Search

Automated driving: high resolution sensing radar is part of the playground

As a matter of fact, human being is responsible for 12.5 fatalities every billion of mile driven in the US. Autonomous driving is the automotive industry ultimate answer to make this number become as low as possible worldwide. Zero fatalities is the target industry is pursuing. Meanwhile advanced driving assistance has become an enabler for road safety improvement.

Yole Développement, releases this month the 2nd edition of its radar report focused on the automotive market segment. This new technology & market analysis, Radar & Wireless for Automotive: Market and Technology Trends 2019 points out the latest technical innovations and the status of the radar industry for automotive applications. Analysts invite you to get a comprehensive vision of the automated driving and its impact on this market.

Even from SAE level 0, an automatic emergency braking (AEB) system is now requested by NCAP and NHTSA to achieve the best score rating. As Automatic Emergency Braking test scenarios are becoming more complex over time from low speed AEB (so called AEB city), now including road intersections cases even under night and obscure lighting conditions and next year head on situation, there is a need for more sensors with better resolution and certainly sensor fusion.

As a first answer to safety incentive policy, OEMs have been committed to enable AEB feature on their cars

Cédric Malaquin

Technology & Market Analyst, RF Devices & Technology at yole développement

Number of them have made it available as a standard equipment for every car. As an example Nissan launched 7 models in 2018 equipped with a standard AEB feature for the US market representing 1 million of cars, 68% of Nissan sales for this market. Other features such as Blind Spot Detection (BSD), Rear Cross Traffic Alert (RCTA), Adaptive Cruise Control (ACC), Traffic Jam Assist (TJA), Highway Pilot (HP), Lane Keeping Assist (LKA) and Driver Monitoring (DM) complement the ADAS package.

Moving forward on automation, level 2+/level 3 cars are now available with Tesla model 3 or Audi A8 as the two most famous examples. Dramatically different in terms of conception, these two cars heavily rely on Radar sensing technology. Along with cameras, ultrasonic sensors as well as a Lidar, the Audi A8 can handle highway pilot and traffic jam assist without requiring the driver to handle the wheel for quite a long time. However the driver may be able to quickly take back the control of the car and a driver monitoring system is needed still most likely with a camera. At level 4 and beyond, the main embedded sensors will definitely be highly resolute, Radar ahead as it can efficiently sense in all environment conditions.

Radar is a polyvalent sensor. It can address from Long and Mid-range to Short and eventually Ultra-Short-range. Stepping back on Radar technology, since the beginning the vast majority of short-range and some of the mid-range radars were using the 24 GHz so called ISM (Industrial Scientific and Medical) frequency band. These systems were built around discrete components as for ZF/TRW with its AC100 or for Continental with its SRR-2. Representing almost 50 %, the discrete component and ICs were the biggest part of the cost breakdown.

The 77 GHz frequency band is mainly used on Long and Mid-Range Radar sensing to enable ACC and AEB, as high equivalent isotropic radiated power (EIRP) of 55 dBm is authorized worldwide on this band. Indeed, high power and linearity were the main specification given by Original Equipment Manufacturers (OEMs). Moreover, the available bandwidth of 2 GHz gives better range and angular resolution than the 250 MHz allowed on the ISM band. The main drawback of this technology was the cost which has partially been solved thanks to SiGe Monolithic Integrated Circuit (MMIC) technology introduction instead of GaAs technology. Radar SiGe MMIC mainly supplied by NXP and Infineon are now massively implemented by market leader Tier1s like Continental, Bosch, Denso or Aptiv (formerly Delphi). Indeed, SiGe is mainly an 8-inch wafer technology and the die packaging require special adjustment. The two main players have built these MMIC on fan-out packaging technology providing high thermal dissipation along with high RF performances. Fan-Out Technology like Redistributed Chip Package (RCP) from Nepes or embedded Wafer Level Ball grid array (eWLB) from Infineon allows the spreading of the ball array all around the die leaving an air gap below the dies thus reducing the parasitic effect of the PCB substrate.

It has been a first game changing step for the ADAS market, but with automated drive level 2+ and beyond will come the time to integrate more digital features into the RFIC, still with cost and RF performance in mind. RFCMOS has begun to satisfy these requirements. Digital content integration is indeed the main advantage of CMOS technology, and the price in high volume manufacturing is another one with medium to advanced technology node available in 12-inch wafer foundry. The challenge is on the RF design which has begun to be solved by the industry. RFCMOS performance is very close to SiGe in terms of transmit power, the remaining challenge will be to maintain a constant transmit power along the full frequency and temperature range. Several major players, Texas Instrument, Analog Devices and NXP have announced devices that bring a real game changing in the radar sensor market. Texas Instrument RFCMOS chip bring high resolution capability while integrating the transmitter, the receiver, the VCO, an Analog to Digital converter (ADC), a Digital Signal Processing (DSP) and a Micro Controller Unit (MCU) in a single transceiver packaged with a low cost Flip Chip Ball Grid Array (FC-BGA) technology. This is a huge step forward in Radar cost reduction from a system perspective. NXP RFCMOS chip include the receiver, the transmitter and the VCO features for the moment. The Tier1 Hella will introduce this new technology from 2021. Meanwhile, SiGe technology still is evolving and serving the market with more integrated designs. For instance, RXS8160PL chip from Infineon or STRADA770 chip from ST Microelectronics include the receiver, the transmitter, the VCO and an Analog to Digital converter on the same die.

