How did you come up with initial robot concept?

When we started talking about the contest back in January, I had the basic idea in my head fairly quickly: something that would make contact with the plate to stabilize the drill. The main inspiration was taken from the intuitive experience that it is hard to hold a position with an outstretched arm. Humans solve this problem by resting their hand when performing precision tasks such as writing, so we carried this idea over to the robot.

In mechanical terms, resting your hand reduces the amount of load-bearing structure between your tool and the workpiece. The kinematic chain is shorter and the flow of forces more direct. As such, there are fewer things that can vibrate and the stiffness is higher. And we knew that high stiffness would be key to drilling quality holes.

Further, to keep with the analogy, resting your hand while writing almost removes your arm from the process. You write with your hand, and not your arm. Similarly, in our solution the function “Drilling” would be fulfilled by the end effector, and the function “Positioning” only by the arm. If there is one thing that remains firmly hammered into my head after my engineering classes, it’s separation of function.

With our design goal formulated, we soon converged on the three-pronged frame with a separate pneumatic actuator. Instead of having all the motors in the robot arm actively compensate for the forces from the drilling, the end effector would transmit them right back to the workpiece and the arm would do nothing but move the end effector and hold it in place.

In our final design, the robot applies force during the drilling process to keep the end effector stationary. Ideally though, the robot would only be responsible for positioning. We were planning to take the next step to transfer the load off the robot and onto the end effector and the workpiece, but sadly, could not get the necessary parts to Japan in time, and simplified the idea in late April.

Everything else was mostly decided in order of simplicity. We already had the robot arm and small powerful cameras in the lab, so it was clear we would use that. A lot of the code originally comes from our research projects, that we tried to stay close to (for example, the vision system is repurposed eye-tracking code).

As an anecdote: another of the original ideas was to fine-tune the position of the drill by using the end effector. We were basically thinking of using a whole robot hand to adjust the position of the drill. However, unless the robot cannot ensure the positioning on its own, there is no advantage to adding that functionality to the end effector, so we dropped it in favor of the fixed rods with springs (rubber) on the feet. Just another lesson to be learned from Separation Of Function, the benevolent goddess of good design!

What sort of prep work went into the project prior to the competition?

After fleshing out the main idea between January and March, we decided to go for it and write a solid application. We spent a long weekend 3D printing a proof-of-concept, cutting together this flashy video and writing our application in research paper format — since that’s what grad students do. When we got the OK from Airbus to compete on March 27, I doubt we realized how much time would be spent on the project until we started our weekly meetings with our Airbus contact.

It turned out that our biggest problem was logistics. I would estimate a good 40% of our time was spent calling the airport, customs, shipping companies, university offices, and insurance companies for quotes and details about how to get our robot from Japan to Sweden. Shipping would take a whole week, a visiting PhD student needed the robot for his experiments, and with us having to work with the robot before the contest, we had to think creatively and find a workable solution fast.

The biggest issue was that our robot weighed 31 kg on its own, and no single piece of luggage can be over 32 kg in European airports. This left us just 1 kg to package the robot safely. In the end, we made a custom box by cutting a robot-shaped hole into boards of styrofoam with a soldering iron and gluing them together. We even sewed a bag for it.

Teammates Gustavo and Lotfi with the robot and resourceful packaging. Photo: Felix von Drigalski

This is a good moment to remember that this robot arm is easily worth $80,000!

With all the logistics issues, long delivery times, and Golden Week (a string of closely-packed public holidays in Japan) delaying them even more, the time we had to prepare the solution got shorter and shorter. We had parts arriving until the last week before we left, so we had to put everything together in a hurry. We ran the first full trial before the competition on Saturday morning, the day of our flight! It was ridiculously last minute.

