As automakers pursue ever higher levels of autonomous driving and strive to meet increasingly stringent safety regulations, they’re reconsidering their approach to radar-sensor architectures.
The sensor and its downstream components perform three critical tasks: sensing, which involves using radar to detect objects and perform basic preprocessing functions; understanding, which includes functions such as object detection and tracking; and acting, such as applying an automatic emergency braking system.
A common approach to these tasks is to have highly intelligent sensors process data at the edge and stream the processed data to an advanced driver-assistance system (ADAS) electronic control unit (ECU) over a Controller Area Network (CAN) bus or 100-Mb/s Ethernet interface (Fig. 1). The sensors in such architectures include processors that can handle tasks like object detection, classification, and tracking. In addition, they include accelerators to handle compute-intensive tasks such as fast Fourier transforms (FFTs).
Satellite Architecture Benefits
With the proliferation of sensors around the vehicle, the edge architecture is giving way to a satellite architecture in which each sensor streams minimally processed data to a high-performance central ECU over a 1-Gb/s Ethernet interface (Fig. 2). The satellite architecture enables the efficient implementation of advanced sensor-fusion algorithms to improve decision-making accuracy.
A recent Texas Instruments article draws an apt analogy: The human brain makes decisions based on inputs from both eyes, rather than requiring each eye to make a decision independently. The use of a powerful central ECU paves the way for the deployment of sophisticated algorithms that can increase the resolution of distributed-aperture radar as well as implement machine learning (ML) for object classification.
Satellite architectures also enhance modularity and scalability. For example, manufacturers can easily deploy a limited number of sensors on an economy vehicle focused on safety features and scale up to a larger array of sensors to add greater levels of autonomy to a premium vehicle. A satellite architecture facilitates vehicle model differentiation through software, too, while enabling over-the-air software updates to improve performance and safety.
Radar Sensors
To support the evolution from edge to satellite sensor architectures, TI offers sensors for both configurations. The company’s lineup includes long-range, medium/short-range, and ultra-short-range radar sensors as well as radar sensors for imaging and driver vital-signs monitoring. The devices are available with a variety of frequency bands to support compliance with worldwide regulatory requirements. Software compatibility and pin-to-pin device compatibility across the radar-sensor lineup, coupled with application-related reference designs, simplify development and speed time-to-market.
Radar-on-Chip Sensor
A specific device in Texas Instruments’ portfolio is the new AWR2544 76- to 81-GHz frequency-modulated continuous-wave (FMCW) satellite radar-on-chip sensor (Fig. 3). The device includes radar data-processing elements, a set of peripherals for in-vehicle networking, and a launch-on-package (LOP) antenna feature that facilitates the attachment of antennas directly to the device. The AWR2544 is built with TI’s low-power 45-nm RFCMOS process, which enables high levels of integration in a small form factor, minimizing the final design’s bill-of-materials (BOM) count.
The device includes an RF/analog subsystem and radio-processor subsystem as well as a main subsystem based on a user-programmable Arm Cortex-R5F processor that enables custom-control and automotive-interface applications. Simple programming model changes can reconfigure the device for various functionalities, such as short-range, mid-range, or long-range sensing, and customers can employ dynamic reconfiguration to implement multimode sensors.
In addition, the device includes a hardware accelerator block to supplement the main subsystem by offloading common radar processing functions such as FFT, scaling, and compression. Secure variants of the device include a hardware security module, which consists of a programmable Arm Cortex-M4 core and the necessary infrastructure to provide a secure zone of operation within the device.
Evaluation Module
To help get you started with designs based on the AWR2544, TI offers the AWR2544LOP evaluation module, which contains everything you need to begin developing software for the on-chip Cortex-R5F controller and hardware accelerator. The module integrates on-board emulation capability for programming and debugging as well as on-board buttons and LEDs that let you quickly create a simple user interface. The module also includes a 3D waveguide antenna, Ethernet and USB cables, jumpers, and mounting hardware.
The AWR2544LOP can team up with the DCA1000 evaluation module, which provides real-time data capture for two- and four-lane low-voltage differential signaling (LVDS) interfaces from the radar sensor. The DCA1000 can stream the data via 1-Gb/s Ethernet in real-time to a PC running the TI’s MMWAVE-STUDIO tool for capture and visualization. You can apply the data to an application of your choice for data processing and algorithm development.
Full Radar Sensor Portfolio
In addition to the AWR2544, TI offers a full portfolio of automotive mmWave radar sensors. The single-chip AWR294x, for example, also operates in the 76- to 81-GHz band and integrates an FMCW transceiver, radar data-processing elements, and peripherals for in-vehicle networking. The associated TIDEP-01027 reference design demonstrates using the device in corner and front long-range radar systems for functions such as blind-spot detection, front cross-traffic assist, and lane-change assist.
On top of that, the company offers the AWRL1432 single-chip low-power 76- to 81-GHz automotive mmWave radar sensor. This device includes multiple power domains that can be switched off individually depending on use case. Applications include child presence detection, intrusion monitoring, gesture detection, and occupancy detection.
Conclusion
As sensors proliferate throughout vehicles to enhance ADAS applications, automakers are moving away from edge architectures toward satellite architectures where much sensor data processing occurs at a central ECU. TI has a full portfolio of automotive radar sensors for both architectures along with reference designs and evaluation modules to help you manage the edge-to-satellite evolution.
