1. How did you come up with the team name “TeamDARE Eindhoven”?
Our team is called “TeamDARE” which stands for “Daring Autonomous Robots Eindhoven”. We needed a team name when participating in Eurobot in 2005. The name sticked since then.
2. Please give your name and any professional designation and tell me what each of your engineering specialties are and where and how did you each gain them?
TeamDARE consists of 9 experts from various high tech companies and institutes in Eindhoven or her direct vicinity, including ASML, Philips, Prodrive, CCM, DAF Trucks, The Mathworks, and the University of Technology Eindhoven. The expertise that each brings in for the team comes from our jobs at these companies. Although, it also works the other way around. The practical knowledge from TeamDARE is also applicable in our daily work in the high tech industry.
3. What is the TU/e?
It is the (Dutch) abbreviation of the University of Technology Eindhoven
(Technische Universiteit/eindhoven). Two of our members work at the mechanics department of this university. One as the project manager for robotics, the other as a student. All of the team members studied at this university.
4. How did you participate in Eurobot?
Back in 2001, some of the current members of TeamDARE participated in the Eurobot competition as an internship at the TU/e. It was organized by three of its departments: Electrical engineering, Mechanical engineering and Computer Science. Combining these disciplines turned out to be really fun and successful (we finished 2nd).
The second and third time that we participated -in 2005 and 2006- we participated as an independent team and were financially supported mainly by the TU/e, Philips and ASML.
5. What is your industry experience in embedded, intelligent systems? What kinds of systems?
In their daily lives, the members of TeamDARE, work on complex embedded systems such as, for example, the wafer steppers from ASML, MRI scanners from Philips. These systems require high quality standards for mechatronics and software. Also robustness follows from our daily work; a nice example of this is are the high robustness requirements that are required in DAF trucks. They have many moving parts, that always have to work in challenging environment conditions.
Applying such ways of working in our robot project is essential to also make sure the music robots always work straight out of their box, even when set up in stress conditions, i.e. on a dark, loud, and often small podium.
Specifically the expertise we apply comes from the team members working at:
- ASML: the world leader in lithography machines.
- Philips Research: one of the world’s largest corporate research organizations.
- The TU/e: with many experience in robotics projects.
- Prodrive: delivers solutions in electronics design, manufacturing and added value services for OEM's and ODM's.
- CCM: the Centre for Concepts in Mechatronics.
- DAF Trucks: daughter of PACCAR Inc.
- Mathworks: best known from Matlab.
6. Briefly describe some other robots you built with their names.
Before starting the work on the robot band, we participated at Eurobot three times. For this competition a team typically builds two (or back in 2001 multiple) robots that have to complete a task autonomously in 90 seconds against an opponent. The goal of the game is changed each year to make it more easy for “new comers” to come up with a competitive robot. The required autonomy and time limit guarantee that the smartest team has most change of winning.
As mentioned before, we participated at Eurobot in 2001 as an internship at the TU/e. We then built 11 robots that all solved a small part of the problem (in that year, placing flags on cylinders with different height and mobility). Our swarm of robots yielded a 2nd place and the concept was so successful that next year the rules were changed, limiting the number of robots each team could construct.
After this competition, the core members entered the Dutch version of the Robotwars competition with a robot named Cyclone. This was a huge moving cylinder that had a spinning outer shell. This was great fun to built, but not very high tech, as these robots from a technical point of view are not much more complex than a radio controlled car.
So we sent back to Eurobot and participated two more times. Reaching the quarter finals in 2005, and finishing again second in 2006. We stopped participating after three times in search for something completely different.
We then found the Artemis Orchestra Contest in 2008. The contest dictated to make a robot that can play an instrument of choice, without alterning the instrument itself and as part of an orchestra. We decided to compete by building a robot that can play an acoustic guitar. It was built in about four months and we got really enthusiastic responses wherever we played. This was the motivation for us to continue to expand the band, irrespective of the contest (which stopped after 2009).
7. Who were your sponsors for the robot band project? Can you provide links to their websites?
see: http://www.teamdare.nl/ it has a list of our sponsors with links. We have several different types of sponsors. Some provide us the parts we need for construction, a few help out financially, others such as e.g. BocaBearings help with publicity.
8. How did you adhere to simplicity and low cost with the robot band?
Simplicity is not our goal on itself, but more a requirement to end up with a robust solution. As a simple example: a complex system depending on a 100 small components in a cascade, each with 99% robustness, will fail about 63% of the time. Complex systems also take much longer to test and debug.
