A concept
One concept refers to autonomous driving, also known as driverless cars, computer-driven cars, or unmanned vehicles. It refers to vehicles that use advanced driver assistance systems (ADAS) to assist the driver or do not require driver intervention. As automated vehicles, autonomous cars can perceive their environment and navigate routes autonomously, completing the journey without human intervention.
Fully autonomous vehicles utilize hardware devices such as radar (LiDAR, millimeter-wave radar) and onboard cameras, along with software systems, to coordinate and operate, enabling them to complete travel operations safely and reliably, transporting passengers to their destinations without driver intervention.
Two modes
There are two main technological development models for autonomous vehicles: single-vehicle intelligence and intelligent connectivity.
Vehicle-on-a-vehicle intelligence refers to equipping autonomous vehicles with sufficient hardware (such as LiDAR, millimeter-wave radar, and onboard cameras) to provide them with more comprehensive traffic information and enable them to autonomously identify and process that information. Vehicle-on-a-vehicle intelligence focuses more on individual intelligence, placing higher demands on the intelligence of each individual autonomous vehicle.
Intelligent connectivity integrates autonomous vehicles into the transportation network, enabling data such as travel routes, planning, and driving status to be transmitted to the cloud and shared with every participant in the traffic flow, including vehicles, pedestrians, and roads. The development of intelligent connectivity allows autonomous vehicles to obtain more traffic information and make advance adjustments to their travel status, thereby ensuring travel safety. Intelligent connectivity leans towards overall intelligence, encompassing intelligence among all traffic participants, including roads, vehicles, and pedestrians. While intelligent connectivity is currently the widely accepted solution for autonomous driving, it still faces many challenges due to imperfect laws and regulations, unclear plans for traffic infrastructure upgrades, and relatively low public acceptance (related reading: An analysis of the difficulties in developing vehicle-road cooperative systems).
Three elements
The realization of autonomous driving is inseparable from three major technical elements: perception, planning, and control.
For autonomous vehicles to achieve autonomous driving, they need to be able to "see" the road clearly, just like humans do. Perception enables autonomous vehicles to understand and grasp the traffic environment. With the support of the perception system, autonomous vehicles can acquire information such as the position, speed, and possible subsequent behavior of obstacles (vehicles, pedestrians) in the traffic environment; driving areas and traffic rules (lane line detection, traffic light recognition, traffic sign recognition); and they can also use the perception system to understand their own location (positioning), thus providing important road information for further decision-making and planning.
The first step in planning for autonomous vehicles is to ensure they can clearly "see" the information. However, analyzing and making decisions based on this information, and then planning subsequent travel behavior, is an even more crucial step. Just as a person walking on the road needs to plan their path, an autonomous vehicle also needs to plan based on the road information it receives. Depending on the direction of the planning, this can be divided into behavior planning, task planning, and action planning. Autonomous vehicles analyze traffic conditions (vehicles, pedestrians, etc.) based on their travel task and make corresponding judgments, such as overtaking, stopping, or detouring. The planning system is like the human brain, analyzing and judging the road information it receives and adjusting driving behavior according to the travel task. This planning stage is analogous to the human driver's handling of the traffic environment. Planning is vital for autonomous vehicles. For autonomous vehicles to drive safely and handle various traffic environments, they need to be able to react promptly to different scenarios and provide the best solution when faced with decisions such as "passenger priority" or "pedestrian priority."
The control system is the most direct manifestation of the implementation of autonomous vehicles. As the bottom layer of the entire autonomous vehicle system, it plays the role of the "hands" and "feet" of the human driver. The control system reacts to the plans made by the autonomous vehicle, enabling the autonomous vehicle to successfully complete a series of actions such as acceleration, deceleration, and avoidance. The core technologies of autonomous driving control execution mainly include the longitudinal control and lateral control technologies of the vehicle.
Four directions
The development path of autonomous driving has four clear directions, namely the "new four modernizations" of automobiles (electrification, intelligence, connectivity, and sharing).
The development of electric vehicles provides a good platform for the realization of autonomous vehicles. Compared to gasoline-powered cars, which rely on engines for power and are susceptible to uncontrollable factors such as combustion defects during operation, electric vehicles offer greater controllability. Since electric vehicles primarily rely on generators and electric motors to operate, the speed of these motors can be effectively controlled by varying the output voltage, thus enabling efficient vehicle control. Autonomous vehicles need to complete their journeys autonomously; therefore, effective control is essential for improving their safety.
The development of intelligent and electric vehicles still faces challenges such as insufficient driving range and the significant impact of temperature on range. Once these shortcomings are resolved, and electric vehicles can achieve the same performance as gasoline-powered vehicles, automotive intelligence technology will further advance. Intelligence primarily refers to vehicle-level intelligence, a manifestation of advanced driver assistance systems (ADAS). Although many cars now feature ADAS, they only achieve partial autonomous driving at levels L2 and L3, and there is still a long way to go before reaching L5 autonomous driving. Improvements in vehicle-level intelligence technology provide the technological reserves and possibilities for achieving autonomous driving. Gradually improving vehicle-level intelligence technology and realizing autonomous driving under this level is a necessary transition to L5 autonomous driving and an indispensable solution. Therefore, automotive intelligence is a necessary stage in the realization of autonomous driving.
