4D Imaging Radar Will Replace LiDAR?

At CES in 2023, 4D imaging millimeter-wave radars (referred as 4D millimeter-wave radar, 4D imaging radar or 4D radar) attracted attention. Many chip companies have launched or updated their imaging radar solutions, and radar system manufacturers have also brought their imaging radar products. Among them, Mobileye CEO Amnon Shashua said they only want 4D radars except for the front LiDAR by 2025 in his CES speech, which grabs industry insiders’ attention to the 4D millimeter-wave radar. Some industry experts even guess that 4D imaging radar will replace LiDAR in near future.

What is 4D millimeter-wave radar?

4D millimeter-wave radar, also known as imaging radar, can process height information in addition to the original function. It can detect the four-dimensional data of the objects’ orientation, distance, speed, and height. At the same time, 4D imaging millimeter-wave radar has a pixel-level angular resolution, which can resolve the outline of the target object, allowing the radar to achieve an imaging function similar to that of LiDAR.

In terms of performance, 4D radar is an upgraded version of traditional millimeter-wave radar. In terms of cost, the cost of 4D radar is only 10%-20% of that of LiDAR.

What are the differences between 4D millimeter-wave radar and traditional radar?

1 Due to the lack of height information, the traditional radar has only one plane in the viewing angle, and it is impossible to distinguish whether the target object is on the “road” or “in the air”. This makes millimeter-wave radar unable to effectively play its role in autonomous driving perception, especially in static object recognition. 4D radar greatly improves the confidence of millimeter waves in static object recognition.

2 With the ability to provide the height information of the target, 4D millimeter wave radar can capture the spatial coordinates and speed information of the target around the car, and also calculate the pitch angle information of the target, and then provide the environmental information around the car, and provide more functions such as realistic path planning, passable space detection.

3 Traditional radars have disadvantages such as low resolution and high noise. 4D radar can effectively improve the angular resolution and generate more point clouds and further image the basic outline, edge, and shape of the target object, by increasing the number of actual or virtual antennas. Traditional one also has point clouds, but the number is small, and there is no pitch angle information. 4D imaging millimeter-wave radar can provide pitch angle information and more point cloud data. After deep learning, 4D radar can also distinguish different targets such as pedestrians, bicycles, cars, trucks, etc.

Will 4D millimeter-wave radar replace LiDAR?

Whether LiDAR or 4D imaging radar can become the main sensor of intelligent driving vehicles in the future may still depend on specific scenarios. Different solutions will be developed according to different functions and needs. 4D millimeter-wave radar and LiDAR have their own characteristics and can play to their respective strengths. 4D imaging radar cannot replace the long-range high-performance imaging LiDAR, but it is expected that the side radar will be loaded on vehicles in near future.

1 From a technical point of view, the accuracy of 4D millimeter-wave radar is an upgraded version of traditional radar, but it is not in the same order of magnitude as LiDAR. Previously, there was a view that 4D imaging millimeter-wave radar might replace low-beam LiDAR in the future. In terms of price, the price of some 4D radars is not cheaper than Flash LiDAR, and the single-chip 4D radar is still in the early research and development stage. It is estimated that it might take 3 years or even longer to establish a standardized and mass-produced product. What 4D imaging radar might replace in the future should be traditional millimeter-wave radar, instead of LiDAR.

2 From the perspective of point cloud effects, although 4D millimeter-wave radar also has imaging functions, it still cannot compare favorably with the performance of LiDAR in terms of point cloud density and quality, and it is difficult to meet the perceiving needs of high-level autonomous driving. Considering the current time node, there is still a big gap between 4D imaging millimeter-wave radar and the main sensor LiDAR in terms of point cloud density and quality, and 4D imaging radar cannot replace the long-distance LiDAR for forward perception.

3 In the autonomous driving scene, the requirements for forward detection accuracy are extremely high. When the vehicle is driving forward, it must be able to accurately detect and identify the object information including size, distance, orientation, and speed in front. Any false positives or missed negatives may directly lead to accidents. LiDAR can form a 3D image model by scanning the surrounding environment and can generate millions of point clouds per second in the horizontal field of view. Its point cloud density is sufficient to meet the perceiving requirements of various levels of autonomous driving.

LiDAR and 4D imaging radar can complement each other. Previously, the industry had been arguing about the advantages and disadvantages of millimeter-wave radar, cameras, and LiDAR. Finally, everyone found that it is impossible to achieve automatic driving with a single sensor because each sensor has its own advantages and disadvantages. Only by fusing all the sensors can we get the precision and accuracy required for autonomous driving.