As of Sept 1st 2016 I am:
- On sabbatical leave
- Not the chair/director of Computer Science,
- Not a senator of McGill University,
- Not the president of CIPPRS (Image Processing Canada),
- Not on the board of directors of the Rozynski center for fine arts (this happened a bit earlier this summer).
What I will be is visiting Stanford and pursuing a range of projects.
As of Sept 1st 2016 I am:
I'm off to Tokyo to attend a variety of events and see some people. With a 13 hours non-profit stop flight, I really had hoped for an upgrade to business class using my accumulated points. I wasn't so lucky, and so will have to endure a couple of list days on the other end as I recover. I hope the other people at the international conference of experimental robotics don't think I'm an imbecile due to the fatigue state I will be in. For trips this long, and people over 25, or more than 6 feet tall, business class is less of a luxury and more of a survival strategy.
The Symposium on Experimental Robotics was very interesting. It's s small single track meeting where you really get to talk to the individuals who are working on projects with a strong experimental flavor. I have attended the meeting many times and would have to say the quality is probably increasing, perhaps consonant with the increasing impact and importance of robotics world wide. Perhaps the most surprising talk I heard was from Carrick Detweiler's lab where they have a flying vehicle that is used to create controlled burns to manage natural resources. To put that more bluntly, it's a quadrotor that spits out flaming/explosive balls to cause the ground below to catch on fire.
Our own work on using teams of different kinds of robots to monitor/measure coral reefs went well and was delivered by Alberto Li. In that work as use a temp of three different kinds if marine vehicle to collect ocean data at different spatial and temporal scales.
On a personal note I loved walking around Tokyo, which is a city with amazing energy.
Afterwards, I went to visit Kyoto University where I gave a colloquium on research projects in our lab, then I got to visit David Avis.
Now I'm off to Seoul Korea en route to the IROS 2016 Conference (International Conference on Robots and Systems) , a multi-thousand-person robotics meeting. The anti-reflective coating on my eyeglasses has started to peel off, perhaps from the humidity here. I'm going to check out Namdaemun Market, in the center of Seoul, where I can apparently pick up a good pair of glasses fast and on the cheap. That should be an adventure! Namdaemun is the biggest traditional market in Korea and seems to sell everything, with some shops have their own factories.
My sabbatical continues. It has been a difficult and complicated one, however, since there were some health issues within my family, not the least of which was my mother breaking her hip while visiting in the USA. This was, among other things, a great lesson for me regarding the differences between the Canadian and US health systems and the astounding rate at which medical bills could build up in the USA.
Despite that, and in other news, we had great marine field trials in Barbados, my students launched gozazz ride sharing, and we are developing plans to build the second phase of the Canadian Robotics Network.
The first computer I built, or at least one of the earliest, was made without any electrical of electronic parts, and learning to play a game at superhuman level. It's a story I've occasionally told in class, and I thought I should document it here. It also has many common threads with modern reinforcement learning.
The idea was based on an article by Martin Gardner, and is based on explicitly representing every possible state of a game called hexapawn. Each box contained a bead for each possible code (with moves shown on the outside of the box by colored lines, each bead inside having a corresponding color). To play the computer, you select the box matching the current board state, shake it and select a random bead. If there is not bead, the computer resigns (and thus loses). You make that move on behalf of the computer.
If the computer loses the game, you find the last bead that was used, and throw it away, thus preventing the computer from event making that move again. As you play, moves that lead to a bad outcome are gradually pruned away and with time, the "computer" becomes unbeatable.
I've tried to replicate this game as a course assignment a few times in various forms, but nothing is as compelling as a computer made only of wood, beads and paper.