The development of connectivity, or vehicle-to-everything (V2X) technology, addresses the drawbacks of single-vehicle intelligence, such as high costs and excessive sensor hardware. Connectivity, on the other hand, uses V2X to transmit traffic perception tasks from single-vehicle intelligence to other vehicles via V2X. Through V2X (vehicle-to-vehicle, vehicle-to-pedestrian, and vehicle-to-road) technology, dynamic road information is transmitted to vehicles that have already achieved single-vehicle intelligence. This allows vehicles to be informed in advance and respond accordingly while driving on the road, making autonomous driving safer and more efficient. It also promotes traffic management and improves driving efficiency. The era of autonomous driving is inseparable from connectivity.
With the advent of the era of autonomous driving, shared mobility will become an effective way to optimize transportation development and improve transportation planning. Through sharing, autonomous vehicles can achieve maximum utilization efficiency and save more parking space development.
Five modules
The development of autonomous driving is inseparable from the technological advancement of domain controllers. Domain controllers can integrate previously isolated ECUs, controlling all ECUs and sensors in the vehicle through one or more central controllers. This allows for the integration of many ECUs with similar but separate functions onto a more powerful processor hardware platform. The development of domain controllers will gradually reduce the R&D costs of automotive intelligence upgrades and accelerate the implementation of autonomous driving technology. Domain controllers are mainly divided into a centralized electronic and electrical architecture comprising five domains: powertrain domain, body domain, chassis domain, cockpit domain, and autonomous driving domain.
The power domain, also known as the safety domain, is an intelligent powertrain management unit primarily used for powertrain optimization and control. In electric vehicles, it mainly refers to the integration of electric drive and electronic control systems, while also possessing functions such as intelligent electrical fault diagnosis, intelligent energy saving, and bus communication.
The body domain primarily transitions from a decentralized combination of functions to the integration of all electronic control functions of the vehicle body. The control of all body equipment, such as brake lights and tailgate locks, is integrated together, which can effectively reduce the weight of the entire vehicle and reduce the control costs of each component.
The chassis domain is primarily related to vehicle operation, integrating the driving, transmission, steering, and braking systems. It is mainly responsible for the connection of various drive components, power transmission, steering, and braking functions. As the primary control domain for a vehicle as a means of transportation, the development of the chassis domain is a key factor in the feasibility of autonomous vehicles. Similar to the powertrain domain, most control systems within the chassis domain have high safety requirements, needing to meet ASIL-D safety standards (the highest safety level in the ASIL series). Therefore, the chassis domain also has high industry barriers, and currently, most chassis domain controllers are still in the laboratory stage.
The cockpit domain is primarily responsible for the functions of the automotive cockpit electronic system. It integrates functions such as smart cockpits, instrument clusters, central control screens, and head-up displays. Unlike traditional cockpits, which consist of several distributed subsystems or individual modules, the cockpit domain needs to have excellent processing power to realize complex smart cockpit functions such as multi-screen linkage and multi-screen driving.
The autonomous driving domain requires capabilities such as multi-sensor fusion, localization, path planning, wireless communication, decision control, and high-speed communication. It needs external hardware devices such as millimeter-wave radar, lidar, onboard cameras, and inertial navigation to achieve the perception and decision-making capabilities of autonomous driving. Its core is the processing power of the chip, and the ultimate goal is to meet the computing power requirements of autonomous driving, simplify equipment, and improve the integration of the autonomous driving system. The autonomous driving domain is also responsible for ensuring the security of the vehicle's underlying core data and network data in autonomous driving mode, and is a core component driving L3 and higher levels of autonomous driving.
Six levels
The Society of Automotive Engineers (SAE) classifies autonomous driving into six levels, from L0 to L5, based on the degree to which advanced driver assistance systems (ADAS) participate in the driving process.
On August 20, 2021, the recommended national standard GB/T 40429-2021 "Classification of Driving Automation for Automobiles," proposed by the Ministry of Industry and Information Technology and under the jurisdiction of the National Automotive Standardization Technical Committee, was approved and issued by the State Administration for Market Regulation and the Standardization Administration of China (National Standard Announcement No. 11 of 2021), and came into effect on March 1, 2022. This classification divides autonomous driving into six levels: Levels 0-2 are driving assistance, where the system assists humans in performing dynamic driving tasks, but the driver remains the primary driver; Levels 3-5 are autonomous driving, where the system replaces humans in performing dynamic driving tasks under designed operating conditions, and when the function is activated, the system becomes the primary driver.
