Mobileye's first mass-produced Robotaxi: Based on NIO ES8, equipped with Chinese-made LiDAR

At the 2021 Munich International Motor Show (IAA Mobility 2021), Intel, as a chip company, attracted a lot of attention. The reason is that its autonomous driving subsidiary Mobileye has made another big move.

Intel's newly appointed CEO Pat Gelsinger personally took the stage to announce many new developments in Mobileye.

Intel CEO Pat Gelsinger speaks at IAA Mobility 2021 in Munich, Germany

The most eye-catching one is the newly released Robotaxi based on the NIO ES8 model. Mobileye regards it as its first mass-produced Robotaxi ("First self-driving robotaxi").

This Robotaxi will be put into operation in Munich, Germany as early as 2022, and then go to Tel Aviv, Israel.

What’s the secret behind Mobileye’s first mass-produced Robotaxi?

(1) Model selection: NIO ES8

First of all, in terms of vehicle model selection, Mobileye chose the high-end mid-to-large SUV ES8 from China's new car brand NIO as the basic model.

Auto Heart heard that most of the taxi industry in Germany uses luxury medium and large vehicles, such as BBA models, so Mobileye also chose a luxury car with large space, ES8, as its operating vehicle.

Another reason is that in November 2019, Mobileye and NIO reached a strategic cooperation. The two parties planned to build an L4 autonomous driving model based on NIO's NT1 vehicle platform (models based on this platform include ES8, ES6, and EC6). In the end, Mobileye chose ES8;

Earlier, NIO's first mass-produced car, the ES8, was first equipped with Mobileye's EyeQ4 chip. For this Roboataxi, NIO mainly provides vehicles and related engineering services to Mobileye, while Mobileye is responsible for the autonomous driving modification.

We have learned that the first batch may be in the size of several hundred vehicles.

(2) Sensor configuration: Luminar + Yijing Technology LiDAR

In terms of sensor configuration, Mobileye has developed two autonomous driving perception systems for this Robotaxi: one based on lidar and the other based on camera. The two systems do not interfere with each other but are redundant to ensure the reliability of the vehicle's perception of the environment to the greatest extent.

It should be noted that Mobileye's lidar + camera solution is different from other companies' multi-sensor fusion solutions. Its lidar solution or camera solution can run independently to ensure the safety of autonomous driving. Mobileye calls it "True Redundancy".

In addition, this solution does not seem to have millimeter-wave radar.

Let’s take a closer look at the sensor layout:

First of all, the car is equipped with 7 long-range cameras, located on the front windshield, wheel arches and both sides of the top of the vehicle.
 

At the same time, the car is also equipped with 6 short-range laser radars and 3 long-range laser radars. There are two short-range laser radars at the front and rear of the car, and two short-range laser radars are located above the rear wheels; the roof is equipped with 3 long-range laser radars, located in the front and on the left and right sides.
 

In addition, the car is also equipped with 4 surround-view cameras and 1 TFL FOP camera.

Regarding the sensor configuration of this car, the outside world is most concerned about the supplier of LiDAR:

Mobileye CEO Amnon Shashua introduced that the three long-range lidars on the roof are Luminar's semi-solid-state lidar Iris, which has a field of view of 120 degrees × 26 degrees and a detection distance of 250 meters at a reflectivity of 10%. It uses a 1550nm wavelength laser and has excellent performance.

As for the six short-range laser radars on the vehicle body, the official has not been made public. AutoHeart has compared the appearance and confirmed with industry sources:

These six laser radars are produced by the domestic manufacturer Yijing Technology. The specific product model is ML-30s, which is a MEMS laser radar. The customized development version supplied to Mobileye is one.

In terms of specific parameters, as a short-range laser radar, ML-30s has a field of view of 140 degrees × 70 degrees; the equivalent line beam reaches 160 lines, and the angular resolution is less than 0.5 degrees.

Auto Heart learned from insiders that the cooperation between the two parties actually began in 2020.

Later, because the cooperation progressed smoothly, Intel Capital invested in Yijing Technology: On June 18, 2021, Yijing Technology announced the completion of hundreds of millions of yuan in Series B financing, and Intel Capital led the B1 round.

However, for Mobileye, purchasing third-party lidar should be a transitional strategy.

Because we know that Mobileye itself is already developing frequency-modulated continuous wave lidar (FMCW LiDAR) and imaging radar. Mobileye's goal is to create a more cost-effective sensor combination solution through self-research, paving the way for the future creation of consumer-grade L4 autonomous driving cars.

