Radar transceiver: ADAS / Autonomous Driving key components
Blog 1: Why do we need radar?
Blog 2: Basics of FMCW Radar
Blog 3: Distance measurement: How far can the radar detect?
Blog 4: Radar resolution: How accurate is the radar?
Blog 5: Radar architecture: Connect multiple radar sensors
Evolution of architecture
In previous radar blogs, we have covered the significant increase in the number of sensors in general, especially radar modules, in order to achieve a higher level of autonomy and safety. As shown in Figure 1, additional radar modules are expected to be installed in the next few years, based only on level L1 driver assistance and NCAP (new car assessment programme) 1-4 forward monitoring radar. From this configuration, it will be compatible with level L2+ driver assistance and NCAP 4 to 5 for standard cars, and level L3 to L4 autonomous driving and NCAP5 for luxury cars.
With the centralization of vehicle functions, the performance of the central processing unit is expected to improve rapidly. In that case, if you don't do computational processing with the sensor module, you will be able to do data processing more efficiently. This leads to the evolution of the E/E architecture of the car into a decentralized architecture. Of course, the transition to a fully decentralized architecture takes a long time and is expected to be completed after 2030. Until then, partial implementation will proceed in the market.
The first step is to install a domain controller (DCU) for specific functional applications such as ADAS. After that, the number of DCUs will increase depending on the level of autonomous driving. When level L2+ driver assistance is required, a zone controller will be introduced along with the DCU. Ultimately, it will be a centralized E/E architecture connected to the central computer through a zone control unit where the sensor is connected. This evolution is shown in Figure 2. Of course, this requires the rapid complexity of the software and the large capacity of the in-vehicle network.
Figure 2. Megatrends of E/E architecture (Source: David Xu, Renesas Electronics)
In the newly introduced E/E architecture, part of the radar processing is not done on the radar module (edge computing) as shown in Figure 3, and the calculation will be done efficiently by delocalizing it. The amount of processing in each module or control unit is determined by the required performance and the available architecture.
Figure 3. Radar data processing process
Smart radar sensor
Today's radar architecture consists of individual independent radar modules placed around the vehicle. Each module is equipped with a radar transceiver and has the ability to process detected data on-board. As shown in Figure 4, it can be processed with a single chip or another microcontroller on the same module or So I'll process it with C
The processed radar data is transferred from each "smart radar sensor" via the CAN bus and fused with a distant DCU. Figure 5 shows the various processing processes and their execution order.
Figure 5. Radar signal processing with smart radar sensors and domain control units
If the number of sensors is sufficient, you can recognize obstacles around the vehicle. For example, in Figure 6, the ADAS ECU receives information on objects detected by the forward long-range radar (LRR) and information from the short-range and medium-range radar (SRR/MRR) at the four corners, and creates a whole image of the surroundings.
Figure 6. Example of radar architecture with smart sensor (edge processing) and domain controller
Satellite radar sensor
With the introduction of a centralized architecture, some of the data processing will be separated from the radar module in the future. As shown in Figure 7, the satellite radar unit performs only some of the received radar signal processing, such as range FFT, before transferring data to a distant ECU (domain or zone controller). In this case, the radar module itself will not be able to say "smart".
Figure 7. Centrally treated satellite sensor
Part of the data processing is performed on each satellite radar module, and the ECU receives the data and performs the main radar data processing process for each received data set. The processed results can be fused with information from other sensors to increase the accuracy of recognition. Figure 8 shows each process.
Figure 8. Radar signal processing in satellite radar and central ECU
In the implementation of a centralized architecture using a satellite radar module, the data is transferred to the ECU via the vehicle's Ethernet network. Forward surveillance radars and imaging radars (LRR/Imaging) require a large amount of radar data, so smart sensors are used, and all data processing of the radar is still performed at the edge. On the other hand, multiple satellite radars are placed around the vehicle so that it can recognize the 360° perimeter. For example, in the architecture in Figure 9, the corner radar (SRR/MRR) and ultra-short-range radar (USRR) preprocess the received data, compress and transmit the data if necessary, and perform further data processing in the zone ECU. I will.
Figure 9. Example of zone ECU-based radar architecture with satellite module and remote processor
Using satellite radar modules in a centralized architecture gives you a variety of benefits.
First of all, the satellite radar module will be simpler and the size and cost will be reduced. It is also easy to repair and upgrade if necessary. In particular, modules placed near hot parts like headlights will also reduce the problem of heat generation.
By using the automotive Ethernet network for data transfer, you can optimize the cost and weight of the cable. Of course, Ethernet transmission requires security measures. In addition, it is a data format that is easy to store and process together with data from other sensors.
Using the vehicle's control unit for data processing not only greatly improves the processing efficiency of radar data, but also enables more complex operations. Using data fusion with other sensors such as cameras and riders, machine learning and artificial intelligence, you will be able to optimize the sensing and analysis of the surrounding environment, and you will be able to contribute to the realization of more advanced autonomous driving.
A summary
In the future, the E/E architecture of the vehicle will evolve into a centralized computing solution. In a few years, different architectures will coexist, and both smart sensors and satellite radars will be used. The main differences are summarized in the following table.
Table 1. Comparison of smart sensors and satellite sensors
Radar Module "Smart Sensor" Centralized (Satellite)
Processor location Radar module/chip Zonal/Central ECU
Radar output Processed data (objects) Processed data (point cloud) Raw data/Range FFT
Network type CAN 100Mb Ethernet Gb Ethernet
As mentioned earlier, in order to migrate to a centralized architecture with higher computational performance, high-speed communication must be available throughout the vehicle. In the near future, this architecture will depend on the vehicle's Ethernet network. However, MIPI A-phy is also a candidate for future generations.
Renesas has ECUs based on the R-Car Gen4 series and innovative components for smart radar and satellite radar units, and will take on providing advanced solutions for future automobiles. I'm working together.
The Winning Combinations
Imaging radar for ADAS & AD