Baby, You Can Drive My Car

What’s better than engaging cruise control and taking your foot off the pedal? And do you ever wonder how you parallel parked without a back-up camera? Your next car will be even more automated, likely offering an autopilot feature to keep your car a safe distance from everyone else on the road.

Piece by piece, the driverless vehicle is heading your way. Fully autonomous vehicles (AVs) – with sensors linked to artificial intelligence systems – are already here, albeit in disguise; self-driving vans are getting to know Toronto streets, with a human in the “driver’s seat” mainly to avoid alarming other motorists.

Think you’ve heard this before? This time, it’s true. Companies from Tesla to Google’s mobility division Waymo promised AVs by 2020 – until a fatal collision two years ago in Tempe, Ariz., proved that mixing humans and massive mobile robots is more complex than experts thought. Still, the economics are so compelling that scientists and entrepreneurs around the world continue to work on improved navigation systems – and York alumni and faculty have their foot to the floor.

Computers have been flying planes for years

York graduate Rod McPhail (Urban Studies ’84) taught transportation planning at the Faculty of Environmental Studies for seven years and spent nearly four decades as a municipal transportation planner, including 16 as Toronto’s Director of Transportation Planning. Now an independent consultant, he helps cities map out their futures – and he sees AVs as the catalyst for a revolution.

Automated transport is nothing new; rapid transit systems have been fully automated for decades in cities from San Francisco to Vancouver – and apart from takeoff and landing, McPhail notes, “computers have been flying planes for years.” The dream of bringing automation to our roadways is almost as old.

“Back in the ’80s, we were widening expressways to get higher capacity. One of the questions we asked was, ‘What if we could get the drivers out of transportation?’” Replacing human drivers with precise electronic controls could double a lane’s capacity without incurring construction costs.

Now that the technology is close, McPhail insists driverless will win out. He cites three key reasons.

The first is safety. In 2018, Canada recorded 1,922 deaths from motor-vehicle accidents, and 153,000 injuries. The problem is not cars but drivers. A U.S. study found that human error – like speeding, distracted driving or driving under the influence – causes 94 per cent of all crashes. The AV industry is too new to produce meaningful statistics; fatalities are usually measured in billions of kilometres driven (in 2016, Canada had 5.1 fatalities per billion vehicle-kms travelled), and even industry leader Waymo has completed only 32 million kilometres of testing.

Think what a street could look like if there’s not a car at every home

Still, New York City–based research firm ARK Invest estimates a fully autonomous system could reduce accidents by over 80 per cent. A 2017 RAND study suggests we shouldn’t wait: deploying AVs just 10 per cent safer than human drivers would save more lives than waiting until the technology is significantly better.

A second factor is their sustainability. Autonomous and electric vehicles work hand in glove, says McPhail. Battery-powered AVs will help cities meet carbon-reduction targets, while their simpler operating systems mean fewer repairs. And when consumers truly embrace car-sharing, says McPhail, the land now occupied by parking lots and driveways can be
repurposed: “Think what a street could look like if there’s not a car at every home.”

And, finally, there’s the economic incentive. Replacing drivers will save trucking firms, cab fleets, rideshare operations and their customers a lot of money. Secrecy surrounds most companies’ pricing plans, but participants in a 2019 Waymo trial in Phoenix, Ariz., paid about US$5.99 for an 8-kilometre ride – one-third the comparable taxi fare. ARK Invest estimates fully autonomous taxis could charge as little as US$0.35 per mile, about half the “all-in” cost of vehicle ownership – a savings that could spark a lasting movement from owning cars to sharing them. 

One drawback in the shift to AVs will be the loss of jobs such as truck driver and transit operator. But as McPhail notes, “the average truck driver is on the older side. The industry is finding it hard to get young people interested.” Once the AV shift is complete, the savings should fund new types of economic activity, but for now, no one knows what those benefits will be.