In order to implement things as simple as possible, we have long discussions at the beginning of each project, trying to see if a certain part of the problem is best solved in software, or using electronics or with a mechanical solution. A typical example is the accurate positioning of an arm. Often accuracy can be increased using all three disciplines and really depend on the specific use case which is yield the simplest result.
Also low cost is not a goal on itself, but a consequence when depending on sponsors and your own budget.
9. Please discuss the circuit design and layout for the robot band.
Each music robot is roughly designed similarly, from a high level circuit design point of view, consisting of three layers: actuation, scheduling and control.
Actuation is the lowest layer. We use microcontrollers (AVR32 family from Atmel) to control motors, and handle sensor information. On this level the controllers have no knowledge of music and only handle simple actuation commands instantly, e.g. move a motor from A to B with a given profile. The intermediate layer does teh scheduling of actuations and also runs on microcontrollers (low-end 8-bit RISC controllers). These controllers receive a lists of music notes for a certain instrument, including how and when each should be played, e.g. exact timing and volume. These lists of notes are translated to commands that can be executed by the low level layer. Each command is sent at their corresponding time. At the highest level, songs are controlled. Each song is decomposed into instruments and for each instrument the song is converted in lists of notes that can be sent to the scheduling controllers. The highest layer is no longer time critical, i.e. can be non-real time and is therefore easily implemented on a PC. This is important to also be able to add a user interface. This layering is important for two reasons. First, to be able to have precise control and actuation, together with a easy user interface that can start, stop, and add songs. Second, to be able to extent the band to an arbitrary number of instruments.
All circuit boards, are designed by ourselves. Some have even been etched by us.
10. Please discuss the design and mechanical and robotic construction of the robot band.
With respect to mechanics, there has been a steady increase in complexity for each new robot that has been constructed. The guitar is the most simple of the three. It consists of six plectra on little arms that each actuate one string. The arms are rotated with a servo motor. The “left hand” of the guitar player consists of a grid of pneumatic pistons that press down the string at the correct fret position.
The volume of each note is controlled by setting a second servo motor that controls the height of the a plectrum with respect to its string.
The drums conceptually seem simple: one stick for each tom, controlled by a servo motor. But in order to get the timing, volume and the bounce of these sticks correct a very complex motion controller was required. Furthermore, the pedal for the hi-hat and the muting of the crash cymbal was implemented using pneumatics.
The pan flute robot is the most complex robot of the three in terms of mechanics. Instead of “just” having a nozzle for each tube, we wanted to use a single pipe and rotate the pan flute back and forth. In order to get the correct tone out of each tube, the nozzle needed to be controlled over three degrees of freedom: horizontally, vertically and one axis of rotation. In addition, the airflow influences the pitch of the sound. To make things really complex, on each tube of the pan flute a semitone can be produced in addition to a note. For this, a tube needs to be partially obstructed. We implemented this using a small rubber cushion that mimics lips and a second air flow supply.
11. Please discuss the software design for the robot band. What programming languages and software did you use?
The whole band currently runs on 8 microcontrollers and two PCs, due to the layered design, mentioned before. Most of the code on the microcontrollers is written in C, with the exception of some small critical parts that are implemented in assembly. The PC software is mostly written in C++. For some nice 3D visualization on a big screen, we have written a player in C++ and GPU code, this runs on the second PC.
In addition, a lot of software has been written to automate the “song writing”, mostly in C and C++. Basically, these tools convert midis to a format suitable for our robots and can simulate how a song will sound when played on the robots.
12. What instruments are included in the robot band?
As mentioned before, the band consists of a guitar, drum set and pan flute. But conceptually, the 3D visualization can be seen as a fourth instrument. Instead of making sound, however, it shows a 3D rendering of the music that is played. We are currently working on a double bass playing robot and have future plans for the saxophone and piano.
13. How did the robot band overcome the instrumental performance capabilities that you lacked?
With science! Playing music in its basic form is about correctly producing notes on instruments. It all depends on the definition of “correctly” to what level of naturalness the robots can play music. Not being able to compose songs ourselves, we depend on the information that can be found in digital tablature, such as midi. These formats typically contain accurate time and volume information for each note, and depending on the instrument type, also effects like changes in pitch and duration.
When the robots reconstruct this digital tablature they play music. Unfortunately, musicians deviate from the written music in all kind of ways when they put a certain emotion in a song. We have not found any means yet to also include these. We are still searching for a scientific description of emotions in music. Who knows this might give our music robots some soul.