Tesla Obtains Patents for Automatic Driving Data Pipeline and Deep Learning System

2019-12-31

Tesla's neural network seems to be constantly improving and evolving every day, as evidenced by a recent patent that will make Tesla's autonomous driving system more efficient.


Tesla released a patent titled "Data Pipeline and Deep Learning System for Autonomous Driving" on December 26th. The purpose of this patent is to revolutionize and improve the deep learning system used in autonomous vehicles, which utilizes "captured sensor data" to retrieve information.As the data becomes more complex, Tesla realized the need for a customized data pipeline that can maximize the signal information from the captured sensor data and provide higher-level signal information to the deep learning network for deep learning analysis.


The system described in this patent will use all the sensors or cameras on the vehicle to capture images. In this case, it will describe high dynamic range cameras, camera sensors, radar sensors, or ultrasonic sensors. The images will then be decomposed using "high-pass" or "low-pass" filters, and a series of processors will interpret the meaning of the images.


The flowchart below describes the process of how the vehicle learns information."Receiving sensor data" is the first part of this process. The data is then decomposed and preprocessed, and the "deep learning analysis" begins. The results are then passed to the vehicle's artificial intelligence processor for use during vehicle control.


In another process, a series of information retrieved from these images is compared to data collected from other Tesla users worldwide. This will alleviate concerns of drivers about the system executing erroneous processes during autonomous driving. The purpose of this patent is to create a safe driving experience and improve the already reliable performance of Tesla's autonomous driving software, achieving it in a more efficient process than before.By using this process, Tesla is able to maintain as high resolution as possible from the images captured by its vehicle's cameras and sensors. This allows the neural network to learn more efficiently from the data it receives. It also enables the neural network to process better images more efficiently, opening the door to faster autonomous driving. Tesla's Hardware 3 computer provides additional power, which will be very impressive, as it is specifically designed for fully autonomous driving with built-in redundancy.


Building on the foundation Tesla has already laid for its full self-driving suite, the latest patent indicates that the company is now attempting to narrow the performance gap of its software. The addition of this patent will not only create a safer driving experience for Tesla owners, but also bring the future of rapidly advancing fully autonomous vehicles closer to reality.

RECENT AI NEWS

RECENT AI TOOLS