Skip to main content

Tesla Autopilot's groundbreaking new feature gives it an edge over GM's Super Cruise, but Autopilot still has a big flaw compared to its major rival, an industry expert says (TSLA)

Mark Brake / Getty Images

  • Tesla's new feature that allows Autopilot, the electric-car maker's advanced driver-assistance system, to recognize stop signs and traffic lights gives the system the ability to in some cases operate on city streets and residential roads.
  • Advanced driver-assistance systems that can control steering and speed (while still requiring human supervision) were previously limited to highways.
  • But the new Autopilot feature raises safety concerns because it will automatically slow down before a green light if the driver doesn't intervene, said MIT research scientist Bryan Reimer.
  • While Tesla has developed an edge in advanced driver-assistance technology, its cars don't have the hardware necessary to turn Autopilot into a fully autonomous system, said Carnegie Mellon professor Raj Rajkumar.
  • Tesla did not respond to request for comment on this story.
  • Are you a current or former Tesla employee? Do you have an opinion about what it's like to work there? Contact this reporter at mmatousek@businessinsider.com.
  • You can also reach out on Signal at 646-768-4712 or email this reporter's encrypted address at mmatousek@protonmail.com.
  • Visit Business Insider's homepage for more stories.

When Tesla began rolling out a new feature for Autopilot, the electric-car maker's advanced driver-assistance system, in March that allows vehicles to recognize stop signs and traffic lights, it set a milestone in the world of automated driving.

Other, similarly robust driver-assistance systems, like Cadillac's Super Cruise and Nissan's ProPilot Assist, are designed for use on the highway (General Motors has said it's working to expand Super Cruise's capabilities to city streets, but has not given a timeline for when it expects those features to become available). Now, Autopilot can in some cases operate on urban and residential roads. This is previously uncharted territory for what SAE International calls "Level 2" driver-assistance systems, which can control speed and steering at the same time, but requires constant supervision.

The new Autopilot feature, called Traffic and Stop Sign Control, suggests Tesla's Level 2 technology, particularly the computer used to run Autopilot's software and the eight cameras positioned around each Tesla vehicle equipped with the hardware, is ahead of other automakers', said Raj Rajkumar, a co-director of the General Motors-Carnegie Mellon Connected and Autonomous Driving Collaborative Research Lab.

"They're clearly ahead of the competition," he said.

But while Autopilot is very good at driving, the same can't be said for its ability to make sure drivers pay attention to it, said Bryan Reimer, a research scientist at the MIT Center for Transportation and Logistics. Autopilot requires drivers to regularly apply pressure to the steering wheel; if the driver doesn't do so, the system will eventually pull the vehicle over to the side of the road and deactivate. Super Cruise goes a step further, using inward-facing cameras to make sure the driver is watching the road. Autopilot doesn't have that capability, and videos of Tesla drivers appearing to sleep while using the feature show the downsides of relying only on steering wheel contact to monitor driver attentiveness.

"The technology is really good at driving," Reimer said of Autopilot, but it's "not very good at ensuring the driver is performing their functional role in overseeing the system."

Tesla did not respond to request for comment on this story.

Autopilot's new feature raises safety concerns

Traffic and Stop Sign Control raises additional concerns about safety, Reimer said. Unless prompted by the driver, it will automatically slow down before a traffic light, even if the light is green. That could lead to a rear-end collision, Reimer said.

"A typical driver following you, even if attentive, is not going to expect this deceleration," he said.

And while the Tesla driver has consented to using Traffic and Stop Sign Control, which the company says is still in "beta" mode, nearby drivers haven't agreed to participate in the public testing of an incomplete feature, Reimer said. 

Tesla argues that Autopilot already makes its vehicles safer, pointing to comparisons it publishes between the crash rates for Tesla vehicles with Autopilot, Tesla vehicles without Autopilot, and non-Tesla vehicles. Consistently, Tesla vehicles with Autopilot engaged experience collisions at a lower rate than vehicles without the system. But, Reimer said, Tesla doesn't provide enough details about its crash data, like whether the company will record a collision in which Autopilot was deactivated shortly before impact as an Autopilot-related incident.

