(Image from the Internet)
Autonomous vehicles are autonomous vehicles that integrate functions such as perception, decision-making and control, among which the perception system replaces the visual, hearing, touch and other functions of human drivers, and integrates massive traffic environment data collected by cameras, radar and other sensors to accurately identify various traffic elements and provide support for the decision-making system of autonomous vehicles.
camera
The vision sensors used in self-driving cars are mainly industrial cameras, which have high image resolution, transmission ability and anti-interference ability, and can be divided into three categories: monocular, binocular and tri-eye cameras.
1. Monocular camera
A monocular camera is a camera that continuously outputs images using only one optical system and solid-state imaging devices. The structure and calibration are simple, which can effectively avoid the shortcomings of small field of view and difficult stereo matching in stereo vision, but there are irreconcilable contradictions in the measurement range and measurement distance, that is, the wider the camera angle, the shorter the precision detection distance; the narrower the camera viewing angle, the longer the precision detection distance.
2. Binocular camera
The binocular camera is based on human vision research, does not actively project the light source to the outside world, and only relies on the 2 pictures taken to obtain scene depth information to achieve three-dimensional scene reconstruction. Binocular cameras have relatively low hardware requirements, but are extremely sensitive to ambient brightness and computationally complex.
3. Three-eyed camera
The three-eye camera achieves full coverage of different ranges of scenes through cameras with different focal lengths, that is, the wide field camera completes the close-up perception task, the main field camera completes the medium-distance scene perception task, and the narrow field camera completes the long-range perception task, which not only solves the problem that a single camera cannot zoom frequently, but also solves the problem of recognition clarity at different distances. However, due to the fact that multi-channel image data processing is more difficult than single-channel image data processing, three-eye cameras have higher requirements for chip processing capabilities and hardware reliability.
radar
Radar can actively detect the surrounding environment, less affected by the external environment than visual sensors, and is one of the important sensors for self-driving cars. Radar obtains data such as target distance, orientation, and rate of change of distance by emitting electromagnetic waves to the target and receiving echoes. According to the electromagnetic wave band, radar can be subdivided into three categories: lidar, millimeter wave radar, and ultrasonic radar.
1. Lidar
Lidar is composed of two parts: laser detection and laser ranging, which maintains a keen perception of the outside world through real-time feedback, and has the advantages of high resolution, strong anti-active interference ability, good orientation, long measurement distance, and short measurement time. Lidar can be divided into single-line lidar and multi-line lidar. Single-line lidar rotates through a single scan line to obtain two-dimensional information about the object; multi-line lidar rotates through multiple scan lines to obtain the depth information of the three-dimensional space of the object, can measure the basic characteristics and local details of the object, and has high measurement accuracy and reliability.
2. Millimeter wave radar
Millimeter wave radar refers to radar working in the frequency domain of 30 ~ 300GHz, with the advantages of small size, light weight and high spatial resolution, with excellent characteristics such as all-weather and all-day time, can identify multiple small targets at the same time, can penetrate fog, smoke, dust and other environments, accurately measure the relative distance and relative speed of targets, is widely used in self-driving car workshop distance detection, but is susceptible to interference.
3. Ultrasonic radar
Ultrasonic radar working frequency is above 20KHz, mostly used for accurate ranging, the basic principle is to measure the time difference between ultrasonic transmission pulses and received pulses, combined with the ultrasonic transmission speed in the air to calculate the relative distance. Common ultrasonic radar: installed on the front and rear bumpers of the car, used to measure the obstacles in front of and behind the car; installed on the side of the car, used to measure the distance of the side obstacles.
High-precision positioning
Since autonomous vehicles cannot accurately perceive traffic environment information such as obstacles, drivable areas and traffic sign markings like human drivers, it is necessary for global satellite navigation systems, inertial navigation systems, and high-precision maps to organically combine autonomous vehicles with the surrounding traffic environment to achieve over-the-horizon perception and reduce the computational pressure of on-board perception sensors.
1. Global satellite navigation system
Vehicle location information is a prerequisite for the normal operation of autonomous vehicles and a reference benchmark for autonomous driving systems to ensure safe driving. At present, vehicle positioning uses four major global satellite navigation systems such as GPS, Beidou, GLONASS and GALEO. The global satellite navigation system can give more accurate positioning information, but when the data update frequency is low, the carrier moves at high speed or is blocked, it is easy to lose the positioning signal, resulting in an increase in error or even inability to locate.
2. Inertial navigation system
Inertial navigation system is an autonomous navigation system that is not easily disturbed by the external environment, by measuring the acceleration and angular rate of autonomous vehicles, the speed, position, attitude and heading of autonomous vehicles can be obtained through analysis and processing. Inertial navigation systems can provide continuous positioning outputs when the satellite navigation system signal is interrupted, but with an error accumulation effect, the positioning accuracy will continue to decrease with the progress of the positioning process.
3. HD map
High-precision maps formed by fusing sensor data such as lidar, inertial navigation systems, and wheel rangefinders can provide lane-level navigation services for autonomous vehicles, providing information including traffic sign markings, protective facilities, road curvature, heading, slope, and cross-slope angle. Combined with the sensing data of the autonomous vehicle itself, macro road matching, microscopic precise positioning and global environment perception can be effectively realized, and the over-visual distance assistance can be provided for the safe operation of autonomous vehicles.
Internet of Vehicles
Through the all-round network connection of vehicles, vehicles and people, vehicles and vehicles, vehicles and roads, and vehicles and service platforms, the Internet of Vehicles improves the level of vehicle intelligence and automatic driving capabilities, builds a new format of automobile and transportation services, and provides users with intelligent, safe and efficient integrated services. At present, there are two technical solutions for the Internet of Vehicles, including the IEEE-led DSRC technology solution and the 3GPP-led LTE-V technology solution.
1. DSRC technology
DSRC (DedicatedShort Range Communications) is an efficient wireless communication technology, consisting of an on-board unit (OBU), a roadside unit (RSU, Road Side Unit) and a control center, which can realize the identification and real-time data transmission of high-speed moving vehicles within a specific range. The advantage of DSRC is that it is mature in technology, can ensure low latency and safety and reliability, and can meet the requirements of autonomous vehicles for the stability and real-time of networked communication systems, but there are also shortcomings such as small coverage, low transmission rate, easy to be obscured by buildings, slow processing data, and high construction costs.
2. LTE-V technology
LTE-V is based on the existing cellular mobile communication support (3G/4G), and is divided into centralized (LTE-V-Cell) and distributed (LTE-V-Direct) according to the communication mode. LTE-V-Cell takes the base station as the distribution center, requires the support of the existing cellular network, has the characteristics of large bandwidth, wide coverage and other communication characteristics, and can achieve long-distance communication; LTE-V-Direct is independent of the cellular network, supporting direct communication between the vehicle and the surrounding environment nodes (including other vehicles), with low latency, high reliability advantages.
LTE-V started late and is still in the R&D and testing stage, but has formed a complete network system that can be operated, which can provide reliable communication capabilities in high-frequency band (5.9GHz), high speed (250km/h), large traffic flow and other environments, and is more mature in terms of large capacity, low latency, anti-interference and manageability. Low-cost deployment, reusable infrastructure from existing cellular networks, and wide coverage that scales beyond hundreds of meters of non-line-of-sight range.
Reprinted from the forefront of intelligent driving, the views in the text are only for sharing and exchange, do not represent the position of this number, such as copyright and other issues, please inform, we will deal with it in a timely manner.