OUT WHERE THE ROBOTS ROAM

When Pablo Molina and Faizan Sheikh graduated from the University of Waterloo and moved to Ottawa in 2011, they wanted to apply their new robotics know-how to solve a genuine market need. Canada being Canada, their eureka moment conjured up the world's first self-driving snowplow.

But the young inventors soon encountered a problem that people in Ottawa might have trouble giving credence to: There wasn't enough snow.

The revelation came in a restaurant next to a Walmart parking lot, where, as part of their research, Molina and Sheikh were interviewing the owner of a snowplow company. He explained just how small and sporadic the demand for snow removal really is. It didn't make financial sense to build a robotic vehicle, they realized, if it would only be used a handful of times
per season.

But just as their entrepreneurial dreams were turning to slush, the driver pointed to the megastore and said, "Why don't you go indoors? Floors need cleaning every day." Roboticists are always on the lookout for dull, repetitive tasks, and in that moment, Sheikh and Molina found their own little slice of workplace drudgery.

Five years later, the two co-founders of Avidbots are showing me the prototype of their self-driving floor scrubber in what must be the cleanest warehouse in Waterloo Region. The prototype looks like any other floor scrubber—minus the handle—but it does the job slightly faster and a lot more cheaply than a human can, and it never gets bored. The first batch of 10 of the as-yet-unnamed product ships to customers in June, and Sheikh says they've got enough orders to keep shipping 10 a month in the short term, and eventually ramp up production by early 2017.

It wasn't long after they realized they had a viable business that Sheikh and Molina left Ottawa and, after attending a hardware accelerator program in China, moved back to Waterloo to start Avidbots. They wanted access to the engineering talent coming out of the University of Waterloo, and more importantly, to be part of one of the most concentrated and supportive tech clusters on the planet. Where once Waterloo's engineering grads fled the area for flashier jobs abroad, the brain drain has lately been arrested, as more companies are, like Avidbots, capturing venture capital abroad and launching their start-ups back home.

The buzz about Waterloo Region's tech ecosystem has been gaining volume for the last five years. But firms like Avidbots underline a little-noticed shift: the ascent of hardware. The scene's biggest driver has been purring quietly in the background, out at the edge of town where the robots roam.

For a time, Research In Motion, as BlackBerry was known back then, was the sun around which everything else orbited in Waterloo. Today, its former office buildings are being colonized by other companies: Waterloo has reinvented itself as a 24/7 start-up factory. There are more than 1,100 new companies in a place of only half a million people, which makes Waterloo the world's second-densest start-up cluster after Silicon Valley. (Greater Toronto is a dozen times Waterloo's size but has only two or four times the number of start-ups.)

You can visit most of those companies in the course of a short walk. A lot of the action can be found at an old tannery building on Charles Street West, which is home to Velocity Garage, the largest free start-up incubator in North America. The Avidbots principals spent seven months there after their return to Waterloo, at the encouragement of Velocity's director, Mike Kirkup.

When start-ups graduate from Velocity Garage, many of them simply move to another wing of the building, where Communitech offers a co-working space for hundreds of tech entrepreneurs. Here the start-up ecosystem is on full display, with big corporations maintaining outposts to stay abreast of new innovations, angel investors stopping in to check on prospects, and the wider community massing to network.

Continuing out the door of the tannery building, the next stage of evolution for a growing Waterloo start-up is a few minutes away on foot, at 305 King St. West, which houses many of the region's rising stars, like Chalk, MappedIn and Square. Between those two locations and a handful of others—including the Breithaupt Block, home to the new Google Canada headquarters—Waterloo's tech cluster hasn't just taken up BlackBerry's space; it has repurposed the ruins of Waterloo's disappearing manufacturing sector, and reinvigorated a downtown core that five years ago was a place to avoid.

Almost all of these companies originated in the University of Waterloo's engineering faculty, which is the largest in Canada by number of students, and so well regarded that it ranks in the top four feeder schools for tech titans like Google, Facebook and Apple, along with 864 other Silicon Valley companies that recruited here in 2015.

