There is no real “full-stack self-research” – Automobile companies’ autonomous driving strategy report

More and more car companies attach importance to technological independence and increase their promotion of "full-stack self-research" for autonomous driving . From the literal understanding of "full-stack self-research", it turns out that autonomous driving requires upstream and downstream industry chain Multi-party collaboration, and a certain car company has achieved one company to cover all modules of autonomous driving; but this is not the case. Car companies that claim "full-stack self-research" cannot completely cut off cooperation with the supply chain.

So what exactly does “full-stack self-research” mean? Can current car companies call it "full-stack self-research"? Under the trend of "full-stack self-research", what do car companies need from external suppliers? ...

With the above questions in mind, this report will split the "full stack" and analyze it item by item.

In order to ensure that the opinions or conclusions of the analysis report are more accurate, we communicated and discussed with a number of experts who are deeply involved in the intelligent driving industry, and finally came up with this report.

This report is mainly divided into 3 major chapters:

The composition of the “full stack” of autonomous driving

The meaning of each module of "self-developed" by car companies

Laying out "full stack" is not the same as "self-researching the whole stack"

The following is the main content of the report: Analysis of trends in autonomous driving self-research by car companies - Table of Contents

1. The composition of the “full stack” of autonomous driving

1.1 What is "full stack"

2. The meaning of “self-developed” modules by car companies

The main reasons why car companies develop their own modules

2.1 Application algorithm layer

• Development status of various functional modules of autonomous driving

• The main reasons why car companies develop their own algorithms

• The layout of autonomous driving algorithms by different car companies

2.2  Middleware

• Controversy over whether car companies should develop self-driving middleware

• Reasons for self-research—some suppliers’ products have many problems

• Difficulties in self-developed middleware by car companies and possible future layout methods

2.3 OS kernel

• The main reason why car companies develop their own cores

• The main difficulties faced by car companies in developing self-developed autonomous driving OS kernels

• Automobile companies’ choice of different autonomous driving OS kernels

2.4 SoC chip

• The main reason why car companies develop their own smart driving chips

• The capabilities required by car companies to develop self-developed smart driving chips

• The division of labor model of self-developed chips by different car companies

2.5  Domain Controller

• The main reason why car companies develop their own autonomous driving domain controllers

• The capabilities and investment required by car companies to develop self-developed smart driving chips

• How different car companies deploy autonomous driving domain controllers

2.6 Toolchain

• Reasons for self-research—some suppliers’ products have many problems

• Automobile companies’ layout of data closed-loop tool chain

3. Laying out "full stack" is not the same as "self-researching the whole stack"

3.1 “Full-stack self-research” is difficult to reflect the actual layout of car companies in autonomous driving

3.2 "Full stack" is just a transitional form

3.3 Economic accounts in “full stack self-research”

3.4 There is no real “full stack self-research”