As automated and intelligent systems advance to take on new workloads and applications, their detection requirements have evolved beyond simple data to awareness of the surrounding complex and dynamic environment. This presents two challenges for radar. First, conventional 24 and 77 GHz designs cannot readily achieve the data quality and density to support artificial intelligence (AI) and machine learning (ML) processing techniques. Second, the application space is growing rapidly and requires “open box” designs for deeper and flexible integration with more systems.

zPRIME from Zadar Labs closes these gaps by combining the latest RF Si technology with a software-defined architecture that elevates the performance of the radar from hardware into software to create a software-defined imaging radar (SDIR). Zadar Labs sees SDIR as the foundation enabling broad and quick expansion of radar applications.

Figure 1

Figure 1 Two cars and their respective zPRIME radar point clouds. The color reflects target elevation.

In nearly all applications for intelligent systems, the objective is to create a “digital twin” to perceive and model the environment with enough accuracy, precision and recall to be useful for an automated workload. To do this, the sensor array—typically composed of a combination of radar, LiDAR, camera and location sensing—will actively and passively collect feedback from the surrounding environment. To do this intelligently, meaning with awareness of context, the sensors must dynamically respond to understand and optimize for the current scenario. zPRIME uses real-time processing to update the radar configuration within milliseconds, optimizing for the current scenario. Linking the processing and sensor is unique to SDIR and requires higher performance than the typical radar; the data from the radar must be of sufficient quality and density to feed the algorithms used with a cognitive radar.


Zadar Labs’ zPRIME is currently leading by demonstrating the potential of SDIR: zPRIME detects vehicles out to 800 meters with 0.4-degree resolution in azimuth and elevation. While Zadar uses available hardware and components like other 4D imaging platforms, its patented algorithms and software provide significantly better performance. With deep access to the processing chain, the platform is versatile and powerful enough for critical radar applications. The Zadar team developed a proprietary processing stack named zVUE that manages both the signal and data processing, using AI/ML algorithms for dynamic operation and advanced radar features. These include object tracking, classification, simultaneous location and mapping and segmentation. Able to tap into data at any point in the processing chain, Zadar’s solution supports almost any form of sensor fusion. Further, zVUE’s capabilities for higher level data processing are independent of the hardware, meaning they can be applied to other radar platforms with relatively little tuning.

Figure 2

Figure 2 Point clouds of a LiDAR sensor (left) and the Zadar sensor (right).

The zPRIME platform fulfills the stringent radar requirements for fully autonomous operation, called L5 autonomy. zPRIME’s state-of-the-art resolution produces a LiDAR-like point cloud that reveals the outline of a target, whether human or vehicle (see Figures 1 and 2). The radar covers a wide field of view of approximately ±65 degrees in azimuth and ±12 degrees in elevation and has demonstrated vehicle detection out to 800 meters in a suburban environment. zPRIME uses a unique chirp-coding waveform for unambiguous Doppler detection per scan to greater than 220 MPH. It is also near immune to interference: if interference is detected, zPRIME frequency hops and adjusts timing to eliminate continuing exposure. All this capability is packaged in a form factor of approximately 15 × 12 × 3 cm.


zPRIME was designed to house most of Zadar’s SDIR processing at the edge; zVUE provides significantly higher data quality and information to the upper-level system. For autonomous mobility, the data is typically used for perception and navigation. Zadar divides the processing chain into kernels, applying bespoke algorithms to provide advanced information based on their respective locations in the chain. This information includes the following:

  • Estimated vehicle odometry without GPS or inertial measurement unit data
  • Static-dynamic detection performed on each scan to differentiate between static and moving objects
  • Prediction, comprising a tracked object list with optimized cluster and track data
  • Online calibration and tuning. The radar has intrinsic and extrinsic online calibration, and the SDIR software stack adaptively tunes the RF front-end and signal processing kernels to the environment, providing enhanced precision and recall
  • Object classification using the AI radar algorithms in zVUE. Using the detection signatures of the detected objects (see Figures 3 and 4), in addition to temporal characteristics, Zadar’s algorithms use a detection profile for classification.
Figure 3

Figure 3 Range-Doppler profile of a person. The signature of the relative hand movements can be used to classify the object as a person.

Figure 4

Figure 4 Range-Doppler profile of a moving sedan, showing a clear difference in the speed of the wheels vs. the vehicle body.


zPRIME can provide up to 20,000 useful detections per scan. When combined with the Zadar SDIR, the platform greatly reduces computational workload, enabling adaptation to support many applications in multiple industries.

zPRIME’s resolution, feature set and post-processing kernels set a development model for Zadar’s current and future platforms. The Zadar SDIR has the performance to support demanding autonