Most automakers are testing autonomous vehicle technologies in controlled environments under close scrutiny. In stark contrast, however, Tesla and some of its owners are taking self-driving technology development to the streets.
The latest example of seat-of-the-pants vehicle R&D comes from YouTube user greentheonly who posted a video of his hacked Tesla Model S recognizing and sometimes stopping for red traffics lights.
In the video description, greentheonly reports his car is running Tesla firmware 19.8.3. With it, his car reportedly recognizes both red traffic lights as well as stop signs and stops for both. Though, the stop sign recognition is not shown in the video.
If the system isn’t sure whether it is nearing a red light or stop sign, it will approach the intersection slowly until it can make a definitive reading.
However, it’s not perfect. As the passenger-turned-driver notes, he had to intervene to prevent the car from running a red light at the five-minute, 12-second mark in the video.
It’s not clear how greentheonly’s vehicle is capable of stopping itself at red lights. The latest Tesla Autopilot update does include a feature to detect red lights and warn if your vehicle is about to run it. The functionality of automatically stopping the vehicle at the detected light is not yet a part of the Autopilot system, however.
Tesla is not the only automaker playing with traffic light recognition. Audi is testing a system that taps into to the traffic light systems of various municipalities. The German automaker is using this information to calculate how quickly or slowly the car should drive, in order to hit the green light every time.
It’s easier for an automaker like Tesla to simply create a software system that uses digital camera information to read red lights than spend the time tapping into the grid, as Audi has negotiated with various municipalities. However, it’s much less safe to rely on visual information alone. Hopefully Tesla won’t roll this feature out officially until it has a more robust solution in place than just cameras.