TI Smart Sensors Designed for Autonomous Driving Applications

By Sneha Narnakaje / Texas Instruments
Automotive Radar Analysis and Processor Product Manager

The automotive industry is driving innovation and technological advancements in robotics and machine vision. Automakers are designing new vehicles with a variety of technologies to meet the ever-growing consumer demands. This trend has created the need for advanced driver assistance systems (ADAS), which improve safety, comfort, convenience, and energy efficiency.

According to government agencies such as the National Highway Traffic Safety Administration, more than 30,000 people die in traffic accidents in the United States and more than 1.3 million worldwide each year. 94% of these accidents are related to human error, and ADAS that help with warning, braking, monitoring, and steering can reduce driver error.

Many of today's cars have features such as blind spot and lane departure warning, forward and rear cross-traffic collision warning, automatic emergency braking, lane keeping, and adaptive cruise control. These features are differentiated from brand to brand and are becoming a revenue stream for automakers, and some countries are now mandating ADAS on all vehicles by 2020.

ADAS, the foundation of autonomous driving

The demand for ADAS is rapidly increasing due to the rising safety awareness, the influence of regulations, and safety ratings from OEMs. According to the global ADAS market outlook by Research and Markets, approximately 50 million vehicles equipped with ADAS were shipped in 2016, and this number is expected to reach 60 million by 2022. According to another ADAS market outlook by Research and Markets, the shipment volume of ADAS parts is expected to increase from 218 million units in 2016 to 1.2 billion units in 2025. A typical ADAS implements various sensing technologies along with advanced processing and communication functions to automate, adapt, and enhance vehicle systems to improve safety and driving functions.

Automakers are turning to leading semiconductor suppliers to supply automotive electronics, from advanced sensing and imaging/vision technologies to high-performance, low-power processors and in-vehicle networking. The maturation and advancement of ADAS components will ultimately enable semi-autonomous and autonomous vehicles. Figure 1 summarizes the six levels of autonomous driving as defined by SAE International.

Autonomous driving systems are based on various components, including sensors that capture information about the vehicle's surroundings, integrated circuits (ICs) for communication, high-performance processors that analyze sensor data, and microcontrollers (MCUs) that activate and control mechanical operations.


The reason why sensing systems are so important for ADAS and autonomous driving is because they add intelligence to the vehicle, allowing it to accurately perceive its surroundings. While various image sensors are becoming standard for ADAS, new sensing technologies such as radar, laser, ultrasound, infrared, and lidar are also enhancing ADAS.

The automotive industry prefers radar sensors because they can pass through non-metallic objects such as plastics, clothing, and glass, and are largely unaffected by environmental factors such as fog, rain, snow, poor lighting conditions, or bright light conditions. Automotive radar systems can be divided into short-range, medium-range, and long-range radars based on the object detection range. Ultra-short-range radar (USSR) is also emerging as an ADAS application for parking assistance systems. Driver assistance features such as blind spot and lane departure warning use short-range radar (SRR) systems.

These systems are SAE International Level 1 and are designed to report or warn the driver using light-emitting diodes (LEDs) or steering wheel vibrations. Current SRR systems operate at frequencies of 24-29 GHz, but industry experts say they may be phased out in the future due to regulations regarding output power at lower frequencies.

Driver assistance features such as adaptive cruise control and automatic emergency braking use long-range radar (LRR) systems. These systems perform simple vehicle control actions. Current LRR systems operate at 76–77 GHz, but higher levels of autonomous driving require better range and resolution, so forward-facing radar systems will use both 76–77 GHz and 77–81 GHz frequencies when combined with LRR and newer mid-range radar (MRR) systems. At higher levels, radar sensors will be needed to analyze complex scenarios, such as detecting hazards, measuring their properties (distance and velocity), and classifying them into objects with unique properties (distance, velocity, angle, height). Finally, the sensors must support safe operation.

TI's AWR1x millimeter wave (mmWave) sensor portfolio enables developers to drive more safely and easily. Based on the mmWave sensing architecture ( Figure 2 ), the AWR1x sensor integrates radio frequency (RF), analog functions, and digital control capabilities into a single chip.


Automotive radar system developers can achieve functional safety specifications by utilizing on-chip built-in self-test (BIST) capabilities. Devices in this portfolio also integrate customer programmable MCUs and single processing functions via hardware accelerators or DSPs. Optimizing radar sensor designs to the level of integration of the AWR1x can reduce size and power.

Ultra-high-precision automotive radar sensors capable of analyzing complex and intense urban driving scenarios use electromagnetic waves with a frequency of 76–81 GHz to determine the distance, speed, and angle of objects in the sensor’s field of view. Several parameters determine the performance of a radar system in terms of distance, speed, and angle, of which resolution and accuracy are the most important. Resolution is the ability to distinguish two objects by distance, speed, or angle, while the other two parameters are the same for these objects. For example, angular resolution is the ability of a radar sensor to distinguish between two vehicles traveling at the same speed and the same distance apart. Precision is the accuracy of the distance, velocity, or angle measurement of an object. SAE International Level 2 and above require ultra-high precision radar sensors for SRR applications (50 m).

Automotive radar systems use frequency-modulated continuous waveform (FMCW) techniques to measure the range, angle, and velocity of distant objects. In FMCW radar, chirp linearity refers to the accuracy of the distance measurement to an object. Conventional mmWave sensors generate chirps based on open-loop voltage-controlled oscillators (VCOs), which cause chirp nonlinearity, resulting in inaccurate distance measurements. The AWR1x mmWave sensor portfolio is also based on FMCW techniques. This uses a closed-loop PLL to enable 0.01% linearity and precise chirps, resulting in improved range accuracy and increased range resolution. Chirp linearity helps prevent false detections and ghost objects, such as the disappearance or secondary images of the actual target.

