The Understanding to Autonomous Driving Sensor

Environment perception, planning and decision-making, motion control, and multi-level aided driving are all part of autonomous driving. It focuses on visual computing, general computing, and neural network computing for data fusion, and it is complemented by V2X communication and artificial intelligence to enable automatic control. Environment sensing, behavior decision-making, path planning, and motion control are the four fundamental technologies of autonomous driving.

Environmental perception is the process of gathering basic information about the surrounding environment via sensors, and it is also the foundation of autonomous driving. The level of automated driving accomplished varies depending on the autonomous driving route taken, as do the types of sensors used. Let's look at the fundamentals, as well as the benefits and drawbacks of each sensor.

Ⅰ.  Camera

A Lens, an Image Sensor, an Image Signal Processor, ISP, and a Serializer are the main components of a camera. In general, the Image Sensor processes the basic information of the object collected by the lens before sending it to the ISP for serialized transmission. LVDS-based transmission on coaxial cable or twisted pair, as well as direct transmission via Ethernet, are two further options.

In terms of the arrangement, the effect of the viewing angle on the perception range is the most important factor. When determining the size of the camera's photosensitive element, the longer the focal length, the narrower the corresponding viewing angle. However, the resolution of the equivalent image can be substantially improved—that is, you can see clearly, but you can see less.

As a result, cameras with various focal lengths are employed in practice to accomplish various tasks. Medium- and long-range cameras, in general, are configured above the L2 level. Three front-view cameras will be used in high-end vehicles. To acquire information from a full-view perspective.

The camera's primary function in an autonomous driving system is to focus on the following:

Obstacle detection: measuring speed and distance (vehicle use requires binoculars or above);

Detection of lane lines: lane line extraction

Reading road information includes traffic signals and sign identification.

Construction and positioning of supplemental maps;

Other traffic participants are detected and recognized, such as vehicles, pedestrians, and animals.

Ⅱ. Millimeter-wave radar

Radar is essentially an electromagnetic wave that is reflected after being obstructed by objects on its transmission route. The distance, speed, and angle of the item can be determined by collecting the reflected signal. Millimeter-wave radar may produce electromagnetic wave signals with wavelengths of 1-10mm and frequencies of 30-300GHz. The diffracted size decreases as the wavelength decreases, implying that the size of the observable item is finer and has a higher resolution. A millimeter-wave system operating at 76-81 GHz (corresponding to a wavelength of around 4 mm) may detect movements as small as a few tenths of a millimeter.

The radar frequency band, on the other hand, is strictly regulated by the government. Vehicle millimeter-wave radar application frequency bands in various nations are generally centered in 24G, 60G, 77G, and 79GHz.

In terms of structure, FMCW (Frequency Modulated Continuous Wave) is the most common working mode for vehicle-mounted radar. FMCW radar receives and transmits at the same time, therefore there is no blind spot in its range, and it can measure Doppler frequency shift and static target probability directly.

FMCW radar's working premise is that its transmission frequency changes linearly with time, allowing time information to be carried in the broadcast signal. A voltage-controlled oscillator generates the high-frequency signal, which is amplified by the power divider before being fed to the transmitting antenna, while the other half is coupled to the mixer, mixed with the received echo, and then low-pass filtered to produce the baseband difference frequency signal. It is transmitted to the signal processor for processing after analog-to-digital conversion. This signal can be used to retrieve not only time information but also the Doppler effect's distinctive spots. As a result, both speed and distance can be measured at the same time.

The following are the technical advantages and downsides of millimeter-wave radar:

The millimeter-wave radar cannot offer height information, but it can detect the item's position. It is unknown whether the object is suspended in the air.

To identify moving targets, the radar principle determines to rely mostly on the Doppler effect. Ground echo and other information are easily jumbled with stationary objects, causing misjudgment.

Radar has a hard time detecting two cars in neighboring lanes at close range.

The radar's spatial resolution accuracy is average, and the data is difficult to recognize object types with.

