Xpeng Changes Gear After Stealing Tesla Autopilot Coding, “They Don’t Have Our NN Inference Computer,” Says Elon Musk

Featured Image Credit: Wikimedia Commons 

Chinese electric vehicle manufacturer Xpeng is set to add LiDAR to its self-driving cars in 2021. The company said the use of LiDAR will significantly improve positioning accuracy relative to other vehicles and objects, helping to avoid collisions even in the absence of sufficient lighting, the South China Morning Post reported.

Xpeng is having great difficulty in developing appropriate software on its own that will give their car access to full self-driving in the future. LiDAR is a step backwards, and this is due to the fact that their Autopilot team does not know how to do the work for the further development of the technology--tech which they previously presented as their own.

In 2019, the California-based company sued Xpeng, claiming that the Chinese company stole and uses Tesla Autopilot source code in their cars. A former engineer from Tesla's Autopilot team Guangzhi Cao is under suspicion of the crime, who, while working for Tesla, downloaded the source code of autonomous systems onto his personal device. After the data was taken, he got a job at Xpeng. Cao even admitted that he did download some of the Tesla Autopilot source code to his laptop, but claimed to have deleted the data before quitting.

Cao was recruited to Xpeng along with a former Apple engineer who is also accused of transferring trade secrets.

The fact that Xpeng is now having problems with further development of Autopilot indicates that they do not have the necessary knowledge and know-how in this technology, and simply cannot go beyond what was stolen by Cao. Essentially they have an old version of Tesla software and don’t have Tesla's NN inference computer. That is why Xpeng plans to implement LiDAR in its vehicles.

They have an old version of our software & don’t have our NN inference computer

— Elon Musk (@elonmusk) November 20, 2020

Tesla relies heavily on Vision and doesn't use LiDAR. In addition to the high cost, LiDAR does not sufficiently cope with the tasks necessary for self-driving.

Everything that we see on the road is filled with visual information. All signs, turns, and intersections help us navigate. These are all stationary objects that LiDAR can detect. But problems begin to arise when moving objects appear on the road. People, dogs, flying plastic bags--these are all objects that we often encounter on the road and LiDAR cannot determine how they move, and even what kind of objects they are. This can cause accidents on the roads.

At the same time, Tesla's cameras and radar system are able to determine what an object is. As soon as an object enters the field of view, cameras determine what the object is, and then the car reacts to the situation. Tesla's Neural Network is an important component. The system can use the provided data to make rational decisions. And LiDAR has no such adaptability at all.

Tesla has managed to collect a huge amount of complex and unpredictable road data--which will continue to be gathered and processed. This is how the system constantly learns and improves. The more Tesla vehicles on the road, the more data is collected, and the better Autopilot becomes. But, Xpeng does not understand and does not know how to do this, which is why they are falling back on the expensive and insufficient aid of LiDAR.

© 2020, Eva Fox. All rights reserved.

_____________________________

We appreciate your readership! Please share your thoughts in the comment section below.

Article edited by @SmokeyShorts, you can follow him on Twitter

Follow @EvaFoxU