Popular Science ｜ The ＂Brain＂ FSD of Tesla Electric Vehicle

    Xinhua News Agency, Beijing, May 1 (Xinhua) Recently, Tesla, an American electric vehicle manufacturer, issued a document in the official Weibo, saying that it would accelerate the landing of clean energy and autonomous driving technology, which triggered the market’s attention to Tesla’s "fully autonomous driving (FSD)" system. Then what is FSD? How is it different from the traditional assisted driving system?

    In recent years, many automobile manufacturers are developing automatic assisted driving systems, which make the on-board computers share more driving tasks, and Tesla is one of them. Tesla CEO elon musk once said: "Autopilot is a good thing on the plane, and we should also equip it on the car."

    FSD is the abbreviation of automated assisted driving system developed by Tesla, and its beta version was launched in 2020. The ultimate goal of the system is to realize that vehicles can identify road conditions, plan routes, control speed and direction independently without human intervention, and complete the driving from the starting point to the end point.

    However, some people have questioned the name "fully automatic driving", because according to the automatic driving classification standard of the National Highway Traffic Safety Administration of the United States, the vehicles currently running under this system are roughly in Level 2 (L2), that is, partially automated, but drivers still need to actively monitor road conditions and be ready to take over at any time. The truly fully automatic driving without the intervention of the driver is level 5 (L5), which can only be said to be the goal of future development.

    It is reported that the functions that can be realized by the current FSD system include automatic emergency braking (AEB), automatic lane change and traffic navigation, and also increase the ability of semi-autonomous navigation in urban streets and responding to traffic lights or stop signs.

    Many subsystems that make up FSD, such as automatic emergency braking, have matured and even become the common configuration of cars, and the National Highway Traffic Safety Administration of the United States requires it to become the standard for new cars. The composition of other manufacturers’ assisted driving systems is similar, but different from the traditional automobile manufacturer’s development system, Tesla’s FSD is based on neural network artificial intelligence algorithm, which imitates human driving through a lot of training and makes corresponding decisions and operations. Its automatic driving algorithm directly outputs control results according to the information sensed by sensors, imitates the operation of human drivers, and changes the "rule-driven" algorithm into a "data-driven" algorithm, thus simplifying model training.

    Tesla pushed the latest version of FSD V12.3.1 to North American car owners in March this year. Musk said in the previous test drive live broadcast that the FSD V12 version has been able to achieve an "end-to-end" artificial intelligence autopilot solution. Simply put, the scheme can directly output vehicle control instructions such as steering, acceleration and braking after inputting the image data obtained by the camera into the algorithm, which is more like a human brain, and can analyze and output control strategies only by relying on image data input. Previously, Tesla’s autonomous driving system relied on rule-based judgment, relying on car cameras to identify lanes, pedestrians, vehicles, signs and traffic lights, and then engineers wrote codes to deal with various situations.

    However, Tesla’s FSD system is still iterative and only available in the United States, Canada, Australia and some European countries. If FSD is to make continuous progress in the future, its security, data privacy, supervision and other challenges also need to be solved urgently.