Key staff resigned, code was stolen, Musk's "fully autonomous driving" will be delayed!

Publisher:数据梦行者Latest update time:2019-07-11 Keywords:AutopilotM Reading articles on mobile phones Scan QR code
Read articles on your mobile phone anytime, anywhere

Tesla publicly promised earlier this year that by the end of next year, its electric vehicles will have the function of " robotaxis " - that is, they will be fully autonomous under normal circumstances and do not require human intervention. Its CEO Elon Musk said that by 2020, there will be 1 million Tesla cars on the road for "robotaxi" business. Tesla is currently at a critical stage in the development of autonomous driving. According to foreign media reports, several key engineering managers who worked on the development of Autopilot semi-autonomous driving functions left after Musk expressed dissatisfaction with the progress of developing fully autonomous driving functions, according to people familiar with the matter.

Musk19071102.jpg

According to foreign media The Information, in the 18 months between 2016 and 2018, Tesla lost three Autopilot (automatic assisted driving system) software team leaders, and several lower-level managers and engineers also left. In the past dozen months, at least 11 members of Tesla's Autopilot software team have left, accounting for nearly 10% of the entire team, including several long-term members. The remaining members are now directly managed by Musk.


The departures come after Tesla Inc Chief Executive Officer Elon Musk replaced Autopilot software team leader Stuart Bowers in early May. Bowers shared responsibility for the Autopilot team with artificial intelligence (AI) expert Andrej Karpathy, with Bowers focusing on operational issues and Karpathy as the technical lead.


After the personnel changes, Ashok Elluswamy was promoted to lead the perception and computer vision team; Milan Kovac was responsible for firmware and embedded systems engineering. CJ Moore, head of quality control, took over the simulation team previously led by Goldstein. Drew Baglino took over Havelak's team. Baglino has worked at Tesla for 13 years and has been responsible for the engineering of batteries and other components.


In addition to the departure of key members of the Autopilot software team, its head of perception technology, Zeljko Popovic, has also been reported to have left (the news has been confirmed). Popovic built and ran the perception team for Tesla's Autopilot division. He was responsible for developing high-precision US highway maps for Tesla and created a " sensor fusion system" that combines data from the many cameras, radars, and ultrasonic sensors used by Tesla cars. The sensor fusion system enables Autopilot to "see" other vehicles on the road. Popovic's technical background is in robotics.


The reason for leaving is the inability to achieve "full" autonomous driving?


According to people familiar with the matter, several Tesla team members told Musk that it would not be possible to achieve full self-driving capabilities on time, which made Musk very unhappy.


Musk's propaganda that "Tesla cars will achieve full self-driving within a year" is not an easy task. Although in April this year, Tesla successfully demonstrated the fully autonomous driving function to investors near its headquarters. According to an employee, the demonstration was of limited significance because it was limited to a specific route. The company regularly updates the Autopilot software. An industry insider said, "It is very challenging to support 500,000 cars on the road and try to add some new features."


In fact, the safety of Autopilot's existing autonomous driving functions is still questionable. At present, its automatic lane change function on highways is not perfect. Fixing them is like a game of whack-a-mole. Tesla owners sign an agreement to provide feedback to Autopilot, confirm the flaws, and forward them to the team for repair.


The current Autopilot often encounters problems such as failure to automatically change lanes. In some cases, the cars cannot identify vehicles that may block the driving path, forcing drivers to take control of the vehicle to avoid a collision, the employee said.


Musk is trying to position Autopilot as an urban self-driving solution, which is more technically difficult and carries a higher risk of collisions. The current employee said that lane markings on city streets appear and disappear, which poses more problems. The Autopilot system often doesn't know how to identify parked cars on the side of the road because it is not a common feature of highway driving functions, this person said.


Progress in the Autopilot source code theft case: The parties involved admitted the incident but denied stealing secrets


Tesla sued several former employees in March for helping competitors steal secrets, including one case in which a former employee stole trade secrets related to Tesla's Autopilot driver-assistance feature before joining Chinese electric car maker Xpeng Motors. Now there's progress.