Range resolution is a function of RF bandwidth. The AWR1x sensor portfolio supports chirp bandwidths up to 4 GHz in a single sweep, enabling range resolutions of less than 5 cm, which is three times more precise than current mmWave solutions on the market.

Clear velocity is the ability to distinguish objects of similar speed. For a given range resolution and maximum range, increasing the maximum speed requires a higher IF bandwidth. The high-performance radar front-end of the AWR1x portfolio supports an IF bandwidth of 15 MHz, enabling a maximum range of more than 250 m and a speed of up to 300 kph. The combination of IF bandwidth and phase noise performance allows the radar sensor to detect small objects around large objects. Additionally, the built-in 20GHz synchronization capability for phase coherence in the high-performance front-end allows multiple front-ends to be staged to achieve sub-1 degree angular accuracy even in harsh urban driving environments, and to provide better elevation estimation when driving under bridges or in tunnels.

To be able to analyze the complex scenarios surrounding highly autonomous driving, future radar sensors will need to be extremely precise. The capabilities of the AWR1x portfolio enable the design of extremely precise sensors.

Versatile intelligence that adapts to changing environments

Automotive sensor manufacturers are turning to multimode radar systems to address SAE International Level 2 and beyond. In a multimode radar system configuration, sensors are designed to support both MRR and LRR configurations in a single sensor module, which enables automakers to achieve significant cost savings, as two separate sensor modules are no longer needed to support each configuration. Multimode radar system design imposes specific requirements on mmWave technology suppliers, including ease of use, flexible chirp configuration, and monitoring.

The AWR1x portfolio integrates a BIST engine to provide real-time local control of chirp generation parameters. The engine supports dynamic chirp configuration via non-real-time messaging to a local digital subsystem or an external host processor. The BIST engine automatically adapts the sensor to changing environmental conditions such as temperature and aging. This enables self-calibration of drift in RF parameters such as output power and gain. Additionally, the BIST engine enhances safety by continuously monitoring RF and analog subsystems for key RF performance parameters.


While existing mmWave sensing technologies have leveraged real-time baseband architectures, the AWR1x sensor delivers system-level performance benefits with a novel hybrid baseband architecture. Since automotive radar sensors are mounted beyond the bumper, if the sensor can accurately compute bumper reflections, it can remember the bumper signature and recalibrate at every boot. The AWR1x portfolio can more accurately compute nearby objects by leveraging the zero-range magnitude of bumper reflections and the phase of bumper reflections, which is nearly impossible with a practical baseband architecture due to the low frequency of bumper signatures. The hybrid baseband architecture also enables robust radar sensor design by monitoring the image band and detecting interference from other jamming radars without ambiguity about the real object.

Miniaturized, low-power radar sensor made possible by a true single chip

As autonomous driving becomes a reality, radar sensor requirements will be driven by power, size, cost, range, and accuracy. Autonomous driving systems for SAE Level 2 and above require far more radar sensors than current solutions, and today’s high-end vehicles have multi-chip single radar systems. Assuming multiple individual components, these radar systems are large and bulky, and need to be smaller, lower-power, and more cost-effective. To meet the demands of the future autonomous driving market, these sensors must be miniaturized and optimized.

Some of the radar systems on the market today are called single-chip solutions, but in reality, they are not. Current solutions still require a large number of components. Although the number of individual chips has been reduced from three to one, a transceiver with an external MCU or DSP is required to process the radar data.

CMOS technology allows TI to integrate an intelligent radar front end with MCU and DSP functionality into the AWR1x single-chip portfolio. Processing is integrated into the front end, reducing the radar system size and form factor by up to 50 percent. This allows for efficient integration of multiple radar systems. CMOS technology and best-in-class power management techniques enable the low power of the AWR1x sensor, which is essential for the automotive industry’s development of energy-efficient electric vehicles. The low power also has cost benefits, as designers can now choose more economical and lighter housings. It also allows the AWR1x sensor to withstand higher ambient temperatures and improve sensor reliability.

All of these capabilities benefit customers only when delivered in a reliable package that is mass-manufacturable. The AWR1x mmWave portfolio is available in an automotive-friendly flip-chip ball-grid array (FC-BGA) package. This FC-BGA package solution delivers reliable electrical, mechanical and thermal performance, eliminates shielding for dissipation, and eliminates the need for underfill, a material that protects the bottom of the chip to protect interconnects, thereby offering cost advantages over existing packages used in mmWave sensing technology.

Supported by the AWR1x mmWave portfolio
ADAS and body and chassis, interior applications

TI’s AWR1x mmWave portfolio supports high-precision sensing applications in ADAS, body and chassis, and interior applications. The portfolio ranges from a high-performance radar front end (AWR1243) to single-chip radar solutions (AWR1443 and AWR1642). Table 1 summarizes the key features of each sensor in the AWR1x portfolio.

Comprised of three devices, this portfolio supports a variety of ADAS radar sensor configurations, from USRR, SRR, MRR to LRR and imaging. It also enables smart sensor architectures where all radar processing occurs at the edge. In a satellite sensor architecture, radar sensors send object data to a central processor via CAN-FD for further processing and sensor fusion.


Figure 3 shows the AWR1x mmWave sensor categorized into applications such as ADAS, body and chassis, and interior, based on the range and type of target object detected, as defined by radar cross section (RCS).

conclusion

The sensor is unique in this market because it allows developers to choose the solution that best suits their design needs. This level of integration and small footprint allows designers to add new features to existing applications. As the market adapts to ADAS and autonomous vehicles, TI’s mmWave AWR1x sensor portfolio will provide the flexibility needed.

More Information: For more information about this portfolio, visit www.ti.com/mmwave.