The International Conference on Robotics and Automation took place in Hong Kong the week before last and with a few of my students I was able to attend (and we present four papers on our work). Each year this conference gets bigger, and this year there were over 2,000 attendees, which is probably symptomatic of the growing importance and impact of robotics today. There were 19 parallel tracks meaning that for any single attendee there was a good chance there was more than one talk of interest going on at any given moment. On one hand this is bad, since it make it impossible to hear all the relevant talks. On the other hand, it was a lot better than squeezing all the talks into a smaller number of tracks and giving them insufficient time to present the results gully. In short, given the growth in the field, I though this was a good compromise.

Computational and algorithmic issues continue to be the key, and perhaps even growing, theme in robotics, as opposed to the also import issues of mechanism design, dynamics etc. There were several nice talks (and papers) on using computer vision to address issues in robotics. Vision and robotics used to be joined together under the once-ill-defined umbrella of artificial intelligence. Those areas separated into different communities well over a decade ago, but as each matures the links are begging to re-emerge. It is about time to see the areas coming together again. In fact, I think there is a sound argument to be made that vision and AI are both best regarded as sub-fields of robotics, which is the amorphous umbrella that captures humanity’s efforts to replication human and biological capabilities with all the sub-themes interacting. Thus, vision in the context of robotics is become more and more important and mature. Underwater robotics is also a theme of growing prominence, as is the combination of sensing and manipulation.

While in Hong Kong I had a chance to do a tiny bit of sight seeing, including a great trip to the Mainland China city of Shenzhen. Shenzhen is a key industrial center, and the place where many of today’s consumer electronics are fabricated. The abundance of electrics manufacturing, development and prototyping was really impressive. I have often observed that Silicon Valley (where I once lived) has good resources for electronics prototyping and technology development. In Japan, however, the popular level of access and appreciation to such activities is much greater, and in a big department store one can find real beakers, test tubes and soldering irons beside the toys, dresses and dishware (then this is good quality scientific material, not just kiddie-toy stuff that one finds in Montreal). Well, at least for electronics Shenzhen takes this trends to a far far far higher level, with a plethora of devices, tools, resources so tremendously available and accessible that any comparison with North America is almost laughable. It was cool, fun, impressive and maybe a bit frightening. As for the regarding to items and ideas that are supposed to be protected by intellectual property legislation, that’s also an interesting story.

All in all, Hong Kong and Shenzhen combined a top robotics conference with a very educational little outing to where the consumer electronics world has its nexus. In addition, I got to see a few great sights and eat some delicious (and sometimes challenging) food.

This summer I spent some time in the San Francisco Bay area and visited Google's Mountain View "campus" where I went for a drive in one of the self-driving cars. The fact that Google has developed a small fleet of cars that can drive themselves completely automatically has been widely reported. These cars drive themselves on city streets, highways and across bridges completely automatically finally delivering on a widely anticipated vision of the future. For the time being, a human driver is always required at the wheel to take responsibility for the car's actions and assume control in case of emergency, but there is no doubt in my mind that once these cars enter the mainstream it will not be many years before the prospect of a human assuming control on normal streets will be regarded as an antiquated risk to safe driving.

For my drive, which I shared with a couple of friends and colleagues, we had two Google staff members in the front seat. Their job during our drive, as for the drives they said they take every day, is to both supervise the car's behavior and to log and observe any anomalous things that happen on the road or in the car's behavior. For example, during our ride a cyclist's bike fell into the road nearby. The car handled it fine, but the staffers made notes in the log that would allow them to inspect the data later and carefully analyze how the car handled the event.

The car is still experimental and very exotic. One of my former students who work at Google was rather jealous of my opportunity. No doubt, all the Googler's will get their chance some time soon. On the other hand, the experience is very conceptual and rather anti-climactic. Other than chatting with the computer science staffers that are involved, you are just driving around on what feels like a perfectly unremarkable car ride (except for some technically interesting live data displays the software engineers have).