Thinking of prep work: we needed our robot at different heights when the plate was inclined, but I designed our stand for the wrong height! I spent almost two months under the completely unfounded misconception that the plate would be inclined by 15° instead of 30°. This is why our stand looks completely different each round and why we had a new mechanical design problem to solve each time. As if that weren’t enough, we misread the distance between the rails and were off by 50 mm. This is why the stand always looks like a mess, on top of everything else!

What was it like competing with your robot? What have you learned in the manic process? Anything you’d do differently?

When we arrived with our limited preparation on Monday to set up the robot and saw the other teams, we didn’t have high hopes. Everyone looked so well-prepared and primed to win. At that point, we just wanted to deliver something solid and keep our heads held high. We freaked out at the thought of not making it through the demo round, so we dragged the robot back to the hotel and stayed up all night to make sure we would get into the contest.

After passing the demo round (with a bug that inverted our hole pattern; look closely at the interview video), we were so happy that we were actually drilling real holes in a real plate that we just kept drilling all through the first round, almost blind. We didn’t have a good way to recalibrate on the fly that day, so we had to accept that our positioning would be off and hope for the best. In fact, our holes were so far off the mark that two thirds of them gave us negative points and we got a terrible score. However, one of the judges said that the calibration was basically the only problem we had — the holes themselves were “absolutely perfect”. That made us perk up.

We had a look at the other plates. Everyone was having more problems with the drilling than we thought. Most were applying force on the drill by moving their robot arm, and the quality of their holes were suffering because of it. Our drilling process was by far the fastest and cleanest. My design worked. It really was only a problem of calibration and our code. That was the moment we gained hope and decided to take the robot back to the hotel for another grueling night of work.

The next day, we paid for our sleep deprivation: we spent the first half hour of round 2 with the robot not moving at all, until we figured out that we never put all the things we fixed during the night into our live code. After we finally did and finished the round with barely half the plate’s holes drilled, we thought we may well have missed our shot at entering the final. However, we absolutely wanted to prove to ourselves that we could at least properly drill a whole set and take a picture of it home, so we were hoping for a chance to maybe perform on a plate separately from the contest, if it came to that. When all the team leaders were informed that the last round would be between 4 teams and we were in it, I was the first to rush out of the room and prepare (and accidentally spill the news to the neighboring and positively amazing team from India, who were the only ones to drop out that round).

In the end, we had more arcane bugs to sort out during the final and we didn’t get to drill through the entire hole pattern like we hoped. The first 10 minutes we weren’t drilling anything at all, and it was only during the last 20 minutes that we really were on point and drilling at (almost) full speed. It was also the first time I had the time to stand back and talk to spectators about what we were doing. Which is when our drill bit got sucked into the clamp with 3 minutes on the clock. Ah well, things never go as planned. We still managed to drill three more holes, although sadly not the damaged one the bit got stuck on. All of that stress fell off our shoulders when the round was called, and we could turn our attention to celebrate everything and everyone who powered through along with us.

All in all, the competition was a blast! Constant adrenaline in an incredible atmosphere. Probably the best thing was that there was not a shred of negativity anywhere, all of the teams were supportive of everyone’s efforts. Team Sirado even helped us troubleshoot KUKA software we had been struggling to install on our virtual machine the day before the competition! Everyone was simply great! Airbus made an excellent choice putting the celebration right after the final round instead of the awards. That allowed us to celebrate everyone’s efforts rather than our scores.

It was also an incomparable team-building experience. Although we had some stressful moments when our robot did not run in the second round, no one cracked or became emotional or accusatory — everyone stayed laser sharp and focused on solving the problem. We could not have done it without a heavy dose of team spirit.

In all that chaos, we learned our lessons about project planning and software development. Probably the worst mistake we made was working on the same code on multiple machines, with multiple sleep deprived people during the night, then trying to merge all those changes. To illustrate: There is a comment in our code history from 6 AM on Tuesday morning that just reads “only a miracle can save us” (the first day of the contest!!), and our live branch is called “last-minute-firebrigade”. You can imagine the state we were in!