Tesla's lawsuit against former Chinese employee Guangzhi Cao for stealing Autopilot source code and providing it to Xiaopeng Motors (Electronic Engineering Times reported this case on March 22: Tesla sued former employee for stealing Autopilot source code and joining Xiaopeng Motors ) is still under review. According to foreign media reports this morning, the former Tesla employee admitted in a document submitted to the court this week that he uploaded a compressed file containing Autopilot source code to his personal iCloud account while he was still working for Tesla at the end of 2018.

Musk19071103.jpeg

But Cao Guangzhi denied stealing sensitive information from Tesla in the same document. His lawyer argued that Cao Guangzhi "made extensive efforts to delete or remove Tesla-related documents before leaving Tesla." It is reported that Cao Guangzhi is now the "perception director" of Xiaopeng Motors and is "developing and delivering autonomous driving technology for production vehicles."


In new court documents, Cao Guangzhi admitted to "using his personal iCloud account to create backup copies of certain Tesla information in 2018." He also admitted to creating a zip file containing the Autopilot source code in late 2018 and confirmed that Xiaopeng Motors sent him a job offer on December 12. He said he disconnected his personal iCloud account from the computer provided by Tesla "on or about December 26" and remained logged in to the Tesla network between December 27, 2018 and January 1.


While Cao Guangzhi did not specify when he officially accepted the job at Xpeng Motors, Tesla said his last day of work was January 3.


Cao Guangzhi "also acknowledged that he deleted certain files stored on Tesla computers and cleared browser history, but denied that these activities constituted any form of misconduct." He said he did a lot of work to delete files on his personal iCloud, but he did not say whether he deleted all of them.


Cao's lawyers argued that any source code or other confidential information that remained on his devices after he left Tesla was simply "inadvertent." They also argued that Cao "did not access or use any 'Autopilot trade secrets'" or transfer any information to Xpeng Motors after leaving the company.


Will “fully autonomous driving” ever become a reality?


Tesla is currently at a critical stage in the development of autonomous driving. The Autopilot team is facing both internal and external troubles. The industry has had great doubts about when "fully autonomous driving" cars will be "truly" realized. The actual situation is that Tesla's autonomous driving is still hovering at the level of Level 2 and Level 3. Industry experts believe that Musk is exaggerating because he has never been able to provide any substantial evidence to prove that Autopilot has Level 4 or Level 5 capabilities.


Although in recent years, with the accelerated integration of new generation information technology and the automobile industry, domestic and foreign companies have accelerated the development and testing of driverless cars. Even though driverless cars have become a hot topic many times, the editor is not optimistic that it is still some distance away from our real life, at least not as fast as optimists "imagine." Therefore, the radical Musk's "fully autonomous driving without human intervention" will not appear for the time being, let alone "In 2020, there will be 1 million Tesla cars on the road for 'autonomous driving rental' business.

 


Keywords:AutopilotM Reference address:Key staff resigned, code was stolen, Musk's "fully autonomous driving" will be delayed!

Previous article:Velodyne Highlights Advanced Lidar at Autonomous Vehicle Symposium
Next article:University of Michigan develops new software to improve computer vision capabilities of self-driving cars

Recommended ReadingLatest update time:2024-11-16 15:38

Tesla is researching AI-based autonomous driving planning control
Elon Musk recently announced that Tesla is working hard to overcome the final hurdle in autonomous driving technology: vehicle control. He once again predicted that Tesla will achieve fully autonomous driving by the end of this year. As part of this plan, Tesla is moving toward relying more on neural networks to con
[Automotive Electronics]
Tesla is researching AI-based autonomous driving planning control
Latest Automotive Electronics Articles
Change More Related Popular Components

EEWorld
subscription
account

EEWorld
service
account

Automotive
development
circle

About Us Customer Service Contact Information Datasheet Sitemap LatestNews


Room 1530, 15th Floor, Building B, No.18 Zhongguancun Street, Haidian District, Beijing, Postal Code: 100190 China Telephone: 008610 8235 0740

Copyright © 2005-2024 EEWORLD.com.cn, Inc. All rights reserved 京ICP证060456号 京ICP备10001474号-1 电信业务审批[2006]字第258号函 京公网安备 11010802033920号