The Historic Guide to Quadcopters (Build to learn the art)

Hi Guys,

I along with two of my teammates (Marco and Luc) at Eindhoven University of Technology, Netherlands had worked on building our first quadcopter from scratch around 2 years ago.

It was the best thing that happened to me and since then my life has been around understanding the behavior of multicopters and working with them almost every day of my life!!

My report was lying dormant in the shelf without much use. I saw that there was quite some interest at our university for it and since 2 years have passed I am allowed to make it public.

Here it what we did and documented everything 'step by step':

1) We did not know then, any (stable) open source so the only solution what get going ourselves. From accelerometers, gyros plucked from different boards and soldered on PCB to make it work. Wrote device drivers for it.
4) We wrote Kalman filtering, understood DCM algorithm and Quaternion to test sensor fusion. DCM and quaternion were both ported to Beaglebone and MBED microcontrollers.
5) Wrote our Wifi-console-GCS software, accelerometer and gyro calibration using Matrices/ offset computation. Thanks to my great teammates Marco and Luc (the real dutch guys :)
6) Autotuning was a great idea with step response, I thought of it and we did succeed! I am sorry Arducopter, but we were first in that :)
7) One of us attempted to create our own 4 in1 ESCs and flash it with firmware ourselves. We succeeded though with some final ESC response issues!!
8) Everything was done on Linux Beagle bone and then MBED microcontroller.
9) Concept of ball detection and tracking. Image processing using OpenCV :)
9) No GPS, that is why it is historic knowledge!

The result was a self developed software with a semi stable system (semi because there was some drift but it was still responsive and glad that we came so close). It was never published (maybe we were lazy). Link to our historic page, here and an old video here!!

Software/material for your reference will be available, on request. If you are passionate about this art, this will serve as a simple useful guide and history will repeat itself :)

Link to full report here

Enjoy,

Shyam.

Thanks Shyam, Very Nice Work especially for just a 2 quarter class.

I know a little of the history of quads / multicopters, and you are absolutely right, a lot of it is repeating history. For instance the first Quadcopters were made in 1904 (believe it or not, though they were big, flimsey, held together by wire, had funky people sitting on them with death defying flimsey motors and rotors, did not fly except a couple feet above ground, then would crash - hey they still do that a lot).

First comprehensive flight control theory that I saw published on quads (they called them quadrocopters) starting in 2004, Robert Mahoney, Univ of Sidney. He and almost everyone else were using the Draganflyer quads and they did not have much payload capacity, were poorly documented, had very limited control systems, etc, Contacting Draganflyer to get any information on the circuits, software, etc. was useless because they would say "proprietary". But they were really cool and so everyone wanted a better, more payload, more stable one to take aerial photos or videos with, and everyone wanted GPS way point control, because only the Canadians or Germans could fly them well back in 2004 to 2005 (just kidding). Quads exploded on the scene in Germany from 2005 to present and a lot of published work was done in Europe, then finally about 2006 some universities in the US started publishing some decent stuff on quads. The worlds best designed navigation systems for quads were actually in Germany back then, published by the group out of Munich Univ. (Ascending Technologies, GMBH, they teamed with MIT and won several years in a row some international competitions in autonomous robotics using mostly quads) and their multicopters are still really awesome, used by many university researchers. If I had enough money, they would be my first choice for an aerial robotics video platform (no offense to all you other guys with such great systems).

Many university researchers looking at aerial robotics in 2004 to 2008 published on the quads control theory, stability, flight mechanics, sensors, as the MEMs accelerometers, GPS, and Gyros, etc. etc. were getting better and smaller, etc. of quads was into non-linear control theory, linearization of flight envelopes, all kinds of PID schemes. Samir Bouabdallah (EPFL) published the best work in about 2005 to 2007, along with girls and guys at MIT, Carnegie Mellon, Berkely, Stanford Univ, Univ of Penn, So. Dakota School of Mines, Univ of Munich, others at Swiss Federal Inst of Tech (EPFL), and many other universities around the world, have done similar work starting around 2004 to present; but I do not recall seen much with the BeagleBone.

DIY Drones entered the scene in 2007, and what Chris Anderson and his team did was amazing, just using the Arduino. No other team of developers has done so much to make this technology so accessible and inexpensive to so many. DIY Drone open source effort is incredible, compared with what was available until 2007.

About the same time (Oct-Nov 2007) a big IROS IEEE (1000 plus attendees) conference in San Diego was held, featuring many prominent aerial robotics researchers using quads. That seemed to be the turning point for less expensive and capable autopilots. Before that conference a dedicated student or hobbiest could at least keep up with the published work and count the number of vendors / developers of quad / multicopter aerial robotics. By 2008, that was just not possible, because so many people started developing and selling their quads, multicopters, flight controllers, auto-pilots, frames, FPV systems, BOOM - now the sky is the limit.

