The first year of mass production of high-level autonomous driving: A brief discussion on the market trend of "high-precision integrated navigation"

Combined navigation manufacturers are making tightly coupled, low-cost products, while car manufacturers and autonomous driving companies are beginning to develop their own GNSS and IMU algorithms.

2022 is known as the first year of mass production of high-level autonomous driving.

With the implementation of high-level autonomous driving functions, high-computing chips, laser radars, high-definition cameras, and high-precision maps have become indispensable components. In addition, combined navigation that provides high-precision positioning also plays a vital role.

Today we will analyze the technical prospects and market trends of "high-precision integrated navigation".

01. High-precision positioning is indispensable, satellite navigation + inertial navigation is the best combination

Autonomous driving technology mainly includes four modules: high-precision map module, positioning module, perception module, and intelligent decision-making and control module.

It can be said that high-precision maps, positioning, and perception all provide the basis for intelligent decision-making in autonomous driving.

Among them, the collected and produced high-precision maps are used to record complete three-dimensional road information; the satellite navigation system and the inertial navigation system are combined to achieve vehicle positioning with centimeter-level accuracy; traffic scene object recognition technology and environmental perception technology are used to achieve high-precision vehicle detection and recognition, tracking, distance and speed estimation, road segmentation, and lane line detection.

The positioning module, which is what we call the integrated navigation system, is mainly based on the satellite navigation positioning system (GNSS) and the inertial measurement unit (IMU). Combined with high-precision maps and a variety of sensor data, the positioning system can provide centimeter-level comprehensive positioning solutions.

Among all positioning modules, only the satellite navigation system can provide absolute position, while other sensors only achieve relative positioning.

Lane-level positioning, especially the elevation information of on- and off-ramps and viaducts for autonomous vehicles, is very important for map matching.

In addition to location, satellite navigation can also provide accurate time information and speed information, which are indispensable for high-level autonomous driving.

Currently, there are four major navigation systems in the world, including the United States' GPS, Russia's GLOGNSS, the European Union's Galileo and my country's Beidou.

Satellite navigation can output information such as latitude, longitude, and elevation in the WGS84 coordinate system. The Beidou satellite navigation system completed its global networking in 2020, announcing that Beidou can provide all-weather, all-day, high-precision positioning, navigation and timing services on a global scale.

The positioning principle of satellite navigation is that the satellite continuously sends its own ephemeris parameters and time information. After the user receives this information, the three-dimensional position, movement speed and time information of the satellite navigation receiver are calculated.

Through the operation of geometric relationships, it can be understood that, assuming that a satellite navigation receiver is placed at the ground test point at time t, the time △t when the GPS signal reaches the receiver can be measured. Adding other data such as the satellite ephemeris received by the receiver, the following four equations can be determined:

In the four equations, x, y, and z are the coordinates of the point to be measured, Vto is the clock error of the receiver which is an unknown parameter, among which di=c△ti, (i=1, 2, 3, 4), di is the distance between satellite i and the receiver, △ti is the time it takes for the signal of satellite i to reach the receiver, xi, yi, and zi are the spatial rectangular coordinates of satellite i at time t, Vti is the clock error of the satellite clock, and c is the speed of light.

The above four equations can be used to solve the coordinates x, y, z of the measured point and the receiver's clock error Vto, thereby determining the receiver's latitude, longitude and altitude information.

However, due to the interference of the troposphere and ionosphere in the atmosphere on electromagnetic signals, the positioning accuracy can usually only reach the meter level after calculation.

In order to achieve centimeter-level positioning accuracy, other means are needed to assist. Common means include ground-based augmentation systems (RTK carrier phase real-time differential positioning), precise point positioning PPP, satellite-based differential systems, etc. to eliminate common mode errors.

The advantages of satellite navigation products are that they can be used around the clock, there is no need for line of sight between observation points, and they have a wide range of action and high accuracy. However, their defects are also very obvious. They are greatly affected by the environment, including electromagnetic interference, multipath, etc., which will affect their accuracy. The most important point is that there should be no obstruction. In order to compensate for this problem, the inertial navigation system (INS) has become the closest partner of the satellite navigation system.

