Channeling the Future of Radars

The advancements in 4D imaging radars

Car makers are racing to bring the best advanced driver assistance systems (ADAS) and automated driving (AD) solutions to market, and 4D imaging radars are front and center. However, when technologies appear overzealous claims are quick to follow. What challenges do a higher number of channels solve? Does such a solution have any drawbacks? Do you need 2000+ channels for higher levels of automation?

Radars have been a mainstay of ADAS and automated driving since the earliest days. Of course, radars are constantly evolving to meet the demands of ever more complex driving situations. Traditional automotive radars supplied three dimensional measurements. This included the relative speed of objects and their location on the radar’s detection horizon (azimuth).

These measurements are enough for a host of driver assistance (e.g., adaptive cruise control, ACC) and active safety features (e.g., automatic emergency braking, AEB). Such functions have already made roads around the world more comfortable, safer, and saved thousands of lives.

However, to safely increase driving automation, automated systems need more accurate data and more of it. Thus, the 4D imaging radar was born. 4D because they add a new measurement, elevation angle, into the mix. Imaging, because their high-density point clouds reconstruct a more detailed image of their surroundings than traditional radars.

The need for 4D imaging radars for level 3 automation and higher is widely acknowledged in the automotive industry. However, the devil’s in the details. 4D radars come in many shapes and sizes, all have their benefits and drawbacks.

Many suppliers have drastically improved the resolution of their current radar generation by increasing the number of channels their radars emit, to over 2000. Some other suppliers, have followed the alternative path, increasing channel count gradually, nearing only 300 channels now, to balance performance and costs. Both methods result in 4D imaging radars that can enable high levels of automation. So how can OEMs make the most of 4D imaging radars, without ballooning costs and development complexity?

What challenges does a higher number of channels solve?

Automotive radars traditionally operate in the 77GHz range, and their resolution is inherently limited by this wavelength. Increasing the number of channels is one way of countering this. With more channels the radar simply generates more data, trying to build a higher resolution point cloud (image) of the environment.

With a more detailed point cloud these radars, detect smaller objects more reliably. This is necessary for automated driving, especially in urban environments. A higher channel count also opens the opportunity of virtually increasing the antenna aperture, resulting in a wider field of view. This in turn can help improve the angular accuracy of the radar, that is, the positioning of an object within the field of view.

Higher channel counts thus lead to more accurate sensors. Where’s the catch?

Does such a solution have any drawbacks?

Increasing the channel count has consequences. First, the sensor is bigger, having to incorporate more parts. Incorporating radars into vehicle design is already a unique problem for OEMs. Making the package larger can significantly complicate vehicle design, for example affecting air flow when the radar is in front of the cooling radiator. Furthermore, larger sensors do not comply with overall weight reduction targets for vehicles.

Second, making the package smaller is difficult. The need to accommodate a higher number of antennas on a limited surface creates even more challenges. Each antenna must be made smaller and less directive. As a result, the antennas are less isolated from each other leading to potential noise issues.

Far Looking Satellite Camera Family FSC300

Processing the signals from over 2000 channels also brings with it relatively large compute performance demands. Higher performance requires more power, a drawback especially clear for electric vehicles. But performance also generates more heat. Not only does a larger sensor have to fit into the tightly constrained environment of a car body, but it also must dissipate more heat, in less space.

Finally, when it comes to 2000+ channel radars, often not all antennas are active at the same time. As a result, ADAS functions are not actually benefiting from the full accuracy improvements claimed by radar suppliers.

As with all ADAS functions, engineers must find the best compromise. One that enables the highest levels of driver safety and automation, without straining the limits of technology, production, and vehicle design.

At Continental, our radar experts leverage their decades of experience in creating automotive radars to support our Customers in rolling out cutting edge ADAS functions. We support OEMs in avoiding the challenges of high channel count radars, without having to sacrifice performance. Continental’s 5th and 6th generation radars offer outstanding 4D imaging capabilities with far fewer than 2000 channels.

Do you need 2000+ channels for higher levels of automation?

Car manufacturers are already proving that they can reach level 3 (L3) automation with fewer channels. A premium German OEM is launching its L3 pilot in March 2024 for highway driving supported by Continental’s ARS540 radar. Continental’s radars ensure 300 meters of range and a wide field of view, enabling advanced automation and active safety features.

But not only German car makers trust Continental radars. Respected OEMs based in the US, Japan and Europe have also built vehicles using out 4th or 5th generation radars. Furthermore, our radar knowledge also brings advantages to portions of our wide-ranging partnership with Aurora. Among others, the reliability and robustness of our sensors will help enable around the clock trucking, while our industrialization and production expertise will support the production of L4 systems at scale.

Higher levels of autonomy will require more channels to be safe. However, suppliers should gauge the best balance of performance and cost for each generation and only increase channels as needed. For current generation radar use cases, Continental experts believe, the best balance lies far below the 2000 channel mark. Utilizing AI powered perception algorithms, sensor fusion, and new sensor modalities will allow OEMs to achieve higher levels of autonomy. Following this path, they will sidestep the major drawbacks of high channel counts in both design complexity and cost.

The future of ADAS is intertwined with that of radars. Over the past decades, Continental has gained experience as one of the world's most trusted automotive radar suppliers. Through this knowledge, our engineering teams are uniquely positioned to develop sensors that deliver the performance needed for any given function. The alternative? Paying a premium for radar improvements that simply don’t translate to real-world performance gains. Or, researching different wavelengths for new sensors, but that’s a topic for another day.