Spy in a backpack

It must be able to fly in a swarm at great speed, but also to land on a ledge to survey the surroundings for hours at a time. The Atmov is the latest gadget from TU Delft’s MAV-lab.

Check telemetry, check altitude estimation, check waypoint…” For a moment, on a strip of grass at the military training ground of Fort Stewart in the American state of Georgia, nothing else can be heard other than a tinny robotic voice from a computer. AE student Dieter Castelein laughs. “The funny little voice is just a bit of fun”, he says. The checklist itself is quite a different matter: forget something and the small aircraft Atmov (autonomous transition multi-rotor observation vehicle) will crash.

Atmov is the latest technical creation from the MAV-lab, the micro aerial vehicle lab, which is located at the faculty of Aerospace Engineering. PhD candidates and students work on aircraft that can fly autonomously, recognise and avoid obstacles and produce images of the surrounding area.

Castelein shows a short film that reveals how everything is controlled in minute detail. The Atmov takes off vertically with its wing raised. At a height of a few metres it rotates forward ninety degrees and darts off horizontally. It does so almost silently, like a dragonfly.

The aircraft has to fly three kilometres over a wood and land in a fake village where soldiers imitate an Afghan scene. For three hours, it has to stand in front of the church, just beyond the hotel and the football ground, observing – making video recordings – before flying back again.

That was the assignment given by DARPA (the research institute of the US army) to nine AE Bachelor’s students from TU Delft together with nine other teams. Last May, they had two weeks to demonstrate they had developed the ultimate small spy plane. An aircraft that would fit in a backpack.

The majority of machines taking part are of the quadcoptertype, an aircraft with four motors that can take off vertically. This includes the Atmov. But whereas the other quadcopters are only able to hover and slowly travel in this mode to their final destination, the Atmov does something special: it first hovers upwards before flying off like an aeroplane.

“This allows us to fly much more efficiently and quicker”, explains Castelein’s colleague Sander Hulsman.“The other aircraft fly to the village at twenty kilometres an hour, whereas we fly at seventy kilometres an hour. We are also able to stay up for much longer.”

Duct tape

The students watch elatedly how the aircraft disappears from view. It is their third attempt. The aircraft appears to be well programmed, flying to the village completely unassisted. “Still flying”, “doing good” are the occasional messages coming through the walkie-talkie that is used to keep in contact with the tracking car.

At the church, the Atmov crashes. The students think that the motors became overheated. The aircraft is lying in pieces, but the TU Delft team have brought four aircraft along and plenty of duct tape. The students are happy. No other team has successfully reached the village unassisted until now.

Some participants hardly manage to reach the wood, like the team from Singapore with its aircraft with parallel counter rotors. “We call it the flying lawnmower”, laughs Hulsman.

“That machine was well made, but very scary”, adds Hulsman. “Everyone took a step backwards when it took off. It weighed some five kilos and made a racket just like a real aeroplane. The pilot had beads of sweat on his forehead trying to keep it under control. But it crashed almost immediately, because the rotors weren’t correctly aligned and they struck each other.”

The greatest challenge for the majority of teams is communicating with the aircraft. Contact with the ground station is crucial even for many of the autonomously operating aircraft (or rather semi-autonomous aircraft), because flight calculations are made by a computer on the ground. But most of the participants lose contact due to the trees blocking the way. One aircraft after another ends up crashing.

Atmov, on the other hand, does fly completely autonomously and is therefore unaffected by the trees. What does trouble the TU Delft students is the calibration of the motors.

The aircraft crashes once again in the village on the final day. Another Atmov that the students sent up simultaneously – somewhat as a joke – disappears in a lake and has to be fished out by a fisherman. The only explanation for the crash the boys can think of is that it had an argument with a vulture.

The TU delft students finish equal third. The winner is a team from Middlesex University with an aircraft with six rotor blades.


Bart Remes, Project Manager of the MAV-lab, has great expectations for the Atmov, especially for projects in the open air.