Auto Heart has learned that Mobileye's self-developed lidar will be launched together with its second-generation Mobileye Drive autonomous driving kit.

(3) Computing platform: 8 EyeQ5H chips, with a computing power of nearly 200TOPS

In terms of computing platform, Mobileye has specially built a computing platform AVKIT58 for this Robotaxi, which is embedded with 8 EyeQ5H chips.

The length, width and thickness of the entire device are 23×33×9 (cm) respectively. It has a high degree of integration and can be placed in the trunk without affecting passengers' placement of luggage.

The computing power of this computing platform is close to 200TOPS, and it needs to process real-time input from 9 lidars and 12 cameras. It can be seen that the utilization rate of computing power is still very high, but it is not yet known how high the power consumption of this platform will be.

In fact, in addition to the AVKIT58 computing platform specially prepared for Robotaxi, which integrates 8 EyeQ5H chips, Mobileye has also built many computing platforms for different application scenarios like building Lego.

GAS52 computing platform: Integrates two EyeQ5H chips to drive the SuperVision assisted driving system, which will be first installed on the Zeekr 001 model;
 

GAS253 computing platform: It integrates 6 EyeQ5H chips, can be connected to lidar and millimeter-wave radar sensors for overall integration, and is applied to L2++ automatic assisted driving systems. Zeekr's subsequent models will be equipped with this platform, which also means that Zeekr models will definitely be equipped with lidar in the future.

In terms of computing platforms, Mobileye currently mainly cooperates with some Tier 1 companies, who make integrated ECUs and then supply them to car companies. Of course, Mobileye itself also has ECU design and integration capabilities.

(4) Software system: Mobileye Drive

The autonomous driving software system installed on this Robotaxi is undoubtedly Mobileye's self-developed Drive system.

The Mobileye Drive system is also its most core technical capability, which includes powerful computer vision perception technology and, of course, multi-sensor fusion technology.

In addition, in addition to improvements in perception algorithms, Mobileye has also made continuous progress in planning and control algorithms in recent years.

The two core systems of the Mobileye Drive system are the perception system and the high-precision map system.

As we mentioned earlier, Mobileye has built two perception subsystems. One is a subsystem built by computer vision, and the other is a perception system built for sensors other than cameras. These two subsystems are redundant.

In some L2+ level models, the computer vision system is mainly responsible for perception, while in L4 level Robotaxi, both systems are required to work together to ensure safe and stable autonomous driving of the vehicle.

In addition to the perception system, Mobileye is also proud of its high-precision map system, especially the crowdsourced map REM system. With its visual ADAS system installed on more than 100 million vehicles worldwide, Mobileye can collect and map road environment data and output high-precision maps if permitted by local regulations. Moreover, this high-precision map system can maintain high-frequency updates.

There is also an often-discussed RSS Responsibility-Sensitive Safety model in the Mobileye Drive system. This is an open, transparent and verifiable formal model proposed in 2017 by Mobileye, which transforms human concepts and ideas about safe driving into a complete set of mathematical formulas.

To achieve true autonomous driving, the autonomous driving system must have a very clear sense of judgment during the driving process. For example, in different environments, the car itself must know that it has the responsibility to protect the safety of people in the driving environment. At the same time, it must know how to communicate efficiently and safely with other road users in some dangerous environments.

The above is the hardware and software secrets of Mobileye’s first Robotaxi. So, who will operate Mobileye’s Robotaxi fleet in the future?

Who will operate the Mobileye Robotaxi?

According to Mobileye's plan, it will first provide autonomous driving ride-hailing services in Munich, Germany in 2022. Mobileye will form its own Robotaxi fleet and initially purchase ES8 models from NIO.

Specific operations will be handled by Intel's subsidiary Moovit, which was acquired by Intel for US$900 million in May last year. The company mainly provides public transportation, bicycles, motorcycles, online car-hailing and time-sharing car rental services, and provides driving routes and navigation suggestions to hundreds of millions of people around the world.

Today, Moovit has become the vanguard of Mobileye's evolution into a travel service provider.

In addition to Moovit, Mobileye has brought in a new external partner at IAA Mobility 2021 - SIXT, a well-known German international travel service provider. The two parties plan to gradually promote autonomous driving ride-hailing services to the whole of Germany and other European countries in the next 10 years. Their first stop is Munich, Germany.