Sensors have been widely used in Advanced Driving Assistance System (ADAS) and Automatic Driving System (AD), which are mainly used to ensure the safety of passengers and drivers. Nowadays, the combination of camera, radar and lidar is used to scan the vehicle environment and process data in real time. This scheme can respond faster and more accurately than drivers.
At present, sensors can be used for short-range, medium-range or long-range applications of automobiles. Vehicle-mounted radars can be specially used for short-range and long-range applications by being installed in different positions. Vehicle-mounted lidars are mainly used for long-range applications. In the future, short-range lidars will be used for automatic lane change of expressways, intercity driving and even urban road driving. Those applications need to monitor the surrounding environment of the car in real time to avoid any blind spots in the car.
Figure1: applications of sensors
Camera
According to Yole's data, each car produced in 2021 is equipped with an average of 2.6 cameras, and it is estimated that by 2027, the average number of cameras per car will rise to 4.6. The advanced prototypes on display today usually have 11 to 12 cameras, and according to the roadmap of self-driving cars sold to ordinary consumers, each of these models has more than 20 cameras on average.
Due to its high cost and low resolution, thermal imager is only used in a few car models, and its main use is to detect pedestrians or animals on the road. This camera can be massive put into use while the price is low Until now, the price of thermal imager is still very expensive, but it has been deployed in self-driving cars such as Waymo, Cruise and Zoox. It may take several years to become the mainstream device in ADAS.
The development trend of visible light imager is the improvement of resolution. A few years ago, the resolution of the forward-looking ADAS camera was only 1.3 million pixels, but now its resolution has been improved to 8 million pixels, and it is expected to be further improved in the future. Of course, this is not the only trend-the frame rate of the camera is also increasing, and now the frame rate has reached 60 frames per second, and higher frame rates are still being explored.
Fgure2: Camera
At the same time, the dynamic range of the camera is also very important, so that it can operate under low light conditions, and ensure that the car can quickly transition from low light to strong light when leaving the tunnel. In addition, the traffic signal lights are made of LED, and the driving mode of LED will induce the flashing of light, which may be misjudged by the running camera. To address this, the influence of light flicker can be offset by raising the parameters of the camera.
In addition, visible light cameras do not work well in low light or bad weather, while far infrared (FIR) or shortwave infrared (SWIR) cameras can solve this problem. However, the price of these cameras is still too high, and the performance needs to be further improved before they can be deployed in cars. There is no single mode can solve the challenge of automatic driving technology, cameras can improve resolution, speed and dynamic range, but the problem arises in the downstream real-time computing ability. For example, event cameras bring extra performance, but they also require developers to reconsider systems. In any case, the camera must work in conjunction with 4D radar, LiDAR and inertial/GNSS devices. There is no doubt that the progress of ADAS and AV technology will require more cameras and more diversified camera types.
Radar
According to Yole's data, the global vehicle radar market will reach 5.8 billion US dollars in 2021, and it is expected to grow to 12.8 billion US dollars at a compound annual growth rate of 14% by 2027. The main driving factor of radar deployment is related to automobile safety functions, such as AEB (Automatic Emergency Braking) function, which can work together with ADAS front-view camera. For many years, radar has been used for blind spot detection at the rear of automobiles, and OEM manufacturers are increasingly using rake-angle radar. Another driving factor is related to in-cabin sensing. Radar is used to detect the presence of children. Even if children in the car are covered by blankets, CPD (in-car child detection system) can detect them.
At present, the leading enterprises are Continental Group, Bosch, Hella (now Forvia), Denso, Aptiv and Veoneere These six enterprises have about 80% market share, in addition to some new enterprises emerging, such as Arbe, Vayyar, Uhnder and Metawave. They focus on developing imaging radars that can provide better angular resolution, so that they can better sense objects.
In addition to integration, RF (radio frequency) performance has also evolved fundamentally. Initially, the radar used analog beamforming and mechanical steering operations before turning to digital beamforming (full scene illumination). In the era of digital beamforming, MIMO (Multiple Input Multiple Output System) technology is introduced, which not only keeps reasonable physical size, but also increases virtual aperture. At present, all participants in the industry provide MIMO scaling to realize 4D imaging radar with angular resolution lower than one. Not only that, but the overall dimensions and cost are very important. In the next generation radar, the focus will turn to radar signal calculation, which is why the industry began to study machine learning and artificial intelligence algorithms of radar signals.
Lidar
The automotive LiDAR market is expected to reach $2 billion by 2027, up from $38 million in 2021, according to Yole Intelligence. Component types within LiDAR are changing. EEL (Edge Emitting Laser) transits to VCSEL (Vertical Cavity Surface Emitting Laser) at the emitter end, PD (Photodiode)/APD (Avalanche Diode) transits to SPAD (Single Photon Avalanche Diode)/SiPM (Silicon Photomultiplier Tube) at the receiver end, FPGA transits to ASIC at the chip end. These changes are improving the scanning range and resolution, thus improving the data quality of LiDAR. With regard to the type of LiDAR, there is also a transition from "mixed solid LiDAR with mechanically moving parts" to "pure solid LiDAR without mechanically moving parts". In the next 5 to 10 years, it is expected that LiDAR based on optics will appear, and FMCW (Frequency Modulated Continuous Wave) principle will be used for ranging and optical scanning.
Figure3: Advanced Driving Assistance System
At present, in LiDAR market, few enterprises can really deliver products to OEM. VALEO is a leader and has been mass-produced with Audi since 2018. In addition to VALEO, HESAI Technology and ROBOSENSE from China are currently dominant in the market. In addition, other companies including INNOVATION, CONTINENTAL and HUAWEI are also providing LiDAR in small batches, but Luminar and Innoviz are still not mass-produced.
During 2020-2021, LiDAR is the leading category in terms of financing rounds and financing amounts. In 2021, investors invested more than $2.6 billion in LiDAR, but this figure decreased more than 10 times in 2022. Investors are thinking more carefully about their investment targets, hoping to invest in suitable startups whose technology can not only keep low cost, but also achieve mass production, and at the same time meet the LiDAR demand of automobiles.
In deep analysis the car camera and radar market, four or five of them have about 75% market share. In the medium and long term, similar developments are expected in LiDAR field. The integration between LiDAR manufacturers such as Velodyne/Ouster will continue, and Tier-1 will acquire LiDAR.
In terms of sensing sensors, it is necessary to reduce the cost of LiDAR sensors to realize automatic driving function, but the cost is also related to the output. Nowadays, the output of LiDAR sensors is too low, and increasing the output is very important to greatly reduce the cost. In addition, more development is needed to realize automatic driving in bad weather. Recently, Waymo released a video showing that robot taxi can drive automatically without a safe driver in rainy days. Therefore, it is possible for cars to realize automatic driving, which will be related to sensor hardware and related sensing software.
Analysis on the Trends of Three Major Automotive Sensors in 2023-China.exportsemi.com
