Dr. Gilles Labonté, BSc, MSc, PhDProfessor

Department of Mathematics & Computer Science

Office: Girouard 320 

Telephone: 613-541-6000 ext 6093

Fax: 613-384-5792 

E-mail: labonte-g@rmc.ca 

Department of Mathematics & Computer Science

Royal Military College of Canada

PO Box 17000, Station Forces

Kingston, Ontario CANADA

K7K 7B4

Autonomous robot airplane "Harfang-des-Neiges" Project

Artificial neural networks
robot airplane
Control; Sensors: velocity, altitude, roll, pitch.

Aim

Development of artificial neural networks which will take on the sensorial, command and task planning functions required for a robot airplane to become autonomous.  An aim of this project is to explore the limits of abilities of artificial neural networks in this context.  

Phase 1

This project is presently in its first phase.  This consists in

  • creating a prototype from a commercially available airplane, sensors and microprocessors, and
  • developing artificial neural networks which will take on sensorial and command functions which will endow this airplane with the ability to learn to take off, fly and land by itself.
Phase 2
  • Artificial neural networks will be added to the airplane so that it can learn to carry out autonomously a certain number of aerial maneuvers.
  • It should be able to learn to follow a prescribed course, land by itself to refuel, take off again and return to its starting point.
Phase 3
  • The airplane will be endowed with hearing, smell and vision so as to augment its capacity to act autonomously.
  • It will have to learn to report to its home base any anomalous condition it detects, to seek a prescribed object in a designated area, transmit or bring back images of it or bring back the object itself.

Present state of the project

Simulation
  • We have produced a computer program which simulates the behavior of an airplane in its landing phase, in the presence of random winds.
  • This program has a graphic interface which allows a human operator to follow the position and attitude of the airplane so as to be able to learn to land it.
  • We have designed an artificial neural network which was able to learn, through imitation of the human pilot, to succeed in landing the plane.
First real airplane

We have purchased a radio controlled trainer airplane which we are presently modifying to achieve the objectives of Phase 1 of the project.

Our airplane: the Kadet LT-40 ARF is a trainer airplane which we have selected to develop our prototype. 

Sensors

The airplane has been endowed with sensors which will allow it to know

  • its altitude,
  • its velocity with respect to air and
  • its angles of roll and pitch.

These sensors are very robust and similar in their operation to those used in nature by living organisms.  Artificial neural networks will process the data provided by these sensors.

Actions

After the sensors have been duly tested, we will endow the airplane with artificial neural networks which will take over the functions of control and task planing.  We will either teach it, or it will be able to learn by itself, to take off, to fly while controlling its attitude and to land by itself.

Participants in the project

4th year students participate in the development of this robot airplane, as part of their last year project.  Here is the list of those who are contributing or have contributed until now.  

Students Contribution

François Allaire (2000) and Patrice Cayouette (2000)

Dept. of Electrical Engineering and of Computer Engineering

Study of the material required for the realization of the first prototype with a real airplane. 

Selection of the airplane type, the sensors and the microprocessors.

Design of the architecture of the global system: airplane, its sensors and controllers. 

Realization of this system and tests of its performances.

Jean-François Latreille (1999)

Dept. of Mathematics and Computer Science

Development of the graphical interface and the simulation software for the landing of the airplane. 

Study of an artificial neural network which can land the computer simulated airplane

Neil Trask (2002) of the Department of Mathematics and Computer Science is an expert pilot for radio controlled airplanes.  He provides freely his expertise and serves as trainer for our pilots.

General information

There is no doubt that aerobots (flying robots) are destined to a bright future.  Some projects similar to ours exist elsewhere.  Until recently, most UAV (Unmanned Aerial Vehicles) were constructed at great expenses for very specific needs.  However, before long, aerobots will be produced commercially and used commonly to accomplish multiple tasks.

Some possible uses

Some of the links below lead to a site belonging to an entity not subject to the Official Languages Act. Information on this site is available in the language of the site.

Planet Exploration

In its year 2000 budget, NASA has included $50 millions for the development of an airplane for the exploration of the planet Mars.  This will be a small unmanned airplane which will be launched from orbit around the planet.  It will then opens its wings and fly in the tenuous atmosphere of Mars.  It will explore the Valles Marineris at an altitude of many thousand feet, on the 17 December 2003, exactly one hundred years after the historic flight of the Wright brothers.  Cameras placed on board will photograph the canyon.  Such an airplane has to be autonomous because of the too long transmission delay between the arrival of the sensor data to earth and the return of the control signal to Mars.  See the Nebweb site or the NASA site

Search and Rescue

A robot plane can rapidly and systematically explore a large area to find victims of an accident or a natural disaster.  This would allow humans to consecrate themselves to helping the victims instead of spending much time seeking for them.  Such airplanes could fly in conditions which would be too dangerous for humans, such as when chemical or nuclear contamination occurs.  

Aerial Surveillance

Such airplanes would be precious tools for the inspection of gas or oil pipelines, of electrical transmission lines, of bridges, of roads after an earthquake, of large spaces...  They could then fly essentially without interruption, landing at fueling stations on their way, and taking off again, all by themselves.

Weather Prediction

The University of Washington is presently improving UAV's which could crisscross the Pacific Ocean to gather the data that is required for the prediction of weather on the west coast of America.    This same university has achieved, on the 21 August 1999, the first crossing of the Atlantic by a robot airplane robot (see University of Washington and Robotic Airplane Completes Historic First Trans-Atlantic Flight). This airplane is the first UAV and the smallest airplane to ever cross the 2000 miles between Newfoundland and Scotland.

Some relevant Internet sites