The attraction is mutual. Silicon Valley holds a lot of appeal for young engineers, and for a long time the region struggled to hold on to its talent. With the engineering faculty's mandatory co-op program, along with the university's business-friendly policy on intellectual property, its funding for Velocity and its affiliation with a similar jump-start program at the Accelerator Centre, U of W has played a huge role in seeding the cluster.

To reach critical mass, Waterloo needed more than all those measures—it also needed some intrepid founders to build businesses locally. One of those was Clearpath Robotics, which is a leader in Waterloo's emerging hardware sector, and a sort of older brother to Avidbots.

Like Avidbots, Clearpath makes a self-driving workhorse for industry, albeit a much sturdier one, the aptly named Otto. A low-lying black disc not much larger than a pallet, Otto looks as sleek as any smartphone. Yet it is one powerful robot, carrying up to 1,500 kilograms. Otto started shipping out to some of Clearpath's Fortune 500 customers this spring. It's currently transporting medical equipment at a repair facility in Wisconsin for GE; elsewhere in the same state, John Deere is using Otto to transport materials on an assembly line. Clearpath also makes a smaller model for lighter loads.

But Clearpath's CEO, Matt Rendall, says that Otto is not just carrying boxes; it's making supply chains more flexible. "Our hypothesis is that if a factory is like a city, and traditional automation is like a metro system, a really busy city also needs taxis to deliver ad hoc, point to point, and more efficiently," he says. "This is automation and infrastructure that you don't bolt to the ground, and that means when you change daily, so does your automation."

Waterloo has seen its share of automation that you do bolt to the ground, and much of it is on display at New Age Robotics, a repair shop and reseller about a 20-minute drive from Clearpath. Here, the history of robotic arms for automotive manufacturing is arranged in colourful rows, like so many mechanical birds with their heads bowed. New models mingle with old, all of them the same basic shape: one or two large, multi-jointed arms on a heavy base, created to do one thing—paint a door, weld a rivet—over and over again. Autonomous technology in drones and wheeled vehicles lets robots move around better, but when it comes to manipulating or manufacturing things, the physics isn't much different than it was in the 1960s.

Andrew Goldenberg of the University of Toronto, who helped design the Canadarm, says that robotics is at an awkward stage: The software side has vastly improved, but the underlying electromechanical systems haven't. "We can do wonders, but in reality, the mechanical part is lagging," he says. Thus, a program equipped with machine learning can beat a Go champion, but a humanoid robot can barely stay upright.

While robots that drive and fly are currently finding eager reception in such industries as manufacturing, agriculture and mining, walking robots have farther to travel to reach their markets. "We still don't have a motor that is of the same weight and size and that can generate the same power as the human body," says Waterloo professor Dana Kulic, who develops algorithms for robots to incrementally learn and imitate human action.

Although the electromechanical challenge is daunting, roboticists are eager to overcome it, because humanoid robots open doors to applications that wheeled and winged devices can't reach, like assisting the elderly and disabled or performing search-and-rescue operations. "In those situations, the environment has been designed for the human form. If you have a robot that's human-shaped, it'll be easier for it to access the same tools and use the same space," says Kulic. It's easier to make robots that fit the world than a world that's fit for robots.

But the real world is a complicated place—just ask the developers of the Google car, who have to write code that can handle every possible situation, including the time one of their vehicles encountered someone in a motorized wheelchair chasing a family of ducks with a broom. This is why self-driving machines fare better in the controlled setting of a warehouse or factory. Otto will never have to identify a pile of leaves or avoid a stray cyclist. "Clearpath and Avidbots have found two environments where there's a limited number of confusing cases, and so they're able to automate to a level that's reliable," says Steven Waslander, who directs the Waterloo Autonomous Vehicles Laboratory at the University of Waterloo.

This is what makes Otto special: not the physical robot, but its mapping and localization capabilities, according to Karamdeep Nijjar at iNovia Capital, which participated in a $14-million investment round for Clearpath last year. Clearpath isn't the first company to offer a self-driving vehicle for industry. Amazon and Staples have been using Amazon's Kiva bot in their warehouses, but that system relies on grids of stickers on the floor, each printed with a bar code, to navigate. You can drop Otto into any industrial setting and it will figure out the lay of the land on its own. So while the robot is no slouch, the real value proposition is in the software.