Most of the research published were done as components of PhD and Master's theses. The best optical navigation theses that I have seen published were by Roland Sigwart's (EPFL) students, guys working with Samir Bouabdallah (now teaching) and Andre Noth, and others after them starting around 2005 also. The best published aggressive quad maneuver algorithms and best stability algorithms are by Univ of Penn (i.e. especially Daniel Mellinger, Alex Kushleyev, and Vijay Kumar at the GRASP Lab) - they have pushed the envelope far beyond others to the point that it is almost unbelievable since about 2010. Many researchers have demonstrated quads with visual navigation, GPS stabilization, auto-tuning routines, SLAM for object avoidance, etc., but I do not think an open source i.e. DIY Drones etc., developers have caught up to those efforts yet, but they are amazingly closer every day. Multicopter systems that cost many thousands of dollars employ some of the most recently published systems for automomous navigation in indoor and outdoor environments, but DIY Drones (Jdrones, 3DR, etc.) are right there too.

I just read and follow some of it and crash my multicopters then re-build, etc. Maybe Open Pilot (Revolution) has autotuning too, but I have not followed that much. Anyway I am sure some-one of the 40,000 plus members of DIY Drones knows a lot more about this than I ever will and hopefully will send you some follow-up info.

Replies

W. Joe Taylor November 20, 2013 at 10:36pm

Thanks Shyam, Very Nice Work especially for just a 2 quarter class.

I know a little of the history of quads / multicopters, and you are absolutely right, a lot of it is repeating history. For instance the first Quadcopters were made in 1904 (believe it or not, though they were big, flimsey, held together by wire, had funky people sitting on them with death defying flimsey motors and rotors, did not fly except a couple feet above ground, then would crash - hey they still do that a lot).

First comprehensive flight control theory that I saw published on quads (they called them quadrocopters) starting in 2004, Robert Mahoney, Univ of Sidney. He and almost everyone else were using the Draganflyer quads and they did not have much payload capacity, were poorly documented, had very limited control systems, etc, Contacting Draganflyer to get any information on the circuits, software, etc. was useless because they would say "proprietary". But they were really cool and so everyone wanted a better, more payload, more stable one to take aerial photos or videos with, and everyone wanted GPS way point control, because only the Canadians or Germans could fly them well back in 2004 to 2005 (just kidding). Quads exploded on the scene in Germany from 2005 to present and a lot of published work was done in Europe, then finally about 2006 some universities in the US started publishing some decent stuff on quads. The worlds best designed navigation systems for quads were actually in Germany back then, published by the group out of Munich Univ. (Ascending Technologies, GMBH, they teamed with MIT and won several years in a row some international competitions in autonomous robotics using mostly quads) and their multicopters are still really awesome, used by many university researchers. If I had enough money, they would be my first choice for an aerial robotics video platform (no offense to all you other guys with such great systems).

Many university researchers looking at aerial robotics in 2004 to 2008 published on the quads control theory, stability, flight mechanics, sensors, as the MEMs accelerometers, GPS, and Gyros, etc. etc. were getting better and smaller, etc. of quads was into non-linear control theory, linearization of flight envelopes, all kinds of PID schemes. Samir Bouabdallah (EPFL) published the best work in about 2005 to 2007, along with girls and guys at MIT, Carnegie Mellon, Berkely, Stanford Univ, Univ of Penn, So. Dakota School of Mines, Univ of Munich, others at Swiss Federal Inst of Tech (EPFL), and many other universities around the world, have done similar work starting around 2004 to present; but I do not recall seen much with the BeagleBone.

DIY Drones entered the scene in 2007, and what Chris Anderson and his team did was amazing, just using the Arduino. No other team of developers has done so much to make this technology so accessible and inexpensive to so many. DIY Drone open source effort is incredible, compared with what was available until 2007.

About the same time (Oct-Nov 2007) a big IROS IEEE (1000 plus attendees) conference in San Diego was held, featuring many prominent aerial robotics researchers using quads. That seemed to be the turning point for less expensive and capable autopilots. Before that conference a dedicated student or hobbiest could at least keep up with the published work and count the number of vendors / developers of quad / multicopter aerial robotics. By 2008, that was just not possible, because so many people started developing and selling their quads, multicopters, flight controllers, auto-pilots, frames, FPV systems, BOOM - now the sky is the limit.

Most of the research published were done as components of PhD and Master's theses. The best optical navigation theses that I have seen published were by Roland Sigwart's (EPFL) students, guys working with Samir Bouabdallah (now teaching) and Andre Noth, and others after them starting around 2005 also. The best published aggressive quad maneuver algorithms and best stability algorithms are by Univ of Penn (i.e. especially Daniel Mellinger, Alex Kushleyev, and Vijay Kumar at the GRASP Lab) - they have pushed the envelope far beyond others to the point that it is almost unbelievable since about 2010. Many researchers have demonstrated quads with visual navigation, GPS stabilization, auto-tuning routines, SLAM for object avoidance, etc., but I do not think an open source i.e. DIY Drones etc., developers have caught up to those efforts yet, but they are amazingly closer every day. Multicopter systems that cost many thousands of dollars employ some of the most recently published systems for automomous navigation in indoor and outdoor environments, but DIY Drones (Jdrones, 3DR, etc.) are right there too.

I just read and follow some of it and crash my multicopters then re-build, etc. Maybe Open Pilot (Revolution) has autotuning too, but I have not followed that much. Anyway I am sure some-one of the 40,000 plus members of DIY Drones knows a lot more about this than I ever will and hopefully will send you some follow-up info.

This reply was deleted.

ArduCopter User Group

Replies