Intelligent driving distance estimates
To ensure the safety of autonomous vehicles, which means that autonomous vehicles in motion the process does not hit any person or thing, so the vehicle is in motion need to know the distance between the object and its surrounding the vehicle itself. To achieve this goal, it is necessary to measure in advance the distance between the vehicle and the vehicle in front of an obstacle, so as to ensure the safety of automatically driving a car.
Autopilot system will depend on accurate, reliable and continuous position information of other road traffic participants, these participants include traffic: pedestrians, bicycles and other vehicles. Usually solve this problem is the use of in-vehicle distance measuring sensor. Radar, lidar and a vision-based system can detect an object within the line of sight. These non-cooperative sensors contrast, collaborative strategy is to follow the sensor all the way participants actively estimated relative position information. The method may be used based on collaborative communication compensate clogging truck vehicle distance measuring sensor detection range, viewing angle defects and facilities. Integration of collaborative strategies and uncoordinated strategies to consider from the perspective of positioning accuracy and robustness seems to be able to get the maximum benefit. A target relative positioning and distance measuring sensor technology can be summarized as the following figure, first of all into two categories: non-cooperative and cooperative positioning positioning. Non-cooperative positioning comprising: radar, laser scanners, and TOF vision camera; Collaborative locating comprises: a ranging-based repeater (radio frequency signal directly in accordance with the estimated relative distance) based on the relative positioning of GNSS (GNSS transmitted through communication the measured location-related information, including the location information and the original GNSS GNSS obtained solver, such as pseudoranges).
There are several ways to achieve automatic driving car distance estimation, they must abide by the following principles:
① This method has the accuracy and stability; ② resolved based on different hardware and software solutions; ③ the use of sensors of different works; ④ in doing research, a priori assumptions to be as small as possible.
The main method of estimation autonomous vehicle distance is achieved using a sensor (such as cameras, radar, etc.) around the vehicle detected object distance, thereby realizing distance estimation purposes. The methods are summarized as follows:
1. estimating the distance based on the millimeter wave radar millimeter wave radar is technically very mature, adaptive cruise earliest start field applied. After the launch of the program at Infineon 24GHZ single-chip radar, millimeter-wave radar is applied to each module of the ADAS, on a global scale, millimeter-wave radar shipments reached ten million. In the autonomous vehicle distance estimation are used in the millimeter-wave radar as the sensor, to the surrounding obstacle identification and location work. Today, the millimeter-wave radar share of global markets is a major supplier of first tier foreign monopoly, such as Bosch, Continental, etc., and with the development of the domestic market, such as domestic suppliers of Chinese domain car also in the layout millimeter-wave radar development effort. According to statistics, the domestic market size of 2018 millimeter-wave radar about 70 million, up more than doubled in 2017, while in the year 2020 expected to be approximately 24 billion by 2025 expected to be approximately 32 billion. Millimeter-wave radar, working in the millimeter wave band i.e. radar detection, and its essence is an electromagnetic wave, a wavelength of about 1-10mm, millimeter-wave radar emitted by the millimeter wave, and then receiving an echo, according to the measured time difference between transmission and reception the location and distance of the front obstacle. The method of estimating the distance based on the millimeter wave radar, FMCW modulation method mainly to measure the distance, its schematic diagram is shown below:
The principle is a signal formed by an oscillator varies continuously, for the received signal and the signal emitted form the frequency difference between them, and showed a difference between transmission time and reception time of the millimeter-wave frequency difference linear correlation, only the frequency difference measurement, measurement can be achieved in front of the vehicle and the object distance is estimated.
ADAS widely in millimeter-wave radar applications, mainly for automatic emergency braking (the AEB), forward collision warning (the FCW), active control lane (the ALC), blind spot monitoring (the BSD), lane change assistance (LCA) and the like, having certain anti-interference ability, easy to penetrate the rain and snow, adaptability better, you can work around the clock, technology is relatively mature, low cost and other advantages. But also it has a low resolution insufficient. Millimeter-wave radar has become an indispensable autonomous vehicles estimated distance of the main sensor.
2. Based on estimates from the laser radar
with autopilot evolving the car, laser radar because of its unique 3D modeling environment has become more autonomous vehicles and L3 essential sensor. From the mechanical to the mixed solid, to pure solid-state laser radar, laser radar of the cost with the development of technology in the continuous reduction, is developing towards miniaturization, ASIC integrated direction, will be the core part of the autopilot sensor. Lidar categories as shown below:
At present the majority of autonomous vehicles is a mechanical test vehicle laser radar, but the high cost of mechanical laser radar, and the complex production process, short life, the future of autonomous vehicles is difficult to meet demanding requirements. Mixing the solid-state laser radar lidar belonging to the mechanical transition from the solid-state laser radar pure intermediate product, and the solid-state laser radar mainly the MEMS , and the OPA 3D Flash three , which can be automated debugging, and there are no rotating mechanical components, in cost, practicality has increased dramatically, solid-state laser radar will be the future development trend of laser radar.
