With the rapid development of technology, some new terms such as cloud computing, big data, and artificial intelligence have entered the public’s field of vision. Artificial intelligence is another technological revolution after mankind has entered the information age and is receiving more and more attention. As the extension and application of artificial intelligence and other technologies in the automotive industry and transportation, unmanned driving has received close attention from the industry, academia, and even the national level in the world in recent years.
Self-driving cars rely on artificial intelligence, visual computing, radar, monitoring devices, and global positioning systems to work together to allow computers to automatically and safely operate motor vehicles without any human active operation. Autonomous driving technology will become a new development direction for future cars. This article will mainly introduce the application fields of artificial intelligence technology in automatic driving, and make a simple analysis of the development prospects of automatic technology.
Artificial intelligence is a fast-developing science that started late. Since the 20th century, scientists have been seeking ways to give robots wisdom. The concept of modern artificial intelligence was developed from the British scientist Turing’s search for intelligent machines. It was not until 1937 that Turing’s paper “Ideal Automata” gave artificial intelligence a strict mathematical definition, which actually needs to be dealt with in the real world. Many problems cannot be purely numerical calculations, such as speech comprehension and expression, graphic image and sound comprehension, medical diagnosis, etc.
In 1955, Newell and Simon’s Logic Theorist proved 38 of the first 52 theorems in “Principles of Mathematics.” Simon asserts that they have solved the problem of how the material system obtains the nature of the mind (this kind of thesis is called “strong artificial intelligence” in the later philosophical field), and that machines have the ability to think logically like humans. In 1956, “Artificial Intelligence” (AI) was proposed by John McCarthy in the United States. After an early stage of exploration, artificial intelligence developed in a more systematic direction, and has since become an independent subject.
In the 1950s, the research on artificial intelligence started with games and games as the object; in the 1960s, the research focused on solving general problems with search methods; in the 1970s, artificial intelligence scholars conducted fruitful artificial intelligence research; in the 1980s , Began the research of uncertain reasoning, non-monotonic reasoning, and theorem reasoning; in the 1990s, breakthroughs were made in basic research such as knowledge representation, machine learning, and distributed artificial intelligence.
Overview of the application of artificial intelligence in autonomous driving technology
The development of artificial intelligence for sixty years, with several ups and downs, is now ushering in another craze. Breakthroughs in deep learning, computer vision, and natural language understanding have made many applications that were once fantasy possible, such as driverless cars. Is one of them. As the extension and application of artificial intelligence and other technologies in the automotive industry and transportation, unmanned driving has received close attention from the industry, academia and even at the national level in the world in recent years. At present, artificial intelligence has also been widely used in auto-driving technology.
Autonomous vehicles rely on artificial intelligence, visual computing, radar, monitoring devices and global positioning systems to cooperate. It is a comprehensive system that integrates environmental perception, planning and decision-making, and multi-level assisted driving. It focuses on the use of computers, modern Technologies such as sensing, information fusion, communication, artificial intelligence and automatic control are typical high-tech complexes.
This kind of car can “think”, “judge” and “walk” just like humans, allowing the computer to operate the motor vehicle automatically and safely without any human active operation. According to the classification of SAE (Society of Automotive Engineers), it is divided into five levels: driver assistance, partial automatic driving, conditional automatic driving, highly automatic driving, and fully automatic driving.
The first stage: The purpose of driver assistance is to provide assistance to the driver, including providing important or useful driving-related information, as well as issuing clear and concise warnings when the situation becomes critical. At this stage, most ADAS active safety assistance systems allow vehicles to realize perception and intervention operations. For example, anti-lock braking system (ABS), electronic stability control (ESC), lane departure warning system, frontal collision warning system, blind spot information system, etc. At this time, the vehicle can learn the surrounding traffic conditions through cameras and radar sensors. Then make warnings and interventions.
The second stage: Some autonomous vehicles obtain road and surrounding traffic information through cameras, radar sensors, laser sensors and other equipment. The vehicles will provide driving support for multiple operations on the steering wheel and acceleration and deceleration. However, when the driver receives a warning, when the corresponding action is not taken in time, it can automatically intervene, and other operations are left to the driver to realize man-machine co-driving, but the vehicle does not allow the driver’s hands to leave the steering wheel. For example, adaptive cruise control (ACC), lane keeping assist system (LKA), automatic emergency braking (AEB) system, lane departure warning (LDW), etc.
The third stage: conditional autonomous driving: The driving operation is completed by the automatic driving system. According to the limitation of road conditions, system requests are issued when necessary and must be driven by the driver.
The fourth stage: Highly automatic driving: All driving operations are completed by the automatic driving system. According to the request of the system, the driver may not take over the vehicle. The vehicle can already complete automatic driving. Once there is a situation where the automatic driving system is unable to parry, the vehicle can also adjust itself to complete the automatic driving, and the driver does not need to interfere.
The fifth stage: Fully automatic driving: The ideal form of automatic driving, passengers only need to provide a destination, no matter any road conditions, any weather, the vehicle can achieve automatic driving. This level of automation allows passengers to engage in computer work, rest and sleep, and other entertainment activities without the need to monitor the vehicle at any time.