The most well-known aircraft from the lab is the wing-flapping micro aircraft Delfly. Work on Delfly started in 2005. It has become so miniaturised over the years that it now fits inside a matchbox. It can fly through small openings and therefore get to places where no one or thing can reach. Useful for searching for victims in collapsed buildings, for example.

In the open air, however, the Delfly is not in its element, as it cannot deal with wind. In such cases, an aircraft like Atmov offers more possibilities. In time, Remes hopes to use this aircraft’s flying technique in many of his projects.

“Atmov is the future. You can station these aircraft at a loading station and with a push of a button let them take off as a swarm to carry out reconnaissance flights. The aircraft that are presently being experimented with for swarm flights have to be individually catapulted into the air with elastic.”

One of the running ‘swarm projects’ Remes is referring to is the FireSwarm project. The idea behind it is that a group of aircraft equipped with heat sensors go in search of heathland fires. Remes: “The aircraft know each other’s position and agree upon the direction each will take to investigate further. Initial tests have recently been carried out with the Woensdrecht fire brigade.”

Remes wants to utilise the same technique to deploy a swarm of autonomous aircraft to monitor ships in the neighbourhood of ports. The detection of illegal dumping of oil is one of the important objectives of this so-called 3I-project (an acronym for Integrated Coastal Zone Management via Increased situational awareness through Innovations on Unmanned Aircraft Systems).

This is a European research project being coordinated by Remes and his colleague DrErik-Jan van Kampen, in which those involved besides TU Delft include the French, British and Dutch port authorities and the universities of Southampton and Brest. Earlier this year, the consortium received 1.85 million euros from the EU.

Therefore, there are many possible applications for Atmov. But for now, it is contending with teething problems. This in itself is hardly surprising. The students only came up with Atmov at the beginning of the year and worked out the concept as part of the so-called AE Design Synthesis exercise. Over the course of ten weeks, students have to demonstrate in this exercise that they are capable of applying their knowledge as a team to create a design. The Delfly also resulted from such a student project.

The clever thing about Atmov is that it can turn ninety degrees in the air fully automatically: a meticulous manoeuvre that involves gradually changing from a hovering position to a horizontal flying position. The hovering position uses a different set of propellers than it uses for horizontal flight.

Castelein: “There are more aircraft that can simultaneously hover and fly horizontally. But they do it far less efficiently. With these aircraft it’s actually the motors that turn ninety degrees. That’s not very practical, seeing that hovering requires a different type of propeller than horizontal flight, due to the fact that airflow speed varies greatly in both cases.”

New mathematics

In order to enable the aircraft to make the transition from one propulsion program to the other, the students had to develop a completely new automatic pilot.

The standard mathematical method behind autopilots makes use of so-called Euler angles. Hulsman: “If your aircraft makes a rotation of up to ninety degrees in relation to its starting position, then Euler angles work just fine. At ninety degrees, however, you find yourself in the so-called singularity point.”

In the singularity point, the axes are no longer well defined. Forward flight and backward flight are then written in the same way. Castelein: “If that point is reached, your aircraft will go completely crazy.”

The students had to employ a different mathematical method: the quaternions. “Through this method, you can turn 360 degrees without any problem”, says Castelein. “However, quaternions have never been used before in the open source autopilot paparazzi (the type of autopilot used by the MAV-lab, ed.). We had to rewrite everything.”

In coming years, Remes wants to use the knowledge gained with other aircraft to improve Atmov. He has submitted a grant application to STW. The project has to develop, just as it did seven years ago with Delfly. Today, five PhD candidates and a dozen students are working on Delfly.

What kind of improvements should you be thinking of? The further miniaturisation of the electronics, for example. To illustrate this, Remes shows the Delfly’s IMU (inertial measurement unit), the system that measures velocity, orientation and gravitational forces. It is hardly a centimetre in size.

Most of all, the students themselves would like to see improvements in aerodynamic design, the landing gear, the autopilot software, as well as a lighter composite construction.And who knows, perhaps Atmov can learn from Delfly how to react when it encounters an obstacle, such as a vulture.

Redacteur Tomas van Dijk

Heb je een vraag of opmerking over dit artikel?

Comments are closed.