TESLA’S NAVIGATE ON AUTOPILOT IS LIKE WAZE ON STEROIDS
Tesla’s self-driving future is closer than ever — and probably not what we thought it would look like
The first time the Tesla Model 3 changed lanes by itself to pass a slower vehicle, I was overcome with an ice-cold “this is the future” feeling. Stuck behind a slow-moving garbage truck on New Jersey’s Highway 17, the Model 3 suggested a passing maneuver, I agreed, and lo and behold, automatic lane change. But by the fourth time the car changed lanes, I was left wondering why Tesla’s Autopilot wasn’t smart enough to merge back into the original lane of traffic. Because what is the future if not endlessly nit-picking every technological achievement until we forget why we were amazed in the first place?
Tesla has a lot staked on Autopilot. Elon Musk is reportedly personally involved in testing out and debugging the advanced driver assist system, which uses a host of cameras and sensors to control the vehicle’s speed, braking, and steering. As its safety is debated across the industry — and as competitors race to introduce their own partially autonomous systems — Autopilot is the feature that could give Tesla an edge as it grows from niche company to global powerhouse.
With that in mind, I trekked out to the barren wasteland of northern New Jersey on a recent Thursday to test out “Navigate on Autopilot,” Tesla’s latest update that guides the car from “on-ramp to off-ramp” by suggesting and making lane changes, navigating highway interchanges, and proactively taking exits. Tesla owners have been jonesing for this update for years, though the company describes the feature as a beta version.
Since it was my first time behind the wheel of a Model 3, Tesla only felt comfortable letting me drive from the Courtyard by Marriott in Paramus to the Hilton Garden Inn in Ridgefield Park, an 11-mile journey along New Jersey’s most flavorless highway. I don’t blame them — there is a learning curve involved in driving a Model 3 since so much of the vehicle’s functions have been subsumed by the 15-inch touchscreen. It’s not the most intuitive experience!
Navigate on Autopilot is a subtle improvement, but still feels incredibly consequential. While testing it, I was struck by how Autopilot is evolving from a mostly passive tool that helps ease the stress of highway traffic, to one that is more actively — and sometimes aggressively — engaged in the core aspects of driving.
First of all, it needs to be said: Navigate on Autopilot does not transform your Tesla into a self-driving car. (If you’re reading a headline that claims otherwise, I recommend rethinking your browsing habits.) The car even says so, in the warning screen that pops up on the center display when you first activate the feature: “Navigate on Autopilot does not make your Model 3 autonomous. Like other Autopilot features, the driver is still responsible for the car at all times.”
Tesla recently removed the “full self-driving” option from its website, because the company said consumers found it confusing. But after experiencing Navigate on Autopilot (albeit briefly), I feel Tesla’s self-driving future is closer than ever — and it’s probably not what we all thought it would look like.
It’s the first time Tesla has grafted a layer of mapping data on to Autopilot. Navigate on Autopilot is like Waze on steroids. It offers all the step-by-step navigational help that anyone with a smartphone or a car built in the last decade is accustomed to. But then — with permission from the driver — executes those maneuvers itself. Navigate on Autopilot only works on divided highways, and warns drivers to take control when exiting via a distance countdown (400 feet… 300 feet… 200 feet) to when the feature disables.
In previous versions of Autopilot, the vehicle’s sensor suite — eight cameras, 12 ultrasonic sensors, radar, and the car’s AI computer “brain” — executed all the major functions, such as Autosteer, lane-keeping, and traffic-aware cruise control. Now, that real-time data is being cross-checked against terabytes of mapping data that Tesla collects from the hundreds of thousands of vehicles it has on the road today.
Navigate on Autopilot can be customized to a driver’s preferences, including four settings for speed-based lane changes: Disabled, Mild, Average, or Mad Max. For my first time demoing the feature, Tesla set it to mild, which honestly I preferred. I wasn’t trying to get over my own skis.
Cruising along New Jersey’s Highway 17, we approached a garbage truck that was driving significantly slower than the 60 mph I’d set the Model 3’s adaptive cruise control to maintain. In a message appearing on the left-hand side of the touchscreen, Navigate on Autopilot suggested we pass the truck. All I needed to do was click the turn-signal stalk and the Model 3 completed the lane change itself. The maneuver felt seamless, if slightly robotic.
