[ A Brief Discussion on Environmental Perception in Intelligent Driving ]
Foreword
Intelligent vehicles organically combine "intelligence" and "capabilities" on the basis of traditional automobiles. They require advanced sensors, controllers, actuators, and other devices, and integrate advanced technologies such as modern sensing, automatic control, computers, information and communication, and artificial intelligence. These devices and technologies enable intelligent driving vehicles to perceive complex environments, make intelligent decisions, collaboratively execute tasks, and share internet information. They can even form an intelligent transportation system with highways and traffic assistance facilities.
Intelligent driving technology enables vehicles to mimic the human thought process and behavioral capabilities of "perception-thinking-judgment-execution," thereby achieving assisted driving and even fully autonomous driving. This technology has significant research value and practical engineering application value in improving vehicle comfort and safety, increasing the utilization rate of traffic resources, reducing urban traffic congestion, and promoting energy conservation and emission reduction. Intelligent development is currently recognized worldwide as a key direction for the upgrading and transformation of the automotive industry, and intelligent driving technology will be the strategic high ground of future automotive technology.
As early as 1939, General Motors in the United States first exhibited the Futurama, a self-driving concept car. After the 1980s, some government agencies in Europe and the United States successively launched research programs on intelligent driving vehicles. Currently, driven by a new round of global industrial and technological revolution, research institutions, universities, and enterprises in various countries have launched a research boom in intelligent vehicle technology. In China, universities such as Tsinghua University and the National University of Defense Technology, as well as automotive groups such as FAW and SAIC, are researching intelligent driving technology. Meanwhile, internet-connected electric intelligent vehicle companies such as NIO and XPeng Motors are springing up like mushrooms after rain.
Table 1 Classification of Vehicle Intelligence Level
The Society of Automotive Engineers (SAE) classifies automotive intelligence levels into six levels, as shown in the table. A concise description of these six levels is as follows: L0 is traditional manual driving, entirely reliant on driver input; L1 provides driving support for one of the following: steering or acceleration/deceleration; L2 provides driving support for multiple of the following: steering or acceleration/deceleration; L3 involves environmental monitoring and all driving operations handled by the system, with the driver only providing response decisions; L4 requires no driver response to all system requests, but driving is limited by road and environmental conditions; L5 enables driving under all road and environmental conditions.
Environmental perception is fundamental to the research of intelligent driving technology. Vehicles use sensors to obtain information about the driving environment, such as the driving path, obstacles, and traffic signs, enabling them to make decisions, plan, and control operations. Therefore, environmental perception plays a crucial role in intelligent driving technology and is a prerequisite for achieving assisted driving and fully autonomous driving. The key aspects of intelligent driving environmental perception technology mainly involve three areas: sensors, environmental detection, and positioning and navigation. This article will introduce these three aspects in detail.
1. Sensor
Sensors are the tools for intelligent vehicles to perceive information. The sensor configurations of autonomous vehicles from three major international companies are shown in Table 2. The environmental sensors they commonly use generally include lidar, millimeter-wave radar, cameras, and ultrasonic sensors.
Table 2 Sensor Configuration for Autonomous Vehicles
LiDAR uses rotating mirrors to emit laser light and measures distance by measuring the time difference between the emitted and reflected light. 3D LiDAR mirrors have a certain range of pitch motion, allowing for the acquisition of high-precision 3D information about objects. However, LiDAR is susceptible to interference from rain, snow, and fog, and is expensive. Table 3 shows the parameters, models, and prices of currently mainstream LiDAR systems.
Millimeter-wave radar refers to radar that transmits wavelengths of 1-10mm and operates in the 30-300GHz frequency range. Utilizing the Doppler principle, millimeter-wave radar offers excellent detection of moving targets, is unaffected by adverse weather conditions, and exhibits good penetration and propagation characteristics through smoke, dust, rain, and fog. However, millimeter-wave radar typically has a relatively small field of view, and millimeter waves experience significant attenuation during atmospheric propagation. Common millimeter-wave radar parameters, models, and prices are shown in Table 4.
