Farewell to Mobileye: Tesla's Next-Generation Autopilot Technology Forecast

Tesla's termination of cooperation with Mobileye was interpreted by various experts as being due to a car accident in May. This is nonsense.

There are two reasons for this: First, the direct and primary responsibility for the accident is the owner of the vehicle. Tesla's Auto-Pilot clearly indicates that it is an auxiliary role. Second, as early as before the accident, Tesla has already started to develop new mobile phones that do not rely on Mobileye. Automatic driving technology.

The accident occurred because an American owner gave his driving authority to the Tesla Auto-Pilot function when he was driving on the fast track, and was fully focused on playing games. As soon as a swift road passes, a white container truck turns left and out. At this point, the Mobileye camera equipped by Tesla does not recognize the other party. The radar and ultrasonic sensors also missed the truck because of physical limitations.

In the end, Tesla drilled into the bottom of the truck, the roof was lifted, and the entire vehicle was flattened. The accident caused the attention of the relevant US Department of NHTSA and launched an investigation. Auto-Pilot this configuration to spend more than 20,000 yuan to install the equipment, turned out to be life-saving features?

It can only be said that the Tesla company is still too young. Similar assistive driving functions, such as Mercedes-Benz's DISTRONIC PLUS with Steering Assist (equivalent to Auto-Pilot), automatically detect whether the driver has mastered the steering wheel after it is turned on. If you do not hold it for more than 12 seconds, this function is automatically released. This avoids the driver’s excessive dependence and trust in autonomous driving.

However, in terms of publicity and philosophy, Tesla feels that Auto-Pilot is automatic driving, and owners can use it with confidence. The disclaimer in the manual contradicts Tesla’s outward signals.

Back to the topic. As a monocular visual recognition solution, Mobileye can accurately recognize the front object in most cases. But in the above incident, the sudden departure of the white truck was a special scene that Mobileye could not reach. The specific reason is not repeated here.

What we want to discuss here is Tesla's abandonment of Mobileye's EyeQ3 chip - a visual identification solution that has already shipped over 3 million - what kind of technology will be selected next?

There are numerous signs that the binocular visual recognition system will drive the next generation of Auto-Pilot. As early as 2015, Tesla Model X's official rendering, "leaked" this technical solution.

As shown above, the inside of the Model X windshield (marked with a red frame) is obviously a binocular system, which is also called a stereo camera. Earlier, Elon Musk stated that he wants to further cooperate with Mobileye, but it does not prevent Tesla from developing its own visual identification technology.

For Tesla, Mobileye is most likely just a temporary transition plan. In the vision of Elon Musk, Tesla will launch fully-autonomous models in 2020. And to achieve this, a reliable visual recognition system is crucial.

Mobileye's EyeQ3 is the most advanced visual recognition SoC, and the new generation of EyeQ4 will be available in 2018. Unlike current cameras based on a single camera, the EyeQ4 can handle information from multiple cameras. However, the core technology of Mobileye still depends on image recognition to determine objects, lane lines, speed limit cards, and so on.

Tesla stayed with his cooperation with this Israeli company. It did not deploy Mobileye's original DNN (deep neural learning network) algorithm, but instead designed a set of technologies. The role of the DNN is very important to the monocular system. It can abstract the vehicles seen by the camera and perform 3D modeling.

After parting with Mobileye, the future of Tesla's autonomous driving technology will shift to the stereoscopic camera system used by Mercedes. The stereo camera is composed of two cameras with a space of 15cm between them to simulate the eyes of the person and to determine the depth of field of the picture.

The details of Tesla's binocular system have not yet been announced. With reference to Mercedes-Benz's SMPC (stereo multifunctional camera) technology, this system can perform 3D modeling within 50m, distinguishing between vehicles, pedestrians, and signal lights, and its observation distance can reach 500m.

Compared to monoculars, dual cameras can create depth of field effects and are important for judging front objects and predicting trajectories. In this regard, the latter must be even better. Of course, the cost of the binocular system is also relatively high. Although there are differences in image recognition between these two types of systems, they all need to use DNN to continuously improve the optimization algorithm and enhance the accuracy of recognition.

In addition to the camera, the number of radar and ultrasonic sensors will increase, and the type will increase, providing more comprehensive 360° detection of the vehicle. In order to widen the application scene of automatic driving, Tesla or install an infrared camera to enhance its recognition ability at night.

At the same time, there is an important component for autopilot: HD MAP - to be accurate to centimeters. High-precision maps allow Tesla to accurately position itself, especially in the case of highway ramps.

Bid farewell to Mobileye, for Tesla there was not much loss. Mobileye offers just one SoC solution, and the core DNN is Tesla's own. After turning into the binocular camp, Tesla will create a multi-camera advanced version of Auto-Pilot like Mercedes.

This article from the Tencent News client from the media, does not represent the views and positions of Tencent News

Laptop Stand Keyboard

Laptop Stand Keyboard And Mouse,Laptop Stand Keyboard Set,Laptop Stand Keyboard Stand,Laptop Stand Keyboard Tray,etc.

Shenzhen Chengrong Technology Co.ltd is a high-quality enterprise specializing in metal stamping and CNC production for 12 years. The company mainly aims at the R&D, production and sales of Notebook Laptop Stands and Mobile Phone Stands. From the mold design and processing to machining and product surface oxidation, spraying treatment etc ,integration can fully meet the various processing needs of customers. Have a complete and scientific quality management system, strength and product quality are recognized and trusted by the industry, to meet changing economic and social needs .

Laptop Stand DjLaptop Stand Bed

Laptop Stand Keyboard And Mouse,Laptop Stand Keyboard Set,Laptop Stand Keyboard Stand,Laptop Stand Keyboard Tray

Shenzhen ChengRong Technology Co.,Ltd. , https://www.laptopstandsupplier.com