It's not clear yet how the robot car will enter the mainstream, but at least 9 different US states have already added legislation that allows robot cars to operate on their roads (with a human supervisor at the wheel). The cars have safely driven over 500,000 miles by now, and they have done so with an essentially perfect safety record, which is better than comparable human drivers.

It will be interesting to see if any Canadian provinces take this plunge in the near future, but I rather doubt it since I suspect the required legislation requires the intercession of Google legal staff, and Canadian's are still waiting for Google Voice to come to Canada almost a decade after it's availability in the USA. Moreover, the challenging weather that Canadian's face for most of the winter certainly don't make this a great location for initial introduction. More generally, it seems unlikely that the self-driving car will first appear as a product for the ordinary consumer, due to it's very substantial cost and the importance of a positive initial reception.

Among the advantages the self-driving car will offer include the increased mobility it will offer physically or visually impaired people, the promise of a much safer driving experience, the ability to carry heavier traffic loads on existing roads, and reduced fuel consumption due to more efficient driving behavior. Whatever the path to commercial introduction is, the robot car has really arrived, it's really working now and it was a thrill to get to drive in it.

Adobe has been accused of spying in users by sending back detailed reports of what people are reading, what pages they look at in an ebook, and what books they have in their digital library. This seems like an unacceptable invasion of privacy, and is part of an increasing trend whereby commercial software products and devices report back on detailed user activities, ownership and habits. This can be excused to some extent in the cases of companies and services like Facebook or Google since their entire business model is based on collecting whatever they can about users, and people presumably give up their rights to privacy willingly in exchange to pay for the privilege of using these services.

It's worse for commercial software to snoop on users activities and invade people privacy. This is a trend that should not be tolerated or encouraged and perhaps North American legislators should take a hint from the European Community which has exhibited real leadership in developing legislation that protects user privacy.

Notable, in the case of Adobe the terms of service for the ebook software (DE4) includes wording indicates that they do collect user data and use it for purpose that include developing licensing models with publishers. This is, of course, buried in the typical ream of legalese. Moreover, it isn't completely clear how much data is collected.

References

http://the-digital-reader.com/2014/10/06/

ICRA Seattle was a bigger, brasher and more successful robotics conference than has even been seen before. To no small extent, this is due to the explosive growth in the important and success of robotics in almost every regard. For my entire career I have been hearable, and predicting, the imminent success and growing relevance of robotics to industry and society, and it's finally becoming blatantly obvious to our whole society. The was reflected at ICRA in the growth in the number of accepted and submitted papers, the growth in the number of people attending (about 3,000), the growth in the number of people attending who did not have anything to personally present, and the vast growth in the trade show where recruiting was a major theme.


Aside from a few long invited talks, papers were presented via a very short oral pitch and a subsequent extended interactive session in the same room. While this format has proven popular with the research community, I am not fond of it myself. The primary merit is that the very short presentations minimize the time you need to spend listening to talks on material you don’t care about; you can then go discuss work you are really interested in. In think this is a disservice to the community since sitting through all talks is a great mechanism for getting a cross section of the community including areas you might not naturally attend to. Many related research communities have fragmented as they grew, and failing to pay attention to research themes other than your own is a perfect mechanism to encourage this. In robotics this would be especially unfortunate since developing robotic systems is a highly integrative activity where broad awareness is particularly important.

Despite my concerns over the format, I though the paper quality was very high and heard about some nice work. One of our own papers was nominated for the best paper award, but unfortunately we didn't emerge as the winner. (The paper was Learning Legged Swimming Gaits from Experience by Meger, Gamboa Higuera, Xu, Giguere, and Dudek). It was great to be a nominee, however, and the other 4 best-paper nominees reflected a great selection of research labs and we supervised by some friends I greatly respect. Amusingly, one of the co-authors (Tsotsos and Soatto) of the winning paper was the son of my own PhD supervisor John Tsotsos!

