This report investigates the market for radar technology, specifically focusing on automotive applications. It develops a comprehensive technology roadmap, examining the technology at the levels of materials, semiconductor technologies, packaging techniques, antenna array, and signal processing. It demonstrates how radar technology can evolve towards becoming a 4D imaging radar capable of providing a dense 4D point cloud that can enable object detection, classification, and tracking.
The report examines the latest product innovations. It identifies and reviews promising start-ups worldwide. The report builds a short- and long-term forecast model covering the period between 2019 to 2040. The market- in unit numbers and value- is segmented by the level of autonomy and by passenger vehicles, shared vehicles, and trucks. In the first decade, ADAS (level 1 and 2) will be the primary market drivers whilst in the second decade autonomous vehicles will be.
Radars are a key element of the sensor suite in ADAS and autonomous mobility. This report first examines the role that radars play in various ADAS functions such as ACC, AEB, FCA, BSD, LCW, HWA, and so on. It then examines how the radar content per vehicle- both for short/medium and long-range radars- will increase with increasing ADAS and autonomy level.
The report then examines the drivers and trends in operational frequency worldwide. It examines how device parameters- including centre frequency, bandwidth, measurement time, and virtual aperture- affect key performance indicators (KPIs) such as velocity, range, azimuth, and elevation resolution. The common products on the market today are then reviewed and benchmarked. The value chain- from chip (fabless/IDM/foundry) to module makers is outlined.
Detailed market forecast models are built. These market forecasts first consider how ADAS and autonomy will penetrate into the vehicle market. Here, the report will build a twenty-year market forecast (2020 to 2040), segmenting the vehicle market by level 0 to 5 of autonomy. The forecast model also considers the impact of robotaxis and shared autonomous vehicles on total vehicle sales, predicting peak car sales around 2031/2 in a moderate scenario. These forecasts are built in conjunction with our autonomous mobility team. The forecasts are then converted into radar unit sales. To arrive at market value, we develop a moderate and an aggressive price reduction scenario for short/medium and long-range radars. We also develop forecasts by the semiconductor technology (GaAs, SiGe, and Si).
Unit number market forecasts segmented by the ADAS level and autonomous mobility level. The report also provides forecasts for vehicle and truck numbers segmented by autonomy level as well as radar forecasts in value based on different cost evolution scenarios.
Technology trends: semiconductors, on-chip integration, packaging, and low-loss materials
The radar technology is changing. Indeed, these are very exciting times for radars. We offer a detailed quantitative benchmarking of various semiconductor technology options such as GaAs HEMT, InP HEMT, SiGe BiCMOS, Si CMOS, and Si SOI. We consider maximum frequency, amplifier efficiency, lithographic technology node, function integration capability, volume, and cost.
The report shows how the semiconductor technology has evolved and is likely to evolve in the coming years. It shows how and when GaAs technology gave way to SiGe and how now SiGe might be beginning to give away to Si CMOS (or SOI). It offers a detailed overview of key existing and emerging products on the market covering SiGe BiCMOS as well as Si CMOS and SOI. Here, we consider companies such as NXP, Infineon, ST Microelectronics, ON Semiconductor, Texas Instruments, Analog Devices, Arbe Robotics, Uhnder, Steradian, Oculii, and so on.
The shift towards Si CMOS and similar will enable more function integration into radar chips. Indeed, we show how radars have evolved from having a separate chip for each function to single-chip radars. The latest SiGe BiCMOS and some recent Si CMOS radar chips include multiple transceivers, monitoring functions, waveform generators, and an ADC. The latest Si CMOS generations even include a microcontroller with memory as well as a digital signal processing unit (DSP). This clearly shows the trend towards single-chip solutions which will result in significant cost-reduction and volume-production potential.
Packaging solutions are next considered. In the past, multiple dies were directly mounted on the board (CoB) and wire-bond connected. Today, the chips are packaged using wafer-level-packaging technologies, e.g., WLP-BGA, or flip-chip ball-grid-array (BGA). We provide a benchmarking of discreet die vs packaged solutions. Within packaged solutions, we also compare the high-frequency behaviour of flipchip, fan-out and BGA.
Board-level trends in design, material, and passives are then examined. Here, we see how the board arrangement has evolved. In the past two separate RF and digital boards were utilized. Now, a hybrid board in which the top layer is composed of a special RF material is common. The trend- at least for small antenna array sizes- is to go towards antenna-in-package (AiP) designs. Some such designs are already qualified for automotive use. The long-term possibility of antenna-in-chip in the longer term is explored.