"We need robust, scientific understanding of the data, and Tesla has not been forthcoming with the data," he said.

If the safety implications of Traffic and Stop Sign Control are uncertain, the reasons why Tesla would want to release it are not. Tesla CEO Elon Musk has said the feature will build on what he has described as one of the electric car-maker's biggest advantages over other companies that are developing autonomous-driving technology: data collection. Tesla uses hundreds of thousands of the vehicles in its customer-owned fleet to teach and refine its Autopilot software. Even if a vehicle doesn't have Autopilot engaged, its onboard computer can still compare a decision the driver made to how Autopilot would have responded in the same situation.

The ability to collect data from a relatively large number of vehicles is significant and gives Tesla an edge over its competitors in the autonomous-vehicle industry, said Missy Cummings, the director of the Humans and Autonomy Laboratory at Duke University.

"That is one capability that they have that is superior to most everyone else," she said.

But there's a difference between collecting data and understanding the lessons it holds, Cummings said. 

"There's no magic wand that you can wave and then, instantly, the data reveals its secrets. You still have to do a lot of post-processing and data manipulation to actually learn anything from the data," she said. "Tesla and every other company out there do not have the manpower or the capacity right now to really get through all that data and really learn what they need to learn."

Full autonomy remains out of reach

The scale at which it collects data is a major piece of Tesla's strategy to eventually bridge the gap between Level 2 and Level 4 autonomous-driving technology — which can control all driving tasks within a prescribed area or situation without the driver having to pay attention to the road — by gradually adding to Autopilot's capabilities over time. (In contrast, many of the leading companies in the autonomous-driving industry, like Waymo, Cruise, and Argo AI, are putting all of their effort into developing Level 4 systems). But while Traffic and Stop Sign Control breaks new ground for a Level 2 system,  it doesn't offer any insight into Tesla's progress in developing Level 4 technology, Rajkumar and Reimer said.

Tesla vehicles don't have the necessary hardware to turn Autopilot into a Level 4 system, Rajkumar said, as they have only one radar and no lidar sensors (which emit pulses of light to detect nearby objects). Waymo places three radars and four lidars on its vehicles.

Musk has argued that lidar isn't necessary for a vehicle to be able to drive without human oversight, saying if humans can drive with their eyes, a self-driving car can rely on cameras, radar, and sonar. But, Rajkumar said, there's a big difference between the human brain and artificial intelligence.

"Comparing what humans can do with what computers can do these days I think is extremely misleading," he said. "There's a huge chasm between human intelligence and so-called artificial intelligence."

Musk's confidence in Tesla's approach will be tested during the rest of this year. In April, he predicted that by the end of December, a Tesla vehicle will be able to drive an owner from their home to their office, often without any human intervention. Musk has a history of missing similar deadlines — last year, he said he was "certain" a Tesla vehicle would be able to drive without human assistance by the end of 2019 — though, over the years, Autopilot has gradually moved closer to his vision for it.

For now, Musk's ambitions for Autopilot to become a nearly self-reliant system will likely remain out of reach, Rajkumar said.

"I think it's all hype," he said. "It's not going to happen by the end of the year."

Are you a current or former Tesla employee? Do you have an opinion about what it's like to work there? Contact this reporter at mmatousek@businessinsider.com. You can also reach out on Signal at 646-768-4712 or email this reporter's encrypted address at mmatousek@protonmail.com.

NOW WATCH: A cleaning expert reveals her 3-step method for cleaning your entire home quickly

See Also:

SEE ALSO: Tesla could be getting closer to unveiling an EV battery that could last 1 million miles, experts say

Data & News supplied by www.cloudquote.io
Stock quotes supplied by Barchart
Quotes delayed at least 20 minutes.
By accessing this page, you agree to the following
Privacy Policy and Terms and Conditions.