The big trade show included exhibits from traditional hard-core robotics companies like Adept, long-standing companies with a narrower focus like Da Vinci Surgery, as well as big companies that have started expanding significantly into the robotics domain like Amazon.com.

I also got a chance to visit one of the fulfillment centers operated by Amazon which heavily depends on technology from Amazon Robotics (formerly Kiva). While there are pictures on the web and the natures of these robotic facilities are well known, seeing the place in person was mind-blowing. It was just huge (even though this was a "small" center, and the efficiency due to automation was truly impressive).

Finally, own team won the bid to hold ICRA 2019 in Montreal, which is a big deal in many ways. I'll be Program Chair and Jaydev Desai will be Program Chair.

This year the International Conference on Robotics and Automation was held in Stockholm, Sweden. It was made up of three intensive days of technical paper presentations, and a couple of days of workshops, working group meetings and associated activities. About 3,000 researchers were in attendance which reflected continued growth in the community -- robotics is, of course, very hot right now -- although in absolute numbers the attendance was not as big as it would have been in more affordably and readily accessible places like Seattle (where numbers were through the roof). Keep in mind that the largest number of internationally active researchers is still in North America.

Personally, this year was very gratifying since there were strong participation from members the NSERC Canadian Field Robotics Network (NCFRN), including my own lab, and a fair number of my former students were also there. As a result, it felt like Canadian researchers really had a strong mutually-supportive presence and a cohort of moral and intellectual support to a degree which I have previously only associated with a few major "teams" on the international stage. The importance of such "community" support is easily underestimated, but I have come to realize that it makes a significant difference in the ability of researchers to flourish and prosper professionally. In addition, scope and complexity of robotics and AI research today means that progress increasingly depends on multi-area research teams that need to be developed and maintained as both social as well as operational entities.

The conference was organized to allow to a few plenary talks an hour long, a number of keynote talks introducing each session, and numerous very short presentations that introduced ideas that were then elaborated in subsequent poster presentations (often called "interactive sessions: since they typically used a live display, as opposed to a plain paper poster). Personally, I greatly prefer longer presentations over short advertisements accompanied by a poster. This is because posters do not allow more than a few people at a time to deeply inspect the material or query the speaker.

Canada has long been recognized as a world-class player in robotic systems and sensors, and the related science it depends on. Companies such as MDA and Syncrude and organizations such as the Canadian Space Agency (CSA) and the Defense Research and Development Canada (DRDC) have long histories of significant investment in autonomous systems. Many smaller Canadian companies are also key players in the autonomous systems market, including Robotiq and Independent Robotics Inc.

Canada has historically had a very strong reputation in robotics and is thus well placed to be world leader in both robotics (including artificial intelligence) and even the the Internet-of-Things (which is basically uses robotics technologies and tools to build small devices that may not be independently mobile). While the expected importance of the field in terms of science and industrial development is apparent, equally clear is the relevance of the area in terms of science education and even our national identity. Robotics is a key motivator for students to study science and engineering. Likewise, student with robotics training are highly sought-after by industry since they have a broad skill base.

Outdoor (field) Robotics is probably the most Canadian branch of Science today! The Canadarm and Canadarm2 robotic manipulators built by MDA Space Missions (previously Spar Aerospace) are Canadian icons. In a 2008 poll reported by the CBC, the Canadarm was identified as the #1 Canadian accomplishment of all time, ahead of peacekeeping, universal health care, insulin, and the telephone.

Canadian industry and academia are at risk, however, of being left behind as the worldwide pace of robotics research, investment and system integration accelerates. This is especially critical since a large number of Canadian industries or sectors are current or potential users of autonomous systems including aquaculture, farming, oil and gas, mining, security, forestry, and materials transport. Moreover, fully functional robotic systems entail substantial integration between different component technologies and this places major emphasis on collaboration between different research groups, agencies and companies, a fact that has been fully understood by government agencies in the USA, in the European Union, in Japan and in Korea where large new national robotics programs are being launched or are underway.