Photo: Felix von Drigalski

What advice would you give to any future robotic challengers out there?

I relay these from the team:

• Do not underestimate logistics!
• Don’t rely exclusively on software, robust hardware design is key!
• Just try and have fun, don’t think it’s impossible without even trying!

And I would stress how much trying is worth it, no matter what. Winning is great and all, but it was just after the final, after close to 60 hours with almost no sleep and all of our holes drilled, that we felt the most relieved and accomplished. We didn’t spend a second thinking about scores or placements — we were just proud that we made something that worked, and we didn’t give up. Thinking up and developing your own solution, working together and seeing it come to life is one of the most satisfying and fulfilling experiences. Making your performance and efforts your motivation, rather than winning, will keep you going when things look dire.

Any closing words?

After the event, we kept getting amused comments about not having Japanese members in our Japanese team. It’s true we come from Germany, Mexico, Belgium and Ecuador, but studying and living in Japan has changed all of us in ways that we sometimes don’t even realize, as anyone who has left home to go abroad understands. As such, we have always said that our team consists of 4 people with 5 nationalities — just no one with a Japanese passport.

So, for the big closing words: If you’re an aspiring youngster reading this and you want to see the world, or if you’re in your undergrad and thinking of going on an exchange, or if you dream of living somewhere else — go for it! Every journey starts with the first step, and it is worth it.

From left to right: Felix von Drigalski, Lotfi El Hafi, Pedro Uriguen, and Gustavo Garcia.

Like this story? Read more about the Airbus Shopfloor Challenge here.

3… 2… 1… drill! Who would have thought that drilling could be so exciting. At the Airbus Shopfloor Challenge last week in Stockholm, 7 teams from around the world competed over 4 rounds of intense drilling. The event was part of the annual International Conference on Robotics and Automation.

The goal of the competition was to build a robot that could drill as many holes as possible in one hour, in the right position, and with the right diameter. Robots were required to be light, modular, and open.

In what turned out to be a suspense-loaded two days, spectators saw the whole gamut of emotions driven by catastrophic failures, live coding of new solutions, sheer joy when the drills powered through, and the exhaustion of pulling all-nighters to compete in what was a very tight race. Above all, it was a chance to see teamwork, innovation, and creativity in action. All the solutions and approaches were profoundly different, and many of them worked really well, although not perfectly. That gave rise to an ideal level of competition: the task was hard enough that no one was perfectly good at it, but it was feasible and teams were able to finish 1-hour rounds with a nicely drilled panel. It was also exciting to see a mix of self-taught undergraduate students battling out against semi-professional teams. Turns out, the pros didn’t win, everyone loves an underdog.

The first day started with a demo round to check that everyone could drill. Team Bug Eaters from the University of Nebraska-Lincoln was the first to leave the competition. After days of hard work, and what turned out to be an excellent learning experience, their robot just couldn’t drill.

They were one of two teams employing a delta-based design for their platform. The theory was that the delta would be more light-weight, and precise – which everyone agreed was a promising solution. It’s also a riskier approach, and the time to fully implement the robot was just too short. Team Akita from Japan, using the other delta-based robot, was not able to drill either because the arms weren’t stiff enough to carry the payload. Their setup was absolutely impressive however, and everyone was treated to a live demonstration of their massively extending arms at the end of the competition. This is probably the only robot of the lot that could realistically reach many areas in an aircraft.

With two teams out of the official competition, 5 were left standing for the first official round of the competition. This was a lucky coincidence as there were really only 5 drilling stations available at the venue. Imagine setting up drilling robots in a hotel room… quite tricky indeed. All the remaining teams based their robots on custom-made XY platforms or commercial robot arms from Denso or Kuka.