Meanwhile, there are bugs to work out. AV sensors – radar, lidar and cameras – remain imperfect; some have trouble “seeing” traffic lights, others balk at left-hand turns. And A.I. control systems may be confused by random events like a swarm of pedestrians. Those on-road trials are part of the answer; A.I. thrives on endless data. But the industry also depends on new ideas.

Jinjun Shan’s specialty is spacecraft design. Growing up in Harbin, China’s space-research capital, he dreamed of being an astronaut. He ended up doing the next-best thing: designing dynamic controls and navigation systems for fleets of small spacecraft, on the theory that multiple smaller (and cheaper) ships can do the same jobs as one big one. Following post-doc research at the University of Toronto’s Institute for Aerospace Studies, he joined York’s Department of Earth and Space Science and Engineering (ESSE) in 2006.

Driving is often a game between you and other drivers

Interest in formation-flying spaceships has dipped, but Shan’s expertise in controls and navigation has found a second life in two sectors: drones and autonomous vehicles. Now chair of ESSE, Shan has received funding for a platform for academics and industry collaborating on “multi-agent systems” such as autonomous truck fleets, or drones that work together to deliver heavy payloads, monitor agricultural data or fight forest fires.

Roads are also multi-agent systems, says Shan: “Driving is often a game between you and other drivers.” He and his team are devising ways to help AVs interact more intelligently – especially at intersections – and increase system efficiency while reducing misunderstandings and accidents. Since A.I. requires massive amounts of information, they’re also investigating new data sources to help AVs’ software predict what other drivers will do next. “We’re working on both the systems design and the programming,” says Shan, “but we’re not yet at the application stage. It will take millions of hours of training to get there.”

Shan admits researchers worldwide are tackling similar ways to make AVs roadworthy. But he sees multi-agency as a niche technology to help vehicle manufacturers reduce sensor costs and solve complex interaction problems. “You never know what will happen,” he says. “You’ve got to think outside the box.”

Recent York grads Raghavender Sahdev and Bao Xin Chen (both earned a master’s in computer science, specializing in vision-based robotics) are betting on a different niche in the AV revolution: driverless trucks. Companies such as Amazon and the U.S. Postal Service are already testing self-driving trucks on U.S. highways (with backup drivers in the cab, usually awake). But cruising the I-75 is easy; navigating rigs through crowded city streets is hard. With their startup NuPort Robotics, Sahdev and Chen are developing autonomous controls for short-haul trucks that carry cargo from container ports to nearby distribution centres.

While many factories and warehouses have robot platforms that follow predetermined guideways, “follow-me” technology saves time and money by enabling mobile platforms to carry loads anywhere a human controller decides to go. The pair’s test platform could carry up to 15 kilograms, but they envisaged industrial versions that could carry up to 100 kilos.

Sahdev and Chen fielded several offers for their vision-based controls. But they demurred, hoping to identify a bigger market to benefit from lower labour costs. After looking at shipping, forestry, even the military, they settled last fall on short-haul trucking, a $60-billion corner of the transport sector. “There’s not a lot of people working on autonomous solutions in this sector,” says Sahdev, “so we moved up from 15-kilo [loads] to 15,000.”

They will save a bunch of money over time

NuPort’s systems will enable conventional trucks to haul shipping containers in autonomous convoys. They’ve already built partnerships with retail, automotive and shipping companies eager to cut costs and reduce accidents. “They’re really excited,” says Sahdev. “This will make their whole supply chain more efficient.”

The company is looking for investors, as Sahdev expects it will take two years to start generating revenue. While he won’t reveal what kind of return on investment clients can expect, he promises “they will save a bunch of money over time.”

He and Chen are also working with industry regulators, calling them “very supportive,” and predicting NuPort-enabled trucks will hit the streets within four years.

And this is why AVs are finally headed your way. Industry wants the ROI, government the safety and environmental benefits. Unlike many entrepreneurs, Shan and Sahdev and Chen don’t have to sell anyone on the concept of change; customers demand it. Says Sahdev, “This market is waiting to be automated.”  ■