarchy of modied PID controllers. By
MARVIN's performance, it has been proven that no more sophisticated
approach to control is required for helicopter ight
control.
I. INTRODUCTION
MARVIN is an abbreviation for Multi-purpose
Aerial Robot Vehicle with Intelligent Navigation.
MARVIN is an autonomously ying robot that can
fulll a complicated search mission purely on the basis
of sensor data, without any human interaction.
The requirements to be fullled by MARVIN are
dened by the mission of the International Aerial
Robotics Competition Millennial Event, as described
in [1]. The mission task consists of nding and classifying
hazards and victims in a simulated disaster
area. There are black drums containing dangerous
materials, with the contents distinguishable by symbols
on the drums' surfaces, and there are res, water
fountains, and smoke threatening the operation of the
robot. Victims may be dead or injured persons, the
survivors recognizable by motion and sound. In the
2000 competition nale, MARVIN was the only participant
to recognize target objects during autonomous
ight. Thus, it won the Millennial Event.
This paper deals with the approach to and the development
of the ight control algorithm for MARVIN.
This ight control algorithm is responsible of
ight stabilization of the inherently unstable air vehicle
and of reaching a single target way-point issued
by the ground station. This means that mission strategy
is not part of “ight control” and consequently not
covered here. Please refer to [2], [3] for details on the
higher levels of control. The following section of this
paper gives a short introduction to the overall system.
Then, the approach to ight control is explained. After
some details about the implementation and the testing
process, the performance of MARVIN's ight control
is described.
Fig. 1. MARVIN in ight
II. SYSTEM DESIGN
Figure 2 provides an overviewof the MARVIN system.
During system development, every detail has
been carefully optimized with respect to weight, prize,
power consumption, and capability. Much care has
been taken to select components that fulll the respective
tasks and integrate smoothly into the system, always
avoiding things being “better than needed” at the
expense of some of the criteria mentioned above. Basically
this made MARVIN's success possible.
A. Air Vehicle
The basis of the MARVIN air vehicle is a conventional
model helicopter. It has a rotor diameter of
1.9 m and is equipped with a two-stroke chain saw
petrol engine producing about 2 kW. Its takeoff weight
amounts to about 11 kg, operation time is approximately
30 minutes. Figure 1 shows a view of MARVIN
in ight.
B. Sensors
The sensors on board MARVIN are:
 A hand-crafted inertial measurement unit consisting
of three magnetic eld sensors, three semiconductor
acceleration sensors, and three piezo-electric rotation
sensors (material cost below $250).
 A ame sensor that detects certain ultraviolet light
characteristic of burning wood, gas, or oil.
 A light barrier RPM sensor for measuring the main
rotor RPM.

Download full report
http://docstocdocs/47008905/Development-of-a-Flight-Control-Algorithm-for-the-Autonomously---rossir24-0004