After the first whistle blow, everyone was silent, waiting for the drilling to begin. It is a race after all! They had to wait a bit longer however as the robots spent the first couple minutes calibrating and figuring out where the panel was so they could drill in the right place. For sensing, teams used a variety of vision systems combined with light patterns or lasers. One by one, they started drilling. Team Vayu was quick to drill, since they had a simpler calibration strategy which was only based on one of the reference holes, instead of the three available. Their control was open-loop, meaning they had no way of checking that they were drilling in the right place, and correcting. After drilling a series of holes, the drill broke, leaving them in a hurry to replace the bit. CriGroup had to do their calibration a couple times before they were ready to start drilling. Once moving, their arm performed beautiful kinematics to go to the different positions in their schedule. They were the only team to drill in an order that seemed completely random. Instead of drilling holes one after the other in a line, they used a traveling sales man algorithm to minimize how much they needed to move the joints of their robot arm.

Team Sirado, and team Naist seemed on track with their sophisticated drilling heads. Sirado’s driller could even sprinkle lubricant and blow away debris as it was drilling.

Team R3, a robotics collective based out of Ryerson University in Ontario, Canada, thought they would outpace everyone with their 7-bit drill XY system. But their robot never really started drilling. An electrical problem seemed to get the best of them that round, it can be a very long hour when things go wrong. Remarkably, the teams were continuously improving their approaches, even in the middle of a round. At one point, CriGroup decided to drill the constrained holes, which are in areas that are obstructed by pegs on the panel. About a millimeter before a crash with the panel, their operator pressed the emergency stop, something was wrong with the code, they had to reset the system.

At the end of the first round, the judges took the panels and started scoring them one by one. This turned out to be a tedious task as each hole needed to be checked for location, precision, and diameter with the help of a perfectly drilled reference panel. A score card was filled and cross-checked. With each panel having up to 255 holes, the judges ended up spending over an hour converging on a good scoring technique. The panel of team Sirado, which seemed nicely hole-ridden and ready to score big, ended up being misaligned due to a shift in calibration. They weren’t the only team with holes in the wrong place.

The judging was strict to replicate conditions on an actual shopfloor. The reality is, only perfect holes make sense for aircraft production. And that was the key message after round one – fast drilling without precision is useless.

So the teams went back to the drawing board for the night. By the morning, team Vayu had figured out how to reduce vibrations, so that the drill wouldn’t break, team R3 fixed their electrical problem, and everyone worked on better calibration. To make things more difficult, panels were tilted 30 degrees.

Team R3 hit the ground running, their drilling of 7 holes in parallel looked like it was going to obliterate the panel in no time. Vayu, after performing even better than in the first round, encountered a new technical problem. One of their cables was a tad too short and got disconnected, but they fixed that too. Team Naist was having trouble drilling due to a software problem. Something was wrong with the positioning of their holes, and they were continuously tweaking the code. There was also an issue with their three-finger solution slipping a bit.

At the end of the 2nd round, it looked like teams that performed best in the 1st round were having trouble, and those that performed worst were catching up. So all in all, a very close race! Four teams were determined to be above the fray, R3, CriGroup, Sirado, and Naist. They were all given a chance to compete in the finals.

Team Vayu didn’t quite reach the same level and finished the race then, which was already a tremendous achievement. They embarked on this challenge as 2nd year aerospace students from India working on their robots from 5pm to midnight in addition to completing all their coursework. They learned control on internet, and built a very clever and low-cost solution with motion controlled independently along 3 axes. Their lesson for next time: “we’d do closed-loop control, use three reference holes for calibration, and reduce vibrations”.

After a few more tweaks to the code, the teams were ready to compete once again in the final round. At this point, it was really unclear who would win. The boards were tilted again 30 degrees in a different direction. This messed up the calibration system on team Sirado’s robot because of light reflections. As a result, they were forced to recalibrate their robot every couple lines of holes. Team Naist looked like they had everything under control, although it took them some time to get started. They were still coding throughout the round. 3 minutes before the end, their drill broke. One team member sprung into action, whipping out a new drill from a pocket, and installing it for a final push to the finish line. Team R3 had swapped their 7-tip driller for a 4-tip driller.

