The Do’s and Don’ts of Tesla’s Autopilot System

August 12, 2016

from Car Connection

Recently, we’ve spent a lot of time writing about Tesla’s semi-autonomous driving system, Autopilot. The technology has been linked to a fatal crash in Florida, which has given rise to three separate federal investigations, plenty of high-profile criticism, and several defenses of the system from Tesla.

Tesla is working on a major update to Autopilot that may improve its functionality, but whether the update will provide more training for users remains to be seen. One of the major critiques leveled at Tesla over the past few weeks has been that Autopilot is still in beta, and yet the company has made it available to consumers without providing adequate instruction on how to use it (or safeguards to prevent them from misusing it, as Nissan promises to do with its system).

If you’re interested in Autopilot but want to know more before engaging it in your own vehicle, check out this clip from the folks at Drag Times. Clocking in at nearly 15 minutes, it’s long by today’s internet video standards, but if you plan to let software shuttle you around at highway speeds, those are probably 15 minutes well spent. Among the major points covered in the clip:

A complete walkthrough of the Autopilot interface: Even the most intuitive software needs some explaining now and then. In Autopilot’s case, knowing exactly what the dashboard illustrations mean can help drivers know when–and when not–to use the system. For example, below 18 miles per hour, black lines along the travel lane indicate that Autopilot can detect the lane edges, suggesting that when the vehicle’s moving faster, Autopilot will be available. Above 18 miles per hour, those lines should switch to blue: if they don’t, it means that Autopilot can’t see the edges of your travel lane.

A list of important “don’ts”: As the speaker explains, Autopilot shouldn’t be used on roads without medians, like small, two-lane highways. In those cases, if the lane markers aren’t well-defined, Autopilot may not function as intended, and the margin for error is very small.

Also, using Autopilot in construction zones isn’t recommended because lane markings may not be visible, or there may be old lane markings present that are no longer used. (Also, there might be barrier walls or narrow/missing shoulders that increase the risk of collisions.) As the speaker says, if it’s confusing to you, it’s going to be confusing to Autopilot.

A list of important “dos”: Tesla Autopilot is made for highways, particularly divided highways with well-defined lanes. If you’re on such a road, though, be sure to keep your hands on the wheel, in case the software loses sight of the lane edges.

A rundown of Autopilot’s limitations: Perhaps most importantly, Autopilot can’t read traffic signals, so it’s not going to slow down or stop at red lights. Worse, because traffic lanes don’t run through intersections, Autopilot may lose its way when traveling through them.

Highway exits can be challenging to Autopilot, too. When you’re traveling adjacent to an exit lane, the lane markings can sometimes make Autopilot want to exit the highway. (Ironically though, when you want to take an exit–even when it’s been programmed into your navigation–Autopilot won’t take it.)

Glare is also an issue for Autopilot, whether it’s caused by driving toward the rising or setting sun or by reflections off rain on the roadway. In situations like that, Autopilot can’t always see the edges of the lane, increasing the opportunity for something to go wrong. Autopilot can rely on the location of vehicles in front of the Tesla to stay in its lane, but that’s not as reliable as reading actual lane markers.


Read the rest of the article here