IARC Postmortem: 2016 in retrospect

We look back at a year of development.

IARC Postmortem: 2016 in retrospect

We know the burning question on all our reader's mind these days: "How can I leverage Deep Learning techniques to increase my Pokemon Go egg hatching rate?" No? Well fear not, because this post will not be about that.

Instead, as many of you that have been following us on Facebook have been asking for, this post will be about the IARC 2016 competition, and also what on earth we have been up to for the last 5-6 months.

And it sure is about time, since many exciting breakthroughs have been made since our last blog post. The technical side of things will have its own blog post coming up soon. It will be packed full of wisdom from our crew members — the type that is only gained in the difficult process of going from prototype to final product. So you should look forward to that!

Pre-summer - Getting bits and pieces to work

As this was our first year of activity, we really had no idea what we were doing. "What algorithms should we use for positioning?" "Which flight controller should we use?" "How many cameras do we buy?" "Oh hey, remember Teknisk Ukeblad? They're visiting next week to do a report".etc.

To give you an image of where we were at, here's an image of where we were at.

Oh. Well, it's not wrong.

Anyway, we were quite busy, balancing work at Ascend with regular school and life in general. Arguably, if we had been given a structure and a system to implement beforehand, things would have certainly gone much faster. But truth of the matter is that a large amount of time - necessarily! - went into researching and teaching ourselves computer vision, ai, control theory - trying and discarding certain algorithms which look promising on paper, but end up being fragile in practice.

As usual, when running a team, you can't just focus on technical stuff.

For instance, we recruited additional members dedicated to handling economy and marketing, since we saw that our current work balance probably wasn't good. We also designed some neat'o marketing material, such as these rad sweaters and rollups.

The members who were working on core portions of the positioning system, and the AI, were also busy writing their masters theses. Which is kind of a time sinker - or so we hear.

We wrote the team paper for the IARC symposium, which was due to be delivered 1. June - aka. mid exam period where you probably don't have time to write papers. So we purposefully worked on it in small amounts leading up to the exam period, and finished it after members had more time on their hands.

Summer - Working at the Innovation Lab

Perhaps the most exciting thing, was the offer we got from Kongsberg - our main sponsor - to be allowed to work at their Innovation Lab during the summer.

This sounded pretty swell to all of us, since it meant we would have a proper office - you know, with real windows, and coffee machines, and desks! This also meant that we would get access to full-scale testing areas - which we had been in desperate need for - as well as high quality tools and 3D printers for building the drone parts.

I guess we came a bit early, since the whole building itself had veryrecently been refurbished. In fact, we got several visits during the summer from people who were putting on the finishing touches. That includes automatic sunblinds, a remote-controlled safety net, and a motion tracking system.

During our stay, we got to focus entire workdays finishing the system, testing thoroughly in full scale, and getting everything ready for the competition. In hindsight, it is really mindboggling how much work had to be done during the summer, taking our drone from mostly research and individual components, to one fully functioning system.

Since we take testing seriously, we knew that we had to test our system full-scale, preferably many times. For this reason, we rented an entire gym a total of 6 days, spread across two weeks, and taped a 20x20 meter grid pattern on the floor (we can now call ourselves grid taping experts, in case the drone career ever fails).

The video above highlights our positioning system, our collision avoidance and the ground robot detector. Have a look!

August - Participating in the IARC 2016

At long last, after a near-fatal incident during a visit from minister Monica Mæland, many late hours at the gym, and ordering enough #09s from Pizzabakeren to cause a significant bump in their monthly revenue charts, it was time to pack our gear and fly to Atlanta city.

Upon arrival, we surveyed the neighborhood of our AirBnB and spotted an IKEA, which gave us a faint feeling of being at home. Unfortunately, those feelings quickly melted away whenever we had to drag our heavy equipment back and forth in the ridiculously high humidity — we had heard tales of the weather in Atlanta in the summer, and we did make sure to bring an abundance of spare T-shirts, but I don't think any of us were truly prepared.

The first competition day was dedicated to the symposium, at which the participating teams get to present their work. We enjoyed listening to the other teams' presentations, getting insight into their solutions and what problems they had run into, and also getting ideas for things we can make for next year. After the symposium we were excited to see everyone's drone in action, as well as eager to talk with the members of the other teams.

And we got a chance to do so later that day at the tech mixer sponsored by - among other - nVidia, who also held a guest presentation about Deep Learning for AI and robotics at the symposium. Cool stuff!

The second day was the day of the competition, held at the Zelnak basketball center. Usually the home to some of the most high-level games known to sports enthusiasts, the center was taken over by a large amount of pale nerds from all around the world, prepared to demonstrate the work they have been refining for the past year, while regular visitors were doing their reps at the benchpress - what a better venue could there be?

We had a reporter from Kongsberg, who tagged along and made this report which pretty much sums up the competition itself. You can watch the video below.

Following the competition we were invited to the awards ceremony, at which the winner would be decided. As expected, however, no team was able to completely solve the mission this year either. Atleast not at this venue; at the time of writing the Asia-venue competition is scheduled to be held September 22-25 in Beijing.

We also got free tickets for the aquarium (it was pretty rad).

Stay tuned!

As a year of work comes to close, we are sad to see some members are leaving, but also excited for the new team. We have many cool things in the loop for the coming semester, such as technical courses, live-coding sessions as well as new equipment and cutting-edge sensors that we are getting in - ready to be experimented with. If you would like be a part of Ascend's 2017 team, check out the Join Us page (we are recruiting!).

Finally, if you - as many others - suffer from FOMO, and would like to hear about the lessons we learned while building this year's drone, stay tuned for next week's technical bonanza blog post!

As always, thanks for reading, and remember - don't PokeGo & drive!