Forward Radar Removal Trend: “Eye” Corner Radar Is Enough

Retiring forward-facing radar on some vehicles may sound unacceptable, but automakers have reasons to do so.

While the front-end radar has numerous benefits as a foundational component of an ADAS system, there are good reasons why removing the forward-facing radar can clearly save hardware cost and weight, where it can be avoided. The advantage is that it saves not only the sensor itself, but also the brackets, wiring harnesses, power supplies, and other expenses associated with the sensor. Doing so simplifies packaging, freeing up the middle section of the grille for more flexible styling and easier thermal management. The architecture can be maintained, thereby reducing software development and integration costs.

Forward Radar Removal Trend: “Eye” Corner Radar Is Enough

OEMs can realize these benefits without compromising safety. OEMs can actually improve safety by using configuration options that are more effective than a single front-end radar and camera in many intersection and turn scenarios.

Two key ADAS technologies are fundamental to achieving this goal: advanced corner radar and sensor fusion. These software and hardware technologies enable performance breakthroughs through the application of AI and machine learning. Level 2+ autonomous driving requires systems that can handle difficult cornering situations that simple forward-facing radars and cameras would struggle with or be unable to handle entirely on their own.

advanced angle radar

In the past, corner radars, sometimes referred to as “short-range radars,” were primarily installed at the rear of vehicles to address blind spots and lane changes. However, this is no longer the case.

Usually corner radars are installed at the first two corners (see the picture below), this wide field of view enables the vehicle to perceive not only objects on the side, but also objects in front of and even behind the vehicle. Compared to forward-facing radars, this type of radar has greater situational awareness of objects to the side of the vehicle or slightly offset from the vehicle. For example, if a vehicle in an adjacent lane begins to perform a “close cut” (move quickly into the lane directly in front of the vehicle), the forward sensor alone may not detect the oncoming vehicle until the vehicle is clearly in the lane, causing the vehicle to suddenly brake The driver may feel that the vehicle does not recognize the lane change of other vehicles or is slow to respond.

Forward Radar Removal Trend: “Eye” Corner Radar Is Enough

broadened horizons

In fact, the two advanced-angle radars provide 250-degree coverage, providing significant overlap in front of the vehicle.

Forward Radar Removal Trend: “Eye” Corner Radar Is Enough

Collectively, the two front-angle radars provide a 250-degree perception that incorporates the advantages of all radars. Radar provides robust distance and speed detection in a variety of environmental conditions, including severe weather (rain, snow, etc.), poor light (night, tunnels, etc.), dust and dirt, while allowing OEMs to encapsulate the sensor behind the dashboard and in a small space.

When automakers add sensors, mmWave radar’s lower computational requirements (an order of magnitude lower than camera-based systems) will reduce cost, power consumption, and heat generation, eliminating the need for cooling. As individuals and governments pay more and more attention to the privacy issues of camera systems, the visual perception will be more inclined to choose radar, because radar does not have the possibility of privacy collection. As automakers continue to install rear corner radars in their vehicles, the combination of two front corner radars can provide 360-degree overlapping perception. In short, radar-based solution perception systems are more robust, economical, and flexible than other solutions.

sensor fusion

However, converting these sensor inputs into a comprehensive view of the vehicle’s surroundings is no easy task. A bigger challenge is that sensors rarely perform as well at the periphery of their field of view as they do on the “boresight” (ie, the axis that extends directly from the radar antenna surface).

Sensor fusion solves this problem, enabling software to combine the vehicle’s surroundings into a single model using input from multiple sensors. In vehicles with radars in the first two corners, the large fields of view of these radars begin to overlap in an area of ​​1.4 meters directly in front of the vehicle. The system can use sensor fusion to coordinate the echoes of two radars in an overlapping area, giving a high degree of confidence in objects in that area. Because each radar has a 150-degree field of view, the overlap area is large. In comparison, current cameras can provide a maximum field of view of 120 degrees, with further increases limited by megapixels and processing requirements to achieve sufficiently high resolution.

Forward Radar Removal Trend: “Eye” Corner Radar Is Enough

In this case, artificial intelligence and machine learning are the foundation to achieve the necessary performance. Artificial intelligence (AI)/machine learning (ML)-enhanced algorithms help vehicles make the most of these radar echoes and quickly and accurately identify objects in a wide field of view over long distances. Even though the echoes may be weak, a properly trained fusion algorithm can extract meaningful data and determine the position, velocity and size of distant objects.

Thus, AI/ML-enhanced sensor fusion enables the creation of next-generation “trackers” that utilize dual-angle radar for forward compliance capabilities. In addition to tracking objects across the combined field of view, the tracker ensures that any object moving within a modest blind spot directly in front of the bumper is recorded by fusing inputs from the front-facing camera and ultrasonic sensors.

The combination of two advanced corner radars and sensor fusion makes sense on a variety of vehicles as this can support basic active safety algorithms and some lower-level ones by reducing the reliance on vision and eliminating the need for forward-facing radar. Vehicle automation.

In short, using a dual-angle radar with sensor fusion and machine learning is an attractive opportunity for OEMs (elimination of forward-facing radar reduces cost and improves performance; this configuration can be applied to multiple vehicle variants) , to simplify packaging, integration and testing), OEMs are looking for a more cost-effective and seamless solution to use in their various models. While forward-facing radar is a pioneer in active safety, it is not a necessity in today’s software-defined vehicles.

Author: Yoyokuo