Tesla Driver Assistance Features Could Re-adopt Radar After Going Camera-Only – MotorTrend

Only one year ago Tesla was touting that it had begun transitioning its so-called Full Self-Driving and Autopilot sensor arrays from ultrasonic sensors (USS) and radar to a camera-only system backed up by an AI called Tesla Vision. Tesla initially began pulling radar units and ultrasonic sensors (which handle long- and short-range spatial and object detection, respectively) out of the Model 3 sedan and Model Y SUV, later yanking them from the larger Model S sedan and Model Y SUV in 2022. Now, it’s been revealed by the Federal Communications Commission (FCC) that Tesla has filed confidentiality requests and has begun seeking an extension for testing and using—you guessed it—onboard radar systems.
In a story posted by Jalopnik, it seems that Tesla has been testing radar systems since earlier this year and filed a 180-day confidentiality request with the FCC back in June. Now, nearly six months later, that agreement is up and Tesla has requested an extension to at least the beginning of February with a potential release date of “mid-January” for its “new” radar system.
Other information on the FCC’s site shows that Tesla’s radar works in the 76 to 77 GHz spectrum and doesn’t pulse to operate, a detail that potentially points to frequency-modulated continuous wave (FMCW) radar. In case you aren’t familiar with how radar works, a system sends out a pulse—a single signal over a set length of time—and waits for that signal to return to calculate distance and direction. With FMCWR, rather than pulsing a signal out and waiting for it to return, this radar system is always sending out a signal and looking at the difference between the frequency of what was emitted and what was reflected back to determine position and distance. This is actually a much cheaper system to use and actually ends up being more accurate than pulsed systems because it’s a continuous system that’s always emitting and listening.
What’s not clear, however, is what the future holds for Tesla Vision, the camera-focused software that relied on machine learning gleaned from Tesla vehicles’ data streams to continuously improve those cameras’ object detection and spatial awareness capabilities. The documentation only shows the testing data and requests for confidentiality, but doesn’t explicitly state that the FMCW radar is going to be a full replacement for those cameras or whether, as before Tesla pulled radar from its vehicles, the radar would augment those cameras.
Tesla still isn’t using LiDAR—a laser-based radar system that Tesla CEO Elon Musk has called a crutch to the automotive industry and its adoption of self-driving vehicles. Given the similar confidence he displayed last year in radar being unnecessary—and how that take has aged—don’t count that more expensive technology out for Teslas quite yet. Tesla Vision was supposed to create autonomous (or at least semi-autonomous) vehicles that see the world much like we do, but it appears that’s proving to be very difficult to accomplish. Look no further than the many incidents with Tesla vehicles running Autopilot and FSD Beta features failing to detect objects or vehicles ahead, or even, um, fake children.
While it’s certainly worthwhile for Tesla to try and pull cost and complexity from its vehicles by ditching radar, there is a reason most every other automaker with advanced driver assistance features (self-steering lane-keep assist and adaptive cruise control, which is essentially the duo that makes up Tesla’s Autopilot widget) continue to run both radar and cameras. Each system has different strengths and can be used to cross-check the other; for example, a camera can confirm whether something the radar “sees” ahead is actually in the lane you’re traveling in based on, say, the lane markers’ path ahead, or what the obstruction might be (a person? an animal? another vehicle?). Radar is much better and accurate at longer-range detection than cameras, even ultra-high-definition ones.
The only other major manufacturer relying solely on cameras is Subaru, with a twist—its EyeSight driver assists rely on stereoscopic vision, in which two cameras angled ever so slightly apart are used to calculate depth of field in addition to “seeing” objects ahead. Even then, Subaru recently added a third forward-facing camera and doesn’t claim anywhere near the capabilities Tesla does for its Autopilot and Full Self-Driving features. It seems Tesla is learning that it might not be able to deliver on those claims (and any future ones) without radar, after all.


Leave a Reply

Call Support:



Useful Links


Top Model Escorts. All rights reserved.