The long road to autonomous driving
From Transformers to the Batmobile, autonomous vehicles may have once been a sci-fi fantasy. But innovative technology is bringing the self-driving car closer to reality and paving the way for an inevitable evolution.
Although features such as lane assist and self-parking are already available in modern cars, the reality is that true autonomy is no easy task and the end goal is still a distant future. No one can know for certain when autonomous vehicles will become readily available. “It’s anyone’s guess,” says Keenan Burnett, team lead for the University of Toronto’s self-driving car team, aUToronto. Burnett says autonomous vehicles could become a reality by 2040.
That said, more realistic advancements in the autonomous vehicle industry have the potential to be implemented much sooner. “Practical applications like autonomous shuttles or guardian systems are probably right around the corner,” says Burnett. Autonomous shuttles would be used for short distance travel, bringing passengers back and forth on a specific route, and can revolutionize transportation for airports, theme parks and hotels. A guardian system is an artificially-intelligent feature added to your vehicle that’s capable of taking away control from the driver in situations where the driver would have otherwise failed to prevent a crash themselves.
Robert Pedretti, an electrical engineer, has been driving his Tesla Model 3 since June, 2018. “When using the car’s autonomous features, at first of course you’re scared,” he says, “but you’re curious and once you’re used to it, it’s scary how comfortable you become.” As impressive as the Tesla Model 3 can be, Pedretti has yet to fully commit to purchasing Tesla’s autonomous feature add- on at its current price of $9,200. “First, I enjoy driving the car, and letting it drive itself is just boring,” says Pedretti. “The price is really high right now and I know the technology will be better and more reliable down the road.”
According to Burnett, the idea of autonomous driving can be broken down into four key attributes: perception, prediction, planning and control. These attributes come together through the collaborative efforts of multiple systems like GPS, LiDAR and camera sensors.
Autonomous vehicles are designed to replicate the functionality of the human brain when driving but as Burnett explains, it’s not enough to just be as good as humans. A self-driving car must be five, even 10 times better. “If you are only as safe as people,” he says, “people won’t buy it.”
Navigation tools like GPS give the autonomous driving system a metaphorical bird’s eye view of exactly where the car is on the world map, seeing key aspects in its path like intersections or railroads. LiDAR creates a 3D picture of the area surrounding the autonomous vehicle. The primary function is to detect and avoid obvious obstructions, such as pedestrians or other cars.
The use of camera sensors surrounding an autonomous vehicle identify specific elements pertaining to the rules of the road like traffic lights, street signs and break lights. All that information is then processed by an artificial intelligence system designed to make the appropriate decisions based on the data it receives.
Today’s leaders in autonomous driving research are Waymo, owned by Google, and Cruise, overseen by General Motors. It’s uncertain which company will revolutionize autonomous driving first, but what is certain is that the road is in no way collaborative. Instead, it is a race with the goal of profitability.
Many factors can, and will disrupt the development of autonomous vehicle technology. The advancements of these systems are not as safe as that of computers or smartphones. Failures or bugs in system updates would not only lead to inconvenience, but injury or death. As a result, many corporations restrict their testing to unpopulated areas with safety drivers in the vehicles. “Most companies are cognisant of safety concerns,” says Burnett. “Uber got into an accident and that basically set them back two years. They only now just started putting their cars back on the road.”
The 2018 Uber self-driving car incident killed 49-year-old Elaine Herzberg from Tempe, Arizona. In this instance, prosecutors ruled Uber not criminally liable for the death its vehicle caused. It was ruled that human error was to blame because the safety driver failing to react. However, Burnett says in the proposed future of autonomous driving, the assumption is that pedals and steering wheels will be removed from vehicles completely. It is then only logical that manufacturers become liable for their products.
While a fully autonomous vehicle may still seem like a distant possibility, Burnett and his aUToronto team are committed to exploring these advancements in the automotive industry. “It’s pretty daunting,” Burnett says. “You are trying to validate that your system is going to work for every possible, conceivable experience.”
This story was originally published in the print version of Emerge magazine.