Cameras are primarily used for lane detection, obstacle recognition, and traffic sign identification. While cameras are inexpensive and the technology is relatively mature, they are significantly affected by ambient light and struggle to accurately obtain 3D information. These issues are being addressed through ongoing research, such as using infrared cameras to reduce the impact of light and employing binocular or tri-lens cameras to acquire 3D information. Table 5 shows the specifications, models, and pricing of mainstream cameras.
Table 3 LiDAR
Table 4 Millimeter-wave radar
Table 5 Cameras
Ultrasonic waves are mechanical waves with frequencies higher than sound waves. Their energy attenuates significantly when propagating through air, therefore ultrasonic radar is mainly used for short-range obstacle detection where high sensing accuracy is not required. There are many types of ultrasonic sensors available at low cost; typically, multiple ultrasonic sensors are installed on a vehicle to avoid collisions at low speeds.
Each sensor has its own advantages and disadvantages. In order to perceive the environment more accurately and avoid being affected by the climate, multiple sensors are usually combined in practical applications.
2. Environmental monitoring
Intelligent driving vehicles need to detect road information and obstacles based on sensors, and calculate the relative distance between the obstacles and the vehicle. The targets that intelligent vehicles need to detect and the corresponding sensors are shown in Table 6.
Table 6 Detection Targets and Sensors
During driving, most environmental information comes from vision. Vision systems offer a wide field of view and are low-cost, making vision-based environmental detection a major direction for intelligent driving detection technology. The main technologies used for target detection in intelligent driving vehicles include background subtraction, feature matching, and statistical learning. Their basic principles and characteristics are shown in Table 7.
There is usually a relative motion between intelligent vehicles and the detected target. Moving target tracking technology can ensure the real-time nature of environmental perception. Currently, the main tracking technologies are the CamShift algorithm and the tracking method based on Kalman filtering. Their basic principles and characteristics are shown in Table 8.
Table 7 Main Target Detection Technologies
Table 8 Main Target Tracking Methods
Table 9 Four Major Global Satellite Systems
Table 10 Major Positioning and Navigation Technologies
3. Positioning and Navigation
Intelligent driving vehicles not only need to detect information about the surrounding environment during driving, but also need to obtain the relative positional relationship between the vehicle and the external environment and determine the absolute position. Positioning and navigation are also key technologies for environmental perception.
Satellite systems play an important role in the positioning and navigation of intelligent vehicles. Currently, there are four major global satellite systems: the US GPS, my country's BeiDou satellite system, Russia's GLONASS, and the EU's GALILEO. Their performance comparison is shown in Table 9.
The most widely used satellite system currently is GPS, which provides continuous, real-time, and high-precision three-dimensional position, velocity, and time information to users worldwide, offering high positioning and speed measurement accuracy. However, in mountainous areas, tunnels, and urban environments with numerous high-rise buildings and elevated highways, GPS signals are susceptible to interference. If a signal is lost, the re-acquisition time is long, severely impacting vehicle positioning and navigation. Other satellite systems also have their own inherent problems; therefore, other technologies are typically combined with satellite systems to improve positioning and navigation accuracy and ensure system reliability. Commonly used positioning and navigation technologies are shown in Table 10.
4. Summary
Intelligent vehicles are highly complex systems, and the primary role of environmental perception systems is to provide accurate and reliable information for their behavioral decisions and path planning. As intelligent driving technology matures, intelligent vehicles will gradually be applied to numerous fields, such as transporting goods in mines and logistics plants, and commuting employees in industrial parks. Companies like JD.com and Alibaba, along with automotive e-commerce entrepreneurs, are exploring new business models and application scenarios. In the near future, the integration of intelligent vehicles and commerce will bring about revolutionary changes to people's consumption habits and behaviors. Researching intelligent driving vehicles will not only promote rapid scientific and technological development and facilitate the rise of emerging industries, but also offer new possibilities for addressing pressing social issues such as more effectively utilizing social resources, alleviating traffic congestion, and reducing energy consumption. For China, developing intelligent vehicles is a hot trend and a significant opportunity for the development of the country's automotive industry. The National Development and Reform Commission, together with relevant parties, is formulating policies and regulations specifically for intelligent vehicles to accelerate their innovative development in my country.