Apart from Radar MMIC evolution, antenna integration is another challenge with technology disruption required. For instance, Aptiv made a differentiation by using cavity waveguides with H-Pol Radiator thus achieving a better form factor. We can expect novel antenna integration for the future.

Radar is best in class for speed measurement at long range and is the sensor of choice for ACC or TJA. However it still lacks of resolution when it comes to nearby objects discrimination and false positive elimination.

Stéphane Elisabeth

Expert Cost Analyst at System plus consulting

As a consequence, there is a need for high resolution Radar at all range. A typical approach with current hardware consist in dedicating a wideband antenna for near field resolution improvement while applying MIMO technics with other antennas dedicated to the long range. It can be found in Bosch current radar sensor. Long range and low resolution or short range and good resolution: that’s the tradeoff. Unless other technics are implemented to improve Radar resolution without sacrificing the range.

Such technic consist in dramatically increasing antenna aperture by scaling up transmit and receive channels of a Radar. Continental built its ARS-4 with such an approach. ARS-4 is able to provide high range resolution with high power at Long Range and Short range on the same board. This device is built around six transceivers with several receiver and transmitters working at the same time on the same clock supply by one voltage control oscillator (VCO) connected to two different kind of patch antenna, one for the Short and one for the Long Range detection. Continental has been very successful with this design. Pushing further away the boundary of scaling, Radar could reach angular resolution below 1° thus approaching a Lidar performance and it could also add elevation capability, a must have for Automated Drive level 3 and beyond. This could be a solution for long range Radar resolution dramatic improvement. New comers Arbe Robotics and Magna are building such systems.

Another approach to improve the short and mid-range radar resolution consist in increasing frequency sweep of the chirp. Using the 4 GHz of bandwidth available on the 79 GHz frequency band has been investigated for quite a long time in Europe and Japan. Due to regulation issues in the US, it has not been launched commercially so far. But with the FCC moving forward on it since end of 2017, it is now close to become a reality. Alps Electrics a Japanese Tier1 supplying GM recently got an FCC approval for 79 GHz Radar trial in the US, opening a path to unlock the US market on ultra-wide band 79 GHz Radar. Other manufacturers such as Ainstein, IntiBeam or WHST also build these types of 79 GHz tiny high resolution Radar. This not only helps in implementing Radar for Parking Assistance features, but it could also be used for Simultaneous Localization And Mapping (SLAM) providing accurate distance information to detected objects in real time. It would be helpful to complement geo-localization technologies for autonomous driving especially in urban canyon condition where GNSS technologies show some accuracy issue. Another advantage of 79 GHz Radar is the mitigation of interference issues that could happen when the streets will be loaded with Radars embedded in the cars. However, some technical challenges remain to be solved to fully exploit the 79 GHz band, for instance the design of wideband antennas.

Beyond the sensing market, the sensor fusion market also is shaken up. On current level 1 ADAS, camera and Radar data are computed at the edge in microcontroller, Field Programmable Grid Array (FPGA) or Vision Processor. Object class, position and speed are the output information separately sent from the different sensors to car’s electronic control unit. As sensors are proliferating around the car, this type of solution will no longer be a viable option in the near future. Aptiv was one of the first player to present a device close to the sensor fusion function by coupling on the same device a camera and a radar sensor with high resolution. Another thing is sensors sampling won’t necessarily be synchronized. Sensor fusion is expected to take over from level 2 with camera fusion first on MobilEye Vision Processor or Xilinx FPGA. From level 4, sensor fusion could include camera, Radar and Lidar data in a fusion platform such as the one promoted by NVidia. This type of solution is already used in high end robotic cars embedding super calculator close to a data center. This additional computing load could be supported by the car’s battery, also in the case of Electric Vehicle which have high voltage battery. However every Watt counts and this solution comes with a cost. High speed connection also remains to be developed in the car. So there is still room for sensor computing at the edge instead of having a centralized computing receiving full sensors raw data.

A later alternative likely more cost effective could be to compute data in the cloud thanks to 5G network and its low latency promises. But then 5G network should be more than ultra-reliable to support automated drive which ultimate goal is to achieve 0 fatalities.

Automated driving is definitely an exciting area with many technical possibilities attracting players from different ecosystems that will adopt different strategies to tackle this hype market.


Authors:

As a Technology & Market Analyst, specialized in RF devices & technologies within the Power & Wireless division at Yole Développement (Yole), Cédric Malaquin is involved in the development of technology & market reports as well as the production of custom consulting projects. Cédric graduated from Polytech Lille in France with an engineering degree in microelectronics and material sciences.

Stéphane Elisabeth, PhD has joined System Plus Consulting’s team in 2016. Stéphane is Expert Cost Analyst in RF, Sensors and Advanced Packaging. He has a deep knowledge of materials characterizations and electronics systems. He holds an Engineering Degree in Electronics and Numerical Technology, and a PhD in Materials for Microelectronics.

Related report:

Radar and Wireless for Automotive: Technologies and Market Trends 2019
The radar and 5G/V2X markets will both grow – one through market pull, the other through prospective enablement

Automotive Radar Comparison 2018
Continental, Veoneer, ZF, Valeo, Bosch, Aptiv, Denso, Ainstein: Discover the technologies used in the main Radar Systems and Chipsets

up