Laser radar based distance estimation autonomous vehicles, the principle is based on laser as a carrier, it is a work in optical frequency band radar. Its working principle is emitting a laser beam to the object, and the received echo is compared with the transmitted signal, for properly treated, obtaining information about the object, as the object distance, azimuth and other information.
The laser radar shown in FIG distance is calculated using the time of reentry of the reflected signal laser radar and some distance estimation is performed by the frequency modulated continuous wave (CWFM) method.
Laser radar is used to acquire the depth information, obstacle detection, target recognition, which is a major advantage can be modeled 3D multi-peripheral object to form high-definition images, to facilitate computer processing and recognition, but also possesses good directivity, no electromagnetic interference, access to comprehensive information and accurate detection, etc., but susceptible to environmental impact, fall under adverse weather precision is difficult to identify obstacles and higher costs.
3. The estimated distance camera-based
automatic car driving distance estimation, the camera plays a vital role, is known as the autopilot eyes. Camera most mature technology, the first application in the vehicle, as the main stage ADAS vision sensor, comprising a single camera, binocular cameras.
The camera has a unique visual imaging capabilities, may utilize a plurality of cameras are synthesized on the surrounding environment, but also can recognize road signs, pedestrians, other sensors can be used as redundant equipment, the autonomous vehicles to improve the accuracy of the estimated distance and safety . Mainly collected by the camera lens image, then the image processing by the internal digital signal as the photosensitive member, so as to achieve the object and sensing the surrounding pedestrian. It includes two measurements. ① distance estimation using a known size of the object by known size and focal length of the camera, can calculate the distance between objects.
According to the figure, a known y, f, and then to measure the distance between the preceding vehicle and the width of the camera in accordance with the theorem of similar triangles, so as to achieve the purpose of using the camera to measure the distance of the preceding vehicle. ② The ground plane distance estimate the camera fixed to the vehicle, the use of the ground plane distance estimation.
The height H of the camera is known and the geometry of F, the road surface, it can be used similar triangles theorem of the distance z between the measured object.
Because of its distance from the camera wide angle detection, access to information rich, accurate angle measurement and other advantages it is widely used in the vehicle and the surrounding object estimate, obstacle recognition, lane detection and tracking, driver's condition monitoring. But due to the large amount of computing, the hardware requirements are too high, resulting in poor real-time systems, susceptible to environmental, climate, etc., making it impossible to obtain depth information.
4. Based on the ultrasonic sensor distance estimation
autonomous vehicles distance estimation using a sensor for ranging and target recognition , one of the primary sensor an ultrasonic sensor as a vehicle has been developed. Ultrasonic sensors are mainly used in detecting an obstacle close.
The working principle of ultrasonic sensors is based on sound wave propagation method, follow the same principle of flight time. And which is simple, inexpensive, small size, light weight, low power consumption, can operate under different conditions, environmental adaptability. However, due to the slow speed of sound, resulting in limited FSP rate, only for close work, the maximum distance of 15 meters, reliability increases as the vehicle speed decreases.
5. Based on estimates from the infrared sensor infrared sensor as one-vehicle sensors in autonomous vehicles from estimates also played a role, it is mainly used in terms of infrared imaging, infrared night vision, obstacle detection. The main infrared sensor having a fast data processing, it is possible to more accurately identify the biological advantage, as compared to other low cost sensors. However, it also has directivity and low radial motion discrimination, disadvantages such as short effective distance. 6. distance sensor fusion development is estimated at the various sensors each have their advantages and disadvantages, with the autopilot speed the process, a single sensor is not capable of autonomous driving technology strict requirements for distance estimation, the development of multi-sensor fusion will become the future trends onboard sensors. At this stage the major manufacturers are actively layout, the most reasonable solution to adapt to the trend of the autopilot. Laser radar, millimeter wave radar, a camera, an ultrasonic radar, etc.
Use of sensor fusion, their advantages can be fully integrated, play to their strengths, to achieve optimal results. In recent hotter Tesla example, the L2 level Autopilot2.0 embodiment is mounted on an ultrasonic radar, a millimeter wave radar, cameras, sensors, and is mainly used in highway more congested road, the vehicle speed can be adjusted depending on the traffic situation , to keep traveling in the lane, an automatic lane change, to switch from one highway to another highway. Another example is the well-known A8 AI luxury cars Audi released, is equipped with a variety of sensors, including four-wire laser radar 1, the ultrasonic radar 12, the wide-angle 360 degree video camera 4, forward camera 1, the infrared night vision camera 1 and a long, mid-range millimeter wave radar total of five, showing that combines the advantages of a variety of sensors, greatly improved the car autopilot performance.
In summary, with the rapid development of automated driving a car, ADAS rapid penetration in the production car market, due to the automatic driving distance estimates rely mainly on-board sensors, vehicle sensors so considerable future prospects of the market. Different focus functionality of the sensor is different, but both have advantages insufficient. In the car on autopilot while carrying a variety of sensors, combined with their complementary advantages, will become the future trend of development. The autonomous vehicles were also inseparable from the estimated multi-sensor fusion development. So, the future of autonomous vehicles will be multi-sensor fusion development trend, autonomous vehicles will be more accurate estimate of the distance, the safety performance of vehicles will be further improved.