24-hours drone challenge, will they fly?

Control Systems Technology Weekend 2017 Drone challenge | Eindhoven University of Technology

In the weekend of 18 and 19 November, an unfamiliar object has appeared in the large open space of the Auditorium building. A cage, measuring 9 by 9 meters and with a height of 4 meters, is decorated with flickering lights and large nets for protection. In the cage, marginally stable drones are hovering while various sleep deprived students are anxiously watching. Near the entrance to the Blauwe Zaal, two large screens show a countdown timer. A new sight, but a returning one: these students are participating in the very first Control Systems Technology (CST) Challenge. In 24 hours, the freshly enrolled students in the CST group have to assemble a drone, and design stabilizing algorithms such that the machines can maneuver around the obstacles in the cage autonomously.

Bonding for life
While this may seem like a trivial task, controlling a highly sensitive, 6 degrees-of-freedom machine is not easy. Luckily, various PhD candidates and staff members are around to assist in the process. The aim of the project is not only to finish the challenge, and experience some hands-on control, but also to bring the students together: foreign and Dutch alike. As an introduction, presentations by the head of the group Maarten Steinbuch, and drone company Avular have given a glimpse of the huge field that is high-tech control. As far as we know, such a commitment to early master students is pretty unique on campus. But precisely this promise of bonding and introduction to the group is the main motivation for the organizing committee, consisting of four PhD students assisted by four student-assistants. A weekend like this in an early stage of the master program helps to place the coming courses into context, makes it easier to form groups for coming projects, and allows for a much broader view in selecting internships and graduation projects in the second year. And yes, for that solving a difficult challenge is key. Slowly opening Pandora’s box and tackling the evil drone spirits one by one appears to be a recipe for group bonding for life.

A few more words on this Pandora’s box: why is it so challenging to make a drone do what you want autonomously? Key here are the dynamics of a drone in flight. Our drone has four actuators (the motors with propellers) to control 6 degrees-of-freedom (or DOFs, namely position in the (x,y,z)-plane and rotation around three axes). In practice, a so-called decoupling is used to ‘decouple’ the actuators from the movement of the drone, such that a 1 DOF input signal to the drone also results in movement in only 1 DOF. However, the rotation matrix (Google that if necessary) is nonlinearly dependent on the tilt of the drone. That makes the decoupling very hard: it changes when the drone moves around. As such, selecting the proper actuation to achieve a certain flight path is not trivial. Then there is the difficulty of hovering: when the drone is close to the ground, the air is locally compressed forming a cushion under the machine. This effect is obviously not affecting the drone when its higher above the ground: again this makes the hovering dynamics position dependent, calling for a different control.

Protein bars and extra points
To relax the mind a little, a dinner quiz on Saturday evening questions the students about the CST group, research, and the world. The winners are supplied with protein bars, possibly making a difference later on. Several glasses of beer are poured, motivating the students for the long night ahead in which the real programming has to be performed. Luckily for them, ‘Sinterklaas’ has dropped by with a big hint which makes stabilizing the drones at least seem doable. However, even with this hint, two groups are rewarded in the morning with extra points for coding the whole night straight.

Then, precisely at noon on Sunday, the L(A)UNCH event finalizes the challenge. In the end, all groups are able to at least stabilize the drone in the vertical direction. To compensate for glitches in the position estimates, the groups are allowed two attempts. One team manages to show the jury a horizontal circle trajectory, and another team even tries to use the provided obstacles. Among the other trajectories were figures of 8 and squares. Considering these achievements and the group bonding that was, without any doubt, part of the process, the CST weekend is definitely worth repeating!

Also check out this brief video report by TU/e Cursor.