Expectations are high with the launch of 'FSD Beta Version 9'… Safety issues and other challenges remain.
Tesla CEO Elon Musk announced on September 22, 2020 (local time) at the annual general shareholders' meeting and "Battery Day event" held at the Fremont plant parking lot in California.
Tesla's Full Self Driving beta version 9 was finally released on July 10. The newly released FSD beta version 9 is a standard software for self-driving drivers designed to help vehicles maintain lanes and recognize traffic conditions on roads using cameras installed in cars.
If the current autopilot functionality performs a limited function of adjusting the distance between vehicles within the same car, FSD beta version 9 provides more advanced functions such as automatic change to car, automatic parking, and vehicle call.
As soon as the beta version was released, Tesla's stock price soared 4.38 percent, demonstrating market expectations for self-driving cars. Some hot-tempered Tesla beta test drivers wirelessly updated the beta version and drove out at dawn to perform a driving test live on YouTube.
According to the current YouTube video, the prevailing view is that it is generally easy in terms of performance. Tesla's vehicle, which updated FSD beta version 9, drove smoothly on foggy roads, detected vehicles approaching from the opposite direction, and carried out a right turn smoothly. While driving, the upward light automatically turned off and on and seemed to be running well according to the road conditions.
FSD beta release 9 features
What stands out most is the so-called "mind of car" driver interface (UI). This new visualization view feature provides a computer vision of how cars perceive the surroundings of roads, allowing drivers to better understand self-driving situations.
Previously, the appearance of peripheral objects on navigation was a simple box-like rough graphic, but the new navigation screen shows the vehicles, people and traffic lights on the road quite realistically in the form of rendering. For example, roads and outer boundaries are marked in red and mid-up and down boundaries are marked in purple. In addition, the lines that should not be crossed on the actual road are yellow and appear red when the car in front is braking.
This is possible because of the Pure Vision, Tesla's camera-only auto-controlled auxiliary driving system. Also known as Tesla Vision, the system captures the surrounding driving environment with eight cameras and deep neural networks (DNN), collects data and recognizes surrounding vehicles.
To this end, Tesla has already removed electromagnetic-based radar sensors from Model 3 and Model Y in North America. This means that Tesla no longer relies on radar for autonomous driving. In addition, Tesla unveiled a supercomputer for self-driving with Nvidia A100 GPU in order to use it for deep neural network training that is 1000 times better than radar sensors in June.
Tesla is known to rely on camera vision and deep neural networks for fully autonomous driving because the use of radar and lida degrades the computing power of electric vehicles, reducing power efficiency. In addition, radar lacks precision and requires high-resolution mapping to use lida, which is also said to be expensive to build and maintain.
However, there is still a lot of controversy over eliminating radar and radar and implementing autonomous driving with only cameras. Above all, cameras are affected by the weather and it is difficult to measure distance. In fact, the National Highway Traffic Safety Administration (NHTSA) and the Highway Safety Insurance Association (IIHS), the U.S. NCAP agency, are expected to recover the highest safety rating given to Model 3 and Model Y. This is due to concerns that removing radar from the two models may prevent safety devices such as collision prevention from working properly. On the other hand, radar is excellent at providing data even in bad weather such as heavy rain or heavy snow at night when visibility is low.
For this reason, global competitors such as GM Close, Google's Waymo, and Aurora are using lasers to measure distance and create accurate real-time high-resolution 3D maps around vehicles.
Fully autonomous driving is still progressive.
Despite various controversies, the release of FSD beta version 9 is expected to be a turning point in advancing the era of full self-driving in the future. In addition, if FSD 9 Subscription service, which is expected to be released on Tesla's "Artificial Intelligence (AI) Day," is released, synergy between self-driving technologies and business models will be possible.
The FSD subscription service is expected to gain more power to secure customers as it is intended to expand accessibility for those who want to use FSD but cannot purchase FSD packages immediately. With the FSD subscription option, Tesla owners or tenants only have to pay for FSD features for a limited period of time or if necessary.
From a long-term perspective, what Tesla is trying to achieve through its subscription service called self-driving can be seen as an evolutionary automobile company as software. If the existing automobile industry is a manufacturing-based hardware product, Tesla's autonomous car subscription service is evolving into a service.
Some argue that if Tesla's FSD beta version 9 is released and self-driving cars based on pure computer vision are released, level 5 will be possible within this year. Some say that with the release of FSD beta version 9, it is getting closer to the fourth stage of autonomous driving. Tesla's CEO Elon Musk also claimed that five levels of fully autonomous driving would be possible at the end of this year when announcing quarterly results earlier this year.
Despite these optimistic claims, however, in reality, Tesla will have to wait more patiently for its goal of fully autonomous driving to come true.
Currently, there are six stages of self-driving technology from 0 to 5, and there are five stages of complete self-driving that do not require drivers. Currently, the level of self-driving technology is mostly level 2.5, and there are no complete self-driving vehicles that correspond to level 5 levels yet, and driver intervention is still required. In particular, Musk also submitted his opinion to the California Automobile Administration in May that full autonomous driving technology is unlikely to be ready by the end of this year.
Furthermore, in order for true self-driving to be implemented, it does not just need to have self-driving functions of cars. An overall and cooperative self-driving ecosystem with infrastructure such as driving environment and communication surrounding cars, cameras, lida, radar, AI, and ethical and legal devices that recognize them should be premised.
Above all, it is a safety issue that is holding Tesla back, which is busy on its way. This is because accidents due to self-driving errors are continuing. Recently, a Tesla driver who posted his social media video without taking the wheel using an autopilot function died in a deadly crash in Southern California. Currently, there are 24 cases of Tesla being scrutinized by the U.S. National Highway Traffic Safety Administration for autopilot-related crashes.
This is why some predict that full self-driving services will be introduced in the order of logistics, urban and urban transportation, and passenger transportation, which are relatively free from safety issues in the future.
댓글