Tesla is working on Autopilot’s blinker detection for better lane-merging

Elon Musk recently provided a brief update on one of Autopilot’s upcoming features — blinker detection, which is already “mostly working” in the developer software installed in the CEO’s vehicle. Musk’s update came as a response to SpiderLabs founder Nicholas J. Percoco, who inquired on Twitter about improvements in Autopilot’s capability to merge with traffic on expressways.

The ability to detect turn signals from other vehicles would be an invaluable addition to Autopilot’s repertoire of features. The system’s inability to read blinkers from other cars on the road, after all, has become a common cause of Autopilot disengagements, especially when other cars attempt to merge lanes in front of the electric vehicle. These observations have been aired by Model S, Model X, and Model 3 drivers on both Twitter and the forum community.

This point for improvement in Autopilot is prevalent during rush hour in stop-and-go traffic. When a car attempts to merge into a Tesla’s lane, Autopilot becomes incapable of anticipating what the vehicle would do. With blinker detection features, however, Teslas would be able to make more accurate predictions about the behavior of nearby cars based on their turn signals. Tesla could accomplish this by using its array of 8 cameras, 12 ultrasonic sensors, and forward-facing radar to look at surrounding vehicles. If Autopilot could accomplish this, tasks such as lane-merging would be a lot smoother.

Tesla’s upcoming blinker detection features for Autopilot fit right in with the company’s upcoming Full Self-Driving suite. According to Elon Musk, the electric cars’ first Full-Self-Driving features would be announced by Tesla sometime in August, together with the rollout of the highly-anticipated Version 9 software. As Tesla begins the release of its first Full Self-Driving features, capabilities such as blinker detection would be invaluable in keeping the electric cars safe and functional on the road.

This could be easier said than done, however. During a talk at Train AI 2018 last month, Tesla head of AI Andrej Karpathy noted that it is challenging to train Autopilot in identifying turn signals due to the variability of blinkers from one vehicle to another. Considering that massive data sets are required to properly train Autopilot in identifying turn signals and understanding what they mean, it has become quite difficult to get enough training data (turn signals, after all, are usually turned off). Musk’s recent tweet about the upcoming feature, however, suggests that improvements have been made on this front. 

As noted by Elon Musk in a recent tweet, Tesla has focused primarily on safety with software Version 8. Earlier this year, Tesla rolled out an Autopilot update that introduced vast improvements to its driving-assistance capabilities, thanks to an extensive rewrite of the driver-assist system. With its latest update, Tesla has opted to increase the rate and assertiveness of Autopilot nag — a system which reminds drivers to keep their hands on the wheel. Once software Version 9 rolls out, however, Tesla buyers who purchased the Full Self-Driving option would finally start seeing their investments pay off.

Simon Alvarez: Simon is a reporter with a passion for electric cars and clean energy. Fascinated by the world envisioned by Elon Musk, he hopes to make it to Mars (at least as a tourist) someday.
Related Post
Disqus Comments Loading...