A software-differentiated hardware company faces all the same challenges as a standard software company, plus quite a few more. Venture capitalists are more attracted by the low overhead and recurring revenue streams that software companies enjoy. Hardware companies need more people: the usual array of software developers, but also electrical and mechanical engineers. They also need more cash than a software company, which—thanks to the availability of cloud services—has few upfront costs. A hardware company has to build prototypes, inventory, a supply chain, and a lot of other things that cost money. Software companies go to market faster and see returns sooner, while hardware companies have lumpy revenue streams and find it difficult to keep sales high on the second or third versions of their product.

But hardware start-ups have their advantages too. They can succeed with a small number of customers willing to pay a high price for a premium product, whereas software companies rely on massive user adoption and heavy traffic. There's a higher barrier to entry in hardware than software, but that cuts both ways: In hardware, there's less competition.

And a decade into the smartphone era, app fatigue is making it much harder for new software start-ups to find their market. The recent spike in major hardware acquisitions, such as Apple buying Beats Electronics for $3 billion (U.S.) and Facebook absorbing Oculus for $2 billion (U.S.), suggests more people are betting on software-differentiated hardware to answer the "what's next?" malaise lately hanging over Silicon Valley.

In the same way that the Cold War and the space race spurred advances in computer technology that made Silicon Valley what it is today, competition in the smartphone market has accelerated the trend of software-differentiated devices by driving down the cost of sensors and computing. Smartphones also illustrate why simple hardware coupled with great software makes a powerful business model. "All the cost and complexity goes into the software, which has a huge upfront cost, but then you can amortize that over millions of units," says Rendall.

Nijjar doubts whether Clearpath would have been able to find its footing before the "smartphone dividend" made sensors and computing so inexpensive. And the trend is only accelerating, as companies like Clearpath and Avidbots search for ways to use cameras to perform the mapping and localization functions that currently require costly laser-based sensors.

There is, however, still the small problem of talent. Even though technology makes physical distance irrelevant and promotes remote interaction, the people building the stuff want to be as close to each other as possible. Disadvantage hardware: It doesn't fit into cool office spaces, and instead is confined to beige buildings at the edge of town. When you deal in tangible products, you need warehouses and shipping bays more than couches and climbing walls.

Kurtis McBride of Miovision, which makes a traffic optimization platform for city planners, aims to solve this problem by creating a rallying point for the hardware cluster that's closer to the start-up action downtown. Catalyst137 will be a 475,000-square-foot facility only a few minutes away from Communitech. There will be space for circuit-board production, rapid prototyping and varied product testing (environmental, thermal, etc.). The site will also have a "hackable streetscape" with sensors throughout, so hardware start-ups can test their devices in a real-world simulator.

There's one more hump for the sector to get over: scale. Waterloo's tech cluster may be the second densest in the world, but it doesn't even rank in the top 25 for the total value of its companies. The combined GDP of the Ontario tech corridor is about $350 billion, which, if it were a single cluster, would put it in the top 10 municipalities in North America. Waterloo would be in a much better position if it mimicked Silicon Valley's relationship to San Francisco. Hello, Toronto.

Here's where it might actually be an advantage to be big, but not too big. Toronto's real-estate prices may seem stratospheric compared to Waterloo, but both cities offer cheaper digs for both companies and people than Silicon Valley.

But the bond won't be realized until transit is improved enough to allow an easy daily commute between the two regions. The trip by car currently takes anywhere between 40 minutes and three hours. A drive to Toronto often begins with a traffic jam before you even leave Waterloo.

San Francisco has the Caltrain connecting it to Silicon Valley. "If you look at the distance between Toronto and Kitchener/Waterloo, it's basically the same," says Rendall. There aren't any plans to build high-speed rail, but the mayors of Toronto and Kitchener have been pushing for improved transit connections. For now, that might just mean moving the lone rail offering—which is not part of Ontario's Go commuter system but an 11 a.m. Via train—to a better time for commuters.

Of course, self-driving cars will solve the problem eventually. But if Waterloo and Toronto wait for that, today's upstart robotics scene will become one more Canadian tech wave that we let slip away.