The bell rang, everyone cheered, proud of the tremendous amount of work and time invested, and completely unaware of who was going to win.

After more scoring and deliberation, the winners were announced at the conference award ceremony. R3 came in a close fourth place, very impressive for a team of undergraduate students. Team Sirado came in third, with their near professional solution and specialised drilling head. CriGroup, came in second, they proved to be quite reliable throughout the competition and had a good working solution that was modular, light-weight, and open. Team Naist came in first, quite a comeback, and a clear reward for their constant improvements throughout the competition.

All in all, the competition was a great success. None of the solution were perfect, but combining the best of each could prove to be a helpful strategy on the shopfloor. Winning teams received up to €20K in prize and most important, an experience of a lifetime. They can all be proud of their tremendous accomplishments. It was impressive to see all the team’s solutions evolve in real time, with each round bringing a new series of working solutions to a concrete real-world problem.

You can also find our live coverage on twitter at #AirbusShopfloorChallenge or below.

#AirbusShopfloorChallenge Tweets

The Airbus Shopfloor Challenge invites robotics teams from around the world to create innovative robotic solutions for a real-life manufacturing challenge, and compete live at the IEEE 2016 International Conference on Robotics and Automation (ICRA) 15-21 May 2016 in Stockholm, Sweden. Deadline to apply is March 15, 2016.

About the Airbus Shopfloor Challenge

With a backlog of almost 7000 aircraft on order, at Airbus we are constantly innovating to find ways to improve our manufacturing processes by integrating emerging robotic technologies into our production lines.

Each aircraft is assembled through hundreds of thousands of point-based process steps. Most tasks involve a drilling process, a point-checking (i.e. measurement) process, and a tightening process. While some operations are automated, many remain manual due to a number of constraints – including space and weight restrictions. Using lightweight automation to perform such tasks would reduce the volume of the most repetitive, physically challenging labour of thousands of aircraft operators.

Because each generation of our manufacturing lines has a lifetime of more than a decade, robotic solutions need to be integrated into existing production environments, with no dedicated physical infrastructure. Accuracy requirements are stringent, and quality is paramount. Since errors in a single step can lead to costly fixes and even disruptions in production, solutions have to meet high reliability standards. In addition, they have to be cost effective in order to be widely deployed in existing Airbus factories.

Competing teams will design and build an autonomous lightweight robotic system able to perform accurate drilling compliant with aeronautic standards. Robots will perform several rounds of a simplified drilling task on an artefact representing part of the aircraft fuselage. Success will be measured based on the number of holes drilled within a specified time and accuracy. Some of the holes will be more challenging due to access constraints. Weight cost of the robot will also be taken into account.

Airbus will send panels to qualifying teams. We are willing to support contestants with equipment and material, subject to their proposal and needs.

The winning Airbus Shopfloor Challenge team will receive a cash prize, but more importantly, the team will have the opportunity to develop their idea for commercial application within Airbus. Other teams with promising solutions may also be invited to develop their idea further with Airbus.

Challenge details

The competition will run live during ICRA 2016. It will consist of multiple 60-minute rounds of drilling holes in an aluminium panel representing an aircraft part. The final round will be a confrontation between the top two teams.

The robot must be able to drill over the whole surface while installed on a fixed station. Multiple arms/effectors are allowed as a solution, as long as adequate fleet/synchronisation management is performed.

Hand-held electrical tools are currently used as end effectors, but participants are free to propose their own system. Manual lubrication is allowed. Airbus will work with qualifying teams to supply drilling tools and cutters.

No data transmission or remote control will be permitted during a round.

All algorithms must use an open source system as the main driver.

To qualify for the contest, the robot has to be:

• Modularly designed, at least at the arm level
• Weigh less than 100kg

Learn more about prizes, rules and how to apply here.