The battle for the autonomous driving industry is about to break out, and LiDAR has become a "gold mine" behind it

Three years ago, Innoviz launched two MEMS solid-state lidar samples: InnovizPro and InnovizOne.

InnovizPro is a high-performance solid-state solution based on MEMS scanning technology that delivers outstanding performance and value for automotive, mapping and other applications.

InnovizOne is an automotive-grade solution that is easy to integrate seamlessly into vehicles and provides 3D perception capabilities for Level 3 to Level 5 autonomous driving. In 2021, BMW decided to use InnovizOne lidar on Level 3 autonomous driving cars.

From the past high-profile promotion of L4/L5 autonomous driving to the current development of more lidar solutions with ADAS functions, lidar companies have adjusted their market strategies by "sinking" their products:

At the software level, more emphasis is placed on ADAS functions and developing killer applications (e.g. encoding laser beams);

At the hardware level, there are both lidar solutions suitable for L3 and below, as well as products suitable for L4/L5 autonomous driving.

The purpose of doing so is that LiDAR companies will launch products that meet market demand in accordance with the existing route and development speed of LiDAR technology. On the other hand, they will cooperate with first-tier OEMs and Tier 1 to gradually promote the evolution of autonomous driving.

4. Born for active safety, LiDAR is essentially a 3D sensor

By adding LiDAR to improve the safety redundancy of the ADAS system, this leads to the solution of LiDAR: it is cheap enough to be installed in front of the vehicle. In this way, LiDAR can obtain valuable installation opportunities.

This actually brings a double benefit:

First, the investment is actually shared by consumers (of course, consumers also get better ADAS functions and experience);

Secondly, real scenarios replace simulations to obtain real massive data.

In order to make it easier for car manufacturers to accept LiDAR and reduce the difficulty for them to install LiDAR on mass-produced models, it is a natural choice for LiDAR companies to provide ADAS functions for car manufacturers to use.

Since the LiDAR point cloud is 3D and contains information such as distance, size, and position, the amount of data is less than that of vision. Therefore, the technical difficulty of LiDAR companies providing ADAS functions is relatively small and does not require complex image processing.

Take Bosch MPC2 as an example (as shown above), its camera outputs a frame of 1920*1080 pixels, and each frame represents 2 million spatial points (the angular resolution corresponding to each pixel is 0.03 degrees).

Taking the Velodyne 128-line lidar as an example, its angular resolution is 0.2*0.11 degrees. After scanning one circle, the number of points is only 230,000 spatial points (horizontal 360 degrees/0.2 degrees*vertical 128 pixels).

Therefore, there is no need for complex image processing and the computing power requirement is extremely low. LiDAR can use its own embedded processor to complete ADAS algorithm processing.

Fourteen years ago, this new radar sensor invented by David Hall became famous in the DARPA autonomous driving challenge and indirectly promoted the vigorous development of autonomous driving.

Today, universities, car companies, Tier 1s, technology companies, and startups are all willing to pay for this sensor when conducting exploration and research and development related to autonomous driving.

But this has left a stereotype in the outside world: that lidar is a sensor specially designed for high-level autonomous driving.

In fact, lidar is not exclusive to L4/L5 autonomous driving.

From today's perspective, we can have a new understanding: LiDAR is a 3D sensor designed for intelligent driving and active safety.