The realization of autonomous driving
To realize autonomous driving, a vehicle must go through three major links: firstly, perception. That is to let the vehicle acquire, different systems need to be collected by different types of vehicle sensors, including millimeter wave radar, ultrasonic radar, infrared radar, laser radar, CCD CMOS image sensor and wheel speed sensor, etc. The working status of the vehicle and its parameter changes. Secondly, deal with, that is, the brain analyzes and processes the information collected by the sensors, and then outputs control signals to the controlled device. Thirdly, implementation. According to the signal output by the ECU, the car can complete the action execution. Every link is inseparable from the foundation of artificial intelligence technology.
Application of artificial intelligence in autonomous driving positioning technology
Positioning technology is the basis of autonomous vehicles. Currently commonly used technologies include line navigation, magnetic navigation, wireless navigation, visual navigation, navigation, laser navigation and so on.
Among them, magnetic navigation is currently the most mature and reliable solution, and most of the existing applications use this kind of navigation technology. Magnetic navigation technology provides vehicle lane boundary information by burying magnetic signs on the lane. Magnetic materials have good environmental adaptability. It can adapt to rainy weather, snow and ice coverage, insufficient light or even no light. The disadvantages are needed to make major changes to the existing road facilities, and the cost is higher. At the same time, magnetic navigation technology cannot predict the obstacles in front of the lane, so it is impossible to use it alone.
Visual navigation has low requirements on infrastructure and is considered to be the most promising navigation method. Visual methods have received greater attention in highways and urban environments.
Application of Artificial Intelligence in Autonomous Driving Image Recognition and Perception
The perception of driverless cars relies on sensors. At present, the performance of sensors is getting higher and higher, the size is getting smaller and smaller, and the power consumption is getting lower and lower. Its rapid development is an important driver of the unmanned driving boom. Conversely, unmanned driving puts forward higher requirements for on-board sensors, and promotes its development. Sensors used for unmanned driving can be divided into four categories:
Radar sensor： It is mainly used to detect the position, distance and moving speed of obstacles (such as vehicles, pedestrians, road shoulders, etc.) within a certain range. Commonly used vehicle-mounted radar types include lidar, millimeter-wave radar and ultrasonic radar. Lidar has high accuracy and wide detection range, but the cost is high. For example, the cost of 64-line lidar on the roof of Google’s unmanned car is as high as more than 700,000 yuan; millimeter wave radar has relatively low cost and long detection range, which is widely used by car companies. It is used, but the accuracy is slightly lower than that of the lidar, and the viewing angle is small; the ultrasonic radar has the lowest cost, but the detection range is short and the accuracy is low, and it can be used for collision warning at low speeds.
Vision sensor：Mainly used to identify lane lines, stop lines, traffic lights, traffic signs, pedestrians, vehicles, etc. Commonly used are monocular cameras, binocular cameras, and infrared cameras. The cost of visual sensors is low, and there are many related researches and products. However, visual algorithms are susceptible to light, shadow, contamination, and occlusion, and their accuracy and robustness need to be improved. Therefore, image recognition, which is one of the fields in which artificial intelligence technology is widely used, is also a research hotspot in the field of driverless cars.
Positioning and pose sensors： Mainly used for real-time high-precision positioning and attitude perception, such as obtaining latitude and longitude coordinates, speed, acceleration, heading angle, etc., generally including global satellite positioning system (GNSS), inertial equipment, wheel speedometer, odometer, etc. Nowadays, the commonly used high-precision positioning method in China is to use differential positioning equipment, such as RTK-GPS, but an additional fixed differential base station is required, the application distance is limited, and it is easily affected by buildings and trees. In recent years, surveying and mapping departments in many provinces and cities have set up continuous operation reference station systems (CORS) equivalent to fixed differential base stations, such as Liaoning, Hubei, Shanghai, etc., to achieve large-scale coverage of positioning signals. This infrastructure construction is intelligent Driving provides strong technical support. Positioning technology is the core technology of unmanned driving, because with the location information, you can use the rich prior knowledge of geography, maps, etc., and you can use location-based services.
Body sensor： From the vehicle itself, obtain the vehicle’s own information such as vehicle speed, wheel speed, gear position, etc. through the entire vehicle network interface.
Application of artificial intelligence in deep learning of autonomous driving
The driver’s cognition relies on the brain, and the “brain” of the driverless car is the computer. The computer in an unmanned vehicle is slightly different from our commonly used desktops and notebooks, because the vehicle will encounter bumps, vibrations, dust, and even high temperatures when driving. Generally, computers cannot run in these environments for a long time. Therefore, unmanned vehicles generally use computers in industrial environments-industrial computers.
An operating system is running on the industrial computer, and driverless software is running in the operating system. Figure 1 shows the software system architecture of an unmanned vehicle. Above the operating system is the supporting module (here, the module refers to the computer program), which provides basic services to the upper software module.
