Just How Smart Are the Latest Autonomous Vehicles?

by Jen McKenzie

Recently, Intel Corporation purchased Israel-based Mobileye, a company that manufactures automotive components, for 15.3 billion U.S. dollars. Mobileye creates a diverse portfolio of sensor arrays, vehicle networking, road mapping, cameras, and machine learning.

The primary applications for these components are in automated vehicles field. What was their motivation for this deal?

Autonomous vehicles are at our doorstep

That such an influential player as Intel will pay this kind of money in an industry so far removed from its core business illustrates just how serious things are getting in the autonomous vehicle world.

Intel competitors Motorola, Qualcomm and Nvidia have long been key players in systems control for the automakers. Car manufacturers use their chips in navigation, ignition, fuel injection, audio-visual systems and transmission controls. Intel seems to be the only name missing from this list. They have stayed away from the automotive sector up to this point, concentrating on its domination of worldwide computing.

Autonomous Is A Big Deal

This intersection of the technology and manufacturing industries is already quite dynamic. Stock analysts advise caution that many startups based around driverless cars could be overvalued at this stage of the industry’s development. Dozens of firms have announced plans to produce driverless technology by 2021. A few even proposed to field a roadworthy autonomous vehicle by that time.

Autonomous in the Real World

Driverless vehicles are already among us, with some of the latest developments stunning, others potentially disturbing.

On the stunning side, the U.S. Postal Service began a testing program to ship mail between Phoenix, Arizona and Dallas, Texas on driverless tractor trailers. The trucks won’t be 100% autonomous at this point, they will carry an operator to monitor and take over if need be.
Meanwhile, in a suburb of Phoenix, a pedestrian was hit and killed by an Uber autonomous vehicle in testing on public streets.
There is also a lawsuit in progress in California involving an autonomous car that hit a motorcycle during a lane change.

Such instances make the public weary of autonomous vehicles. Besides, how would you program a car in case an accident is unavoidable? To save the occupant of the vehicle at all costs, or risk their lives to avoid a larger tragedy? The thought experiment has made rounds online just a year ago.

When are Big Players Planning to Release their Autonomous Cars:

Tesla CEO Elon Musk makes annual predictions that driverless cars will be available by a year. At this point, he’s at 2023.
Uber released Ford Fusion cars modified with the company’s self-driving tech as part of its Pittsburgh test in September 2016.
Google’s Waymo has logged over two million miles of autonomous driving around San Francisco in cars that have no steering wheel or brakes. Waymo has strongly hinted at 2020.
Toyota has invested over a billion dollars in AI, robotics and road mapping to release a driverless car to the public in 2020.
Nissan has announced plans to release driverless cars in Japan by 2020.
Ford built a new facility to manufacture autonomous vehicles for a fleet service operated by Ford Motor Company in 2021.
Daimler plans to release a Mercedes-Benz driverless truck in 2020.

The Tech Behind Autonomous Vehicles

When Henry Ford perfected the rolling assembly line in 1913, it’s doubtful he ever considered the possibility that Ford cars would drive themselves a century later.

How does a driverless car perform its magic? First it’s important to recognize the differing levels of vehicle automation. The Society of Automotive Engineers lays out six levels of autonomy.

Level 0: Manual control by the driver, no automation.
Level 1: Systems which aid the driver, such as cruise control and lane drift warnings.
Level 2: Steering and acceleration both automated, for instance lane correction, automatic collision avoidance and automatic parking.
Level 3: Environmental detection. While these systems can make some decisions, such as passing slower cars or changing lanes, the driver needs to override the system regularly.
Level 4: No interaction required. The car can operate itself and perform driving tasks on its own, but the option to override for manual control remains in effect.
Level 5: No human control. This is the only true driverless car, with no steering wheel or other human controls. Google’s Waymo cars are an example. While this level has been achieved, the vehicles at this point are still not capable of reacting appropriately to the same endless list of possibilities that humans can.

To perform the complex task of driving an automobile without supervision, Level 4 and Level 5 cars need a CPU to process the extreme amounts of data, cameras of various kinds, accelerometers, energy sensors such as laser rangefinders and radar, motion sensors, ultrasonic sensors for parking or maneuvering and GPS data. This comes with its own set of issues.

Challenges

Toyota probably never dreamed of seeing their light system slowly progressing to having actual smart sensors on their assembly lines that let you know when maintenance is needed.

The fact that all key automotive manufacturers are pouring their resources into producing a driverless car ready to hit the road in just a few years points to the idea that smart, driverless cars are inevitably coming down the road.

But how smart (and how driverless) are they, really?

Vehicle-to-vehicle communication (V2V) and vehicle to an external environment (V2X) technologies are being developed to allow driverless cars to pass and receive data. The problem? A nightmare that haunts every engineer working on the central processing is the idea these systems might suffer a breach, allowing remote car theft, assassination, criminal activity or causing chaos on the roadways.

Still, V2V and V2X systems would allow for cooperation between vehicles and the driving environment. Such a system could also transmit information they had stolen the car, drive it to the police or obstruct its movement. The flip side is that a hijacking from a remote location could disrupt entire groups of automobiles by manipulating traffic controls.

Autonomous Vehicles Might Be Ready Soon, But How About Humans?

The biggest hurdle to adopting autonomous vehicle systems may not be the technology at all, but people. If the driverless car is exceedingly safe around pedestrians, people will learn they can just step right into traffic to cross the street, potentially bringing traffic to a standstill.

Besides, the phasing period, when there will be autonomous and human driven vehicles on the road, will be the most challenging part. We might be close to having fully functional autonomous vehicles, but people will need some time to get on board.