In its testing, Consumer Reports criticized Navigate on Autopilot for occasionally suggesting lane changes when another vehicle was quickly approaching from behind. Tesla said there may be occasions when the car decides to cancel a suggested lane-change if it detects another vehicle speeding up from behind. I didn’t experience this issue in my test drive, but I could see how it could become dangerous, especially if the driver becomes inattentive and fails to check blind spots before confirming the lane change.
Not everyone is thrilled about the new feature. “On my morning commute downtown [Navigate on Autopilot] sticks like glue to the left lane until [a quarter of a] mile from the exit where it finally decides to try to cross [six] lanes,” one Reddit user griped. “That’s about 2 miles too late…” I suppose that’s why it’s still considered to be in beta.
I enjoyed the novelty of a car performing its own lane changes — who wouldn’t? — but I found myself wondering how many prompts were too many and when would they start to adversely affect a driver’s attention. There are plenty of drivers who will tell you the last thing they want is a car that is constantly chirping, vibrating, or flashing messages at you. But for better or worse, that seems to be where we’re heading as drivers. Cars are growing smarter and more connected, and gosh darn it, they’ve got things to say! Of course, just like you can silence your phone’s notifications, you can also disable your smarty-pants car.
Tesla argues that our attention is already divided while driving, between the road, our phones, and the infotainment screens, and this new automated feature is designed to help us focus on the road ahead by giving us less to do.
Which brings us back to that message in the warning screen, and the idea of driver responsibility. Lately, when there’s been a crash involving Autopilot (and there have been a few), Tesla will inevitably release a statement hammering this point: drivers need to pay attention to the road, and Autopilot does not prevent all accidents.
Navigate on Autopilot occupies this weird space between human and automated driving that many experts still struggle to define and regulate. Tesla says Autopilot makes driving exponentially safer. The company recently released its first safety report that found that drivers using Autopilot experienced one accident or crash-like event every 3.34 million miles, as opposed to one accident or crash-like event every 1.92 million miles driven without Autopilot. Other than that, though, the report was pretty light on details.
We don’t have very good research into the correlation between systems like Autopilot and driver attention. The Insurance Institute for Highway Safety recently put a bunch of cars equipped with partial autonomy to the test, and found that the Model 3 (with an older version of Autopilot) performed better than the competition. Consumer Reports, though, gave Cadillac Super Cruise a higher score than Autopilot in its first ADAS ranking. And The Drive’s editor-at-large Alex Roy also prefers Super Cruise, but with caveats.
Tesla is still missing a critical safety feature: active driver monitoring. Cadillac attached an infrared camera to the steering column so it could track whether drivers were watching the road when Super Cruise was engaged. Tesla only requires drivers to apply torque to the steering wheel as indication they are maintaining control of the vehicle. Engineers inside Tesla wanted to add robust driver monitoring systems, but reportedly rejected the idea out of concern that the options might not work well enough, could be expensive, and because drivers might become annoyed by an overly nagging system.
Much like regular Autopilot, Navigate on Autopilot is going to take some getting used to — especially for the uninitiated driver. A couple times, when the lane we were driving in forked into two lanes, the car would lurch without warning to the right if the system determined that was the appropriate lane. Fortunately, Tesla representatives warned me about this, so it wasn’t a total surprise.
Tesla acknowledges there will be scenarios where human driving is more effective than Navigate on Autopilot: for instance, merging with fast-moving traffic, or navigating a lane of traffic that’s completely stalled.
Like most driver assist systems equipped with automatic emergency braking, Autopilot is programmed to focus on maintaining a safe distance from other moving objects — cars — and to ignore stationary objects. That’s led to some deadly incidents, such as the death of Joshua Brown in Florida in 2017 and the fatal March crash in Mountain View, California, that killed engineer Walter Huang.
We didn’t encounter any problematic interchanges or construction zones that could have scrambled Autopilot’s perception during my test drive. The ride was seamless, and I found myself wishing I had more time behind the wheel. Everyone who uses Navigate on Autopilot is teaching the system as a whole how to drive better. And I felt like I still had a few lessons to impart.