The inertial navigation system can provide real-time position and attitude information of mobile carriers and is a completely autonomous navigation method. The disadvantage of inertial navigation positioning is that the measurement error will accumulate over time and become larger and larger.

The combined application of satellite navigation and inertial navigation can just make up for the shortcomings of both. According to the different degrees of information exchange or combination, it can be divided into loose combination and tight combination.

The loose combination is based on INS. When GNSS is working, GNSS signals are used for optimal estimation of navigation information, and the optimal estimation results are used to feed back and correct INS to maintain high precision. When GNSS is unavailable, INS works alone and outputs inertial navigation solutions.

This combination scheme has been widely used in domestic integrated navigation systems. The results show that in areas where GNSS works well or when it loses lock for a short time, the output accuracy of the integrated navigation system is good, generally not lower than the GNSS accuracy.

However, when the carrier performs high-dynamic maneuvers or the GNSS receiver is affected by environmental interference and cannot work for a long time, the accuracy of the system will drop sharply as the system operating time increases, and the reliability and anti-interference capabilities will be poor.

Advantages: Simple calculation and easy to implement; the two systems work independently and the navigation information has a certain degree of redundancy.

Disadvantages: low accuracy and poor dynamic characteristics.

Compared with loose combination, tight combination is a two-way information transmission: on the one hand, GNSS signals (including raw data pseudorange and pseudorange rate) are used to correct INS; on the other hand, INS signals, with the assistance of satellite ephemeris, also calculate the pseudorange and pseudorange rate of the carrier relative to the GNSS satellite, and use this information to assist the reception of GNSS signals and the code loop phase-locking process, thereby enhancing the rapid capture and anti-interference capabilities of GNSS signals, and thus improving the receiver accuracy, dynamic performance and working reliability of GNSS.

When the observable data of GNSS navigation satellites is less than 4, useful information can still be output. However, if the same situation is encountered in the loose combination, the output information of GNSS is unavailable.

Advantages: High precision, suitable for high dynamic environment.

Disadvantages: The algorithm is complex and the design is difficult.

02. Autonomous driving is implemented in multiple scenarios, and the market prospects are promising

There are many scenarios for autonomous driving, including Robotaxi, Robobus, trunk logistics, ports, mines, low-speed delivery, cleaning, etc. Different scenarios have different requirements for combined navigation products, but the dual GNSS antenna solution is indispensable.

For example, the mining scene is relatively open but subject to greater vibrations, which requires earthquake resistance for the product hardware and filtering for the software algorithm.

The port scene environment is changeable and severely obstructed, so it is necessary to add UWB, signs, wheel speed sensors, etc. for multi-sensor fusion.

Scenarios such as low-speed delivery and cleaning are simple and have little change, so they have lower requirements for the accuracy of combined navigation products and rely more on SLAM algorithms.

The Robotaxi scenario is the most complex, and the early algorithm requires high combined navigation accuracy.

In addition to the algorithm, the accuracy of integrated navigation mainly depends on the accuracy of the inertial device IMU. Common IMU products are divided into fiber optic IMU and MEMS IMU. The accuracy of fiber-optic-level IMU devices is generally higher than that of MEMS-level devices. They are currently also mostly used for high-precision map acquisition and as the true value of the product.

MEMS-level combined navigation is the mainstream of L4 autonomous driving at present. The prices generally accepted by L4 autonomous driving manufacturers range from several thousand yuan to tens of thousands of yuan up to more than 100,000 yuan.

The core product suppliers mainly include BDStar, Huace, Daoyuan, NovAtel, Beiyun, Daishi, and StarNet Yuda.

At present, judging from the development trend of combined navigation products for L4 level autonomous driving applications, manufacturers of combined navigation products are gradually making tightly coupled and low-cost products.