On the afternoon of the 2nd of the month, Shen Yiren posted a paper pointing out that after three years of training iteration of the neural network algorithm based on the vision scheme, Tesla is likely to achieve unmanned driving in limited scenarios.
Shen Yiren said that the early stage for Tesla to achieve this goal is that tesla car sales should continue to grow, and there are enough cars in motion to continuously train algorithms.
Tesla CEO Musk has stressed more than once that autonomous driving is close at hand, and one day, Tesla Cars can completely rely on itself to cope with various road conditions.
Industry insiders pointed out that the real difficulties of unmanned driving lie in perception, planning decisions, control and hardware. Among them, planning decision-making and control are more difficult, it requires more data accumulation, and it also needs to go to more scenarios to learn and gradually polish the technology.
Because driverless cars need to capture a lot of data through technologies such as cameras and lidar and provide real-time feedback, this data needs to be processed by trillions of operations per second.
Therefore, the high-speed processing power, reliability and accuracy of the product are particularly important, and it involves driving safety.
Baidu founder Robin Li has said that it will take a long time to achieve L5 (completely unmanned) level unmanned driving. In his view, it is expected that it will take about 10 years for autonomous driving to be completely open to the road without any restrictions. (Fast Technology)
![Feng Yufei, vice president of vivo, was rumored to have left his post, and Weibo certification has been canceled[fig]](data:image/gif;base64,R0lGODlhAQABAIAAAP///wAAACwAAAAAAQABAAACAkQBADs=)