Tesla humanoid robots enter the factory to "work" sorting battery cells: relying on pure vision to autonomously correct
David墨
发表于 2024-5-6 19:10:07
1319
0
0
Tesla has released the latest developments in the second-generation humanoid robot Optimus.
On May 5th, the official Tesla Optimus account posted a 90 second demonstration video of Optimus robots working in Tesla factories on social media platform X, with the accompanying text saying, "We are currently working hard to become useful!"
In the video, Optimus is sorting battery cells on the production line. The video shows that Optimus' end-to-end neural network has been trained to accurately package battery cells in Tesla factories; Optimus runs in real-time on the robot's FSD computer, relying solely on 2D cameras, hand touch, and force sensors; Optimus utilizes its legs to maintain balance while relying on neural networks to drive the entire upper body.
The video also showcases the working process from the perspective of Optimus and introduces that very precise actions are required in sorting operations, and the fault tolerance is very low. Optimus's neural network can automatically locate the next idle slot. In addition, Optimus can autonomously recover from failures.
The video also shows operators wearing VR glasses and gloves, collecting data through remote operation.
In addition, the video showcases Optimus walking in Tesla's office and introduces that Optimus is now not only walking faster, but also walking farther and farther.
It is worth noting that Tesla CEO Elon Musk also revealed on X that Optimus' hands will reach 22 degrees of freedom (the number of independently movable joints) later this year.
The human hand has 22 degrees of freedom (including the wrist). According to information previously released by Tesla, the Optimus hand has 11 degrees of freedom, while many competitors only have 6-7 degrees of freedom. The 22 degrees of freedom of Optimus means that its biomimetic degree will be further enhanced.
Subsequently, Optimus project leader Milan Kovac forwarded a video of Tesla Optimus and posted an article introducing the specific progress of Optimus.
He stated that in the past few months, Tesla has manufactured more robots and collected artificial intelligence data from them. At present, the team has trained and deployed a neural network that allows Optimus to start performing useful tasks such as picking up cells passed down from the conveyor belt and accurately placing them into the tray.
This neural network operates completely end-to-end, which means that it can directly generate joint control sequences using only videos from the robot's 2D camera and onboard ontology sensors. It runs entirely on the embedded FSD computer of the robot, powered by onboard batteries.
Kovac said that although Optimus is not perfect enough and its speed is still a bit slow, its success rate is getting higher and the frequency of failures is getting lower. The team is still training Optimus to recover from failed cases and has shown autonomous corrective behavior.
Kovac also emphasized that Tesla is currently working hard to make Optimus move faster and cope with more complex terrains, but all of this work will not sacrifice its "humanoid" nature. He revealed that more updates will be released soon.
Julian Ibarz, an engineer responsible for Tesla's robotic artificial intelligence, also posted that Tesla has improved the reliability of the entire Optimus team's learning behavior and decided to introduce some robots in the factory to perform a truly useful task: classifying batteries. The autonomous behavior demonstrated in the video can continue to run for several hours until the robot needs to be charged.
Jim Fan, senior scientist at NVIDIA and project leader for general embodied agent research, also praised Optimus' latest progress in a post. He stated that the Optimus team has released an important update, which has allowed us to learn about their "human data collection farm," which in my opinion is Optimus' biggest leading advantage.
The Optimus robot prototype was first unveiled on October 1, 2022 at Tesla's Second Artificial Intelligence Day. Musk has repeatedly emphasized that the Optimus robot project will be more valuable than the automotive business and FSD.
LogoMoney.com 系信息发布平台,仅提供信息存储空间服务。
声明:该文观点仅代表作者本人,本文不代表LogoMoney.com立场,且不构成建议,请谨慎对待。
声明:该文观点仅代表作者本人,本文不代表LogoMoney.com立场,且不构成建议,请谨慎对待。
猜你喜欢
- Tesla Beijing: Model 3 test drive volume increases by nearly 60% on a weekly basis
- Top 20 US Stock Transactions: Tesla Falls Nearly 5%, Domestic Car Companies Launch Price Reduction Promotion Strategy
- AI+sweeping the automotive industry, humanoid robots may form the second growth curve
- Tesla Cybertruck to go on sale in January next year? Official Refutation: False News
- NVIDIA's latest statement! Robot 'ChatGPT Moment' is Coming, Bet on the Next Growth Driver
- The seven giants of the US stock market all rose this year, and Tesla experienced a major reversal in the fourth quarter
- Nvidia plans to release a new generation of humanoid robot computing platform in the first half of next year, supporting multimodal AI models
- Tesla's Shanghai Energy Storage Gigafactory is expected to start mass production early next year
- Global markets: European and American stock markets close lower, Nasdaq, S&P down over 1%, Tesla down over 3%
- Tesla (Shanghai) Co., Ltd. and Tesla Motors (Beijing) Co., Ltd. recall some imported and domestically produced electric vehicles
-
工信部党组书记李乐成会见德国汽车工业协会主席希尔德加德·穆勒 4月27日,工业和信息化部党组书记李乐成在北京会见德国汽车工业协会主席希尔德加德·穆勒,双方就深化中德汽车产业合作进行了交流。李乐成表 ...
- moonlightplay
- 3 天前
- 支持
- 反对
- 回复
- 收藏
-
美国总统特朗普近日在接受媒体采访时表示,他第二个任期不仅治理美国,也治理全世界。 特朗普于4月24日接受了《大西洋》(The Atlantic)月刊采访,这段专访于4月28日发布。 “第一次当总统时,我要做两 ...
- lfancn
- 昨天 12:10
- 支持
- 反对
- 回复
- 收藏
-
4月29日凌晨,阿里巴巴开源新一代通义千问模型Qwen3(千问3),参数量为DeepSeek-R1的三分之一,成本大幅下降。据称,该模型性能全面超越R1、OpenAI-o1等领先模型,登顶全球最强开源模型。 千问3是国内首个“ ...
- 风雨中行走
- 前天 10:32
- 支持
- 反对
- 回复
- 收藏
-
东风有限回应武汉工厂关停事宜 据第一财经,4月29日,东风汽车有限公司证实,该公司武汉工厂目前正常运行,后续也不会关停。东风有限称,该公司将在东风与日产母公司的支持下平稳有序发展,持续加速向新能源 ...
- king19831101
- 昨天 09:56
- 支持
- 反对
- 回复
- 收藏