Support modules include: virtual switch module, used for communication between modules; log management module, used for log recording, retrieval and playback; process monitoring module, responsible for monitoring the operating status of the entire system, and prompting the operator if a module is not operating normally And automatically take corresponding measures; interactive debugging module, responsible for the interaction between developers and unmanned driving systems.
In addition to cognition of the outside world, the machine must also be able to learn. Deep learning is the foundation for the success of unmanned driving technology, and deep learning is an efficient machine learning method derived from artificial neural networks. Deep learning can improve the time efficiency of cars in recognizing roads, pedestrians, obstacles, etc., and guarantee the correct rate of recognition. After training with a large amount of data, the car can convert the collected graphics, electromagnetic waves and other information into usable data, and use deep learning algorithms to realize unmanned driving.
When the driverless car collects data through radar, etc., the original training data must first be preprocessed. Calculate the mean and standardize the mean of the data, do principal component analysis on the original data, and use PCA whitening or ZCA whitening. For example: the time data collected by the laser sensor is converted into the distance between the car and the object; the photo information taken by the on-board camera is converted into the judgment of roadblocks, the judgment of traffic lights, the judgment of pedestrians, etc.; the detection of radar The data is converted to the distance between each object.
Applying deep learning to driverless cars mainly includes the following steps:
Application of Artificial Intelligence in Autonomous Driving Information Sharing
Firstly, use wireless networks to share information between cars. Through a dedicated channel, a car can share its location and road conditions with other cars in the team in real time, so that the automatic driving systems of other vehicles can make corresponding adjustments after receiving the information.
Secondly, it is 3D road condition sensing. The vehicle will combine ultrasonic sensors, cameras, radar and laser ranging technologies to detect the topography and landforms within about 5 meters in front of the car, and judge whether the road ahead is asphalt road or gravel, grass, beach and other roads. According to the terrain, it automatically change car settings.
In addition, the car will also be able to automatically shift. Once it detects a change in the terrain, it can automatically decelerate and return to the original state after the road returns to normal.
The amount of traffic information collected by car information sharing will be huge. If these data are not processed and used effectively, they will be quickly obliterated by the information. Therefore, it is necessary to use data mining, artificial intelligence and other methods to extract effective information while filtering out useless information. Considering that the information that needs to be relied upon during the driving of a vehicle has great temporal and spatial relevance, the processing of some information needs to be very timely.
The advantages of artificial intelligence in autonomous driving technology
Artificial intelligence algorithms are more focused on learning functions, and other algorithms are more focused on computing functions. Learning is an important manifestation of intelligence, and the learning function is an important feature of artificial intelligence. At this stage, most artificial intelligence technologies are still in the learning stage. As mentioned earlier, unmanned driving is actually human-like driving. It is a smart car that learns from human drivers how to perceive the traffic environment, how to use existing knowledge and driving experience to make decisions and plans, and how to skillfully control the steering wheel, throttle and brake.
From the perspective of perception, cognition, and behavior, the perception part is the most difficult, and artificial intelligence technology is the most used. Sensing technology relies on sensors, such as cameras, and is popular in the industry due to its low cost. An Israeli company called Mobileye has done a very good job in the field of traffic image recognition. It can perform traffic line recognition, traffic signal recognition, pedestrian detection through a camera, and it can even distinguish whether it is a bicycle, a car or a truck ahead.
The successful application of artificial intelligence technology in the field of image recognition is deep learning. In recent years, researchers have trained image samples through convolutional neural networks and other deep learning models, which greatly improved the accuracy of recognition. Mobileye’s current achievements are due to the company’s early research on deep learning as a core technology. In terms of cognition and control, the traditional machine learning technology in the field of artificial intelligence is mainly used to establish a driver model by learning the driving behavior of human drivers and learn how to drive cars in a human way.
Challenges and prospects of unmanned driving technology
In the context of the current worsening traffic conditions, the commercialization prospects of “unmanned” vehicles are still restricted by many factors. There are:
In addition, one of the biggest characteristics of “unmanned driving” cars is that the vehicles are networked and highly informatized, which also poses a great challenge to the safety of computer systems. Once the computer program is disordered or the information network is invaded, how to continue to ensure the safety of the vehicle and other surrounding vehicles is also an urgent problem to be solved in the future. Although there are still many challenges in unmanned driving technology, unmanned driving is difficult to perceive, and the emphasis is on “learning”. Sooner or later, the technical level of unmanned driving will surpass humans, because stability, accuracy, and speed are the inherent advantages of machines, and humans cannot compare with it.
Driving is sometimes not a burden. On the contrary, it is a kind of fun, reflecting the ability of human beings to expand their limits. The author believes that completely unmanned driving may be a little far away, but with the improvement of machine learning algorithms and the application of mining, a more grounded human-machine harmonious co-driving is just around the corner. No matter how difficult there are on the road of autonomous driving, I believe that there will always be a day when it will appear on urban roads, and the development of technology is full of passion and motivation. In the near future, perhaps autonomous driving will become the dominant trend.