When a person crosses the road from the side of the road in a lateral direction, the radar cannot tell if it is a human or a shrub on the side of the road, and because of the poor lateral speed perception capacity, the front radar cannot feel the side environment at this time.

Because radar is prone to metal reflection interference, it is easy to misread things on the road, such as cans; radar is also easily impacted by road ramp reflections, resulting in inaccurate obstacle data.

Ⅲ. Ultrasonic radar

Ultrasonic radar is made up of ultrasonic transmitters and receivers, as well as control circuits and power supplies. Piezoelectric conversion devices are used in most ultrasonic transmitters. A 40kHz ultrasonic radar, for example, requires a 40kHz changing voltage signal to be delivered to a piezoelectric ceramic chip, which will then extend and contract according to the polarity of the provided high-frequency voltage and send 40kHz ultrasonic waves. The receiver works on the same principles as the transmitter. Piezoelectric ceramics' reversible properties are employed to convert ultrasonic echoes into voltage signals with the same frequency. A comparable amplification circuit is necessary for processing due to the tiny amplitude of the high-frequency voltage.

The time difference between when the ultrasonic wave is sent out by the ultrasonic transmitting device and when the ultrasonic wave is received by the receiver is used to determine distance with ultrasonic radar. 40kHz, 48kHz, and 58kHz are the common working frequencies of probes. In general, the higher the frequency, the greater the sensitivity, but the narrower the horizontal and vertical detection angles, which is why reversing radars employ 40kHz probes. A little amount of silt has little influence on the ultrasonic radar. The detecting range is 0.1-3 meters, with a high level of precision.

The many forms of ultrasonic radar are categorized into the following groups:

1) Front and rear bumper UPA: the detection distance is 15-250cm, and it is mostly used to assess obstacles in the front and rear of the vehicle.

2) APA mounted on the side with a detection range of 30-500cm The detecting range is longer than UPA, the transmission power is higher, and the cost is more. Automatic parking is the most common application.

The following are the pros and downsides of ultrasonic radar:

Advantages: Ultrasonic energy consumption is low, propagation distance in the medium is long, penetrability is high, the distance measurement method is easy, and the cost is low.

Disadvantages: When detecting distance at high speeds, ultrasonic radar has some difficulties. This is due to the fact that weather conditions have a significant impact on ultrasonic transmission speed. The transmission speed of ultrasound varies depending on the weather, as does the propagation speed. Slower, while the automobile is traveling at a high speed, the ultrasonic distance measurement is unable to keep up with the real-time change in the car's distance, resulting in a substantial mistake. The ultrasonic scattering angle, on the other hand, is large, and the directivity is weak. When measuring a long-distance target, the echo signal will be feeble, lowering the measurement accuracy. The ultrasonic distance measuring sensor, on the other hand, offers a significant advantage in short-distance measurement.

Ⅳ. Lidar

Control hardware DSP (digital signal processor), laser drive, laser emitting light-emitting diode, emitting optical lens, receiving optical lens, APD (Avalanche Optical Diode), TIA (variable transconductance amplifier), and detectors are the key components of LiDAR, according to the signal chain of signal processing. They are all electronic components, with the exception of transmitting and receiving optical lenses. The performance of semiconductor technology is gradually improving while the cost is rapidly decreasing, thanks to the rapid evolution of semiconductor technology. Optical components and rotating machinery, on the other hand, account for the majority of the expense of lidar.

Mechanical type, MEMS, phased array, and floodlight area array is the different types of driving forms (FLASH).

The laser determines the propagation distance between the sensor transmitter and the target object (Time of Flight TOF), analyzes the amount of reflected energy on the target object's surface, as well as the amplitude, frequency, and phase of the reflected spectrum, and thus provide accurate three-dimensional structure information about the target object. The working mode of TOF lidar is identical to that of millimeter-wave radar, which is separated into dTOF and iTOF. In most cases, the direct pulse approach is utilized to measure dTOF. Laser transmitters with wavelengths of 905nm and 1550nm are currently the most common, and light with a wavelength of 1550nm is difficult to transfer in the liquid of human eyes. As a result, under the concept of assuring the safety, 1550nm can considerably enhance emission power. The higher the power, the longer the detecting distance, and the longer the wavelength, the better the anti-interference capacity.

Mechanical type, MEMES, and phased array OPA are the three types of lidar structure. The 64-line radar launched by Velodyne in 2007 is an example of the mechanical type. It revolves at 20 rpm and stacks 64 lasers vertically. The basic idea is to rotate the laser point into a line, convert the line into a surface using 64-line stacking, and then get point cloud data to gain 3D environment information.

The point cloud measurement necessitates accurate installation positioning, and the mechanical structure necessitates a complicated mechanical structure. Given the effects of the environment and aging, the average failure duration is just 1000-3000 hours, making it impossible to meet the automobile factory's minimum requirement of 13000 hours. Furthermore, because LiDAR is mounted on the car's roof, civil engineers must address external maintenance difficulties such as the influence of car washing. As a result, mechanical construction severely restricts the cost and promotion of the application.

The MEMS type drives the revolving mirror with microelectromechanical system technology, and the reflected laser beam points in various directions. Fast data acquisition speed, high resolution, and strong adaptability to temperature and vibration are all advantages of solid-state lidar; through-beam control, detection points (point clouds) can be arbitrarily distributed, for example, in the main scanning of highways far ahead, strengthen the side scan at the intersection for the sparse but not completely ignored side scan. This delicate operation cannot be performed by a mechanical lidar that can only rotate at a steady speed.

Valeo SCALA lidar is an example of a typical application. It's now found in the Audi A8 (the first L3 self-driving vehicle). It's mounted in the front bumper and uses MEMS technology to provide a 145-degree scanning angle and an 80-meter detecting range.

The optical phased array technology's principle To modify the emission angle of the outgoing light, lidar uses optically controlled phase technology. The light interference principle is primarily used. By varying the phase difference of the incident light in separate slits, the position of the central pattern (main lobe) after grating diffraction can be altered. The following are its key benefits and drawbacks:

Advantage:

①Simple construction and compact size: Because no spinning parts are required, the radar's structure and size can be considerably reduced, extending its service life and lowering its cost.

②Simple calibration: Due to their fixed optical structure, mechanical lidars frequently require precise positioning and angle adjustments to adapt to different vehicles. Solid-state lidars may be calibrated using software, which substantially simplifies the process.

③Fast scanning speed: The optical phased array's scanning speed is determined by the electronic properties of the material utilized, rather than the speed and precision of mechanical rotation, and may generally reach the MHz level.

④High scanning accuracy: The optical phased array's scanning accuracy is determined by the control electrical signal's accuracy, which can be on the order of a thousandth of a measurement.

⑤Excellent controllability: The optical phased array's beam direction is totally controlled by electrical signals, and it may be oriented arbitrarily within the permitted angle range, allowing for high-density scanning in crucial locations.

⑥Multi-target monitoring: A phased array can be broken down into numerous tiny modules, each of which can be controlled independently to lock and monitor multiple targets at once.

Disadvantages:

①The scanning angle is limited: changing the phase can only modify the centre pattern by around 60°, and 360° acquisition normally takes 6.

②Side lobe issue: in addition to the primary bright lines, grating diffraction will produce secondary brilliant lines. Because of this issue, the laser will produce side lobes outside of the maximum power direction to spread the laser's energy.

③High processing difficulty: The size of the array element in an optical phased array must not exceed half a wavelength. Because the current operating wavelength of lidar is around 1 micron, the array element size must be less than 500nm. Furthermore, the higher the array density, the more concentrated the energy, raising the processing accuracy requirements and necessitating specific technological breakthroughs.

④Large receiving area and low signal-to-noise ratio: Traditional mechanical radars only require a tiny receiving window, whereas solid-state lidars require the entire receiving surface, which introduces more ambient light noise and makes scanning analysis more difficult.