The material requirements for low insertion loss at high frequencies are analysed. These special materials will need to offer low loss tangents. Crucially, the dielectric constant and the loss tangent will need to remain stable against variations in temperature and frequency. Furthermore, moisture uptake will need to be low and the material will need to be easy- or with well-known modification- processible, e.g., how to make the Cu stick. This study offers a comprehensive benchmarking of a wide range of materials on the market including ceramic-filled PTFEs, LCP, PI/fluoropolymers, ceramics such as LTCC or AlN, glass, etc.
Towards 4D imaging radars
Radar technology is further evolving towards a 4D imaging radar capable of providing a dense 4D point cloud that can enable moving beyond presence, range, and speed determination towards 3D object detection, classification, and tracking.
We highlight and assess the critical impact of increasing the antenna array on azimuth and elevation resolution and on the data matrix and the point cloud. The additional high-res information on azimuth and elevation paves the way towards 4D imaging radars. These emerging capabilities will blur the lines with lidar, potentially allowing radar to encroach into lidar territory without compromising its light-level and weather independence. This will set up an interesting competitive dynamic although lidar will likely retain its dominance some parameters including angular resolution and potentially object classification.
The report offers a high-level review of deep neutral network and deep learning techniques which have been so successful in camera images. The challenges specific to radar data are considered. In particular, we consider how future radars can densify the radar point cloud, bringing its density closer to the point clouds of lidars. We consider the state-of-the-art in 2D and 3D object detection, and outline some approaches aimed at bridging the performance gap between the two. We discuss the challenge of the limited availability of labelled training data, and how some are seeking to build precise radar maps and to develop semi-automated methods of labelling radar data, often using some late-state fusion with data from cameras, GPS and lidars.
The interference challenges are also briefly discussed. This is expected to become a growing challenge as the number of radar-equipped radars on the roads will grow. Various approaches are under consideration. In some cases, the jammed signals are locally reconstructed. In other approaches, a loose or a tight system-level coordination is proposed, akin to what is found in telecommunication systems.
These radar charts compare the status of today's radar with that which is emerging.
Multiple innovative radar start-ups have emerged in recent years. These firms take different approaches. Some are developing radars on advanced SOI or CMOS nodes, supporting very large virtual channel sizes. This, coupled with their developing processing techniques, can enable truly 4D imaging. Others are developing novel techniques such as the use of metamaterials to electronically steer the radar beam.
Not all are automotive focused. Indeed, some are focused on the UWB band, seeking to offer single-chip low-cost high-resolution radar solutions for applications such as drone navigation, vital signs monitoring, human machine interface, medical imaging, smart home, and so on. These start-ups include Arbe Robotics, Uhnder, Steradian, Echodyne, Metawave, Oculii, Vayyar, Lunewave, Zendar, Ghostwave, Novelda, Omniradar (Staal Technologies), and so on.
Detailed market forecast models are built. We consider the diffusion of different levels of ADAS and autonomy into the vehicle market over a twenty-year period. We have selected this long timeframe because higher levels of autonomy will take time to become technologically ready and commercially viable.
Our model, therefore, offer a twenty-year unit number forecast (2020 to 2040), segmenting the vehicle market by level 0 to 5 of autonomy. This model clearly shows that how level 0 will tend towards obsolescence before the 2032-2034 period. It shows how level 1 will slowly give way to ADAS level 2, enabling this level to become the dominant level of automation in the short and medium terms.
Our model then considers the rise of higher levels of autonomy (level 3, 4, and 5). In particular, it considers the impact of shared autonomous vehicles and robotaxis on total demand for vehicles, showing that a peak car sales scenario can be anticipated in 2031/2. The comes about because a shared vehicle can service a higher mileage of travel demand than a private vehicle. The total vehicles sales are then forecast to fall beyond this point, creating complex and far-reaching questions for the global automotive industry.
We translate our vehicle and truck unit number forecasts into radar units. Here, we consider the radar content- for short/medium and long-range radars- per vehicle for each level of autonomy. The increase in radar content per vehicle will compensate for the emergence of peak-car. We also develop market value forecasts, considering a moderate and an aggressive price erosion scenario for short/medium and long-range radars.
Finally, we also segmented the unit number forecasts by semiconductor technology, showing how a technology transition has already taken place and how we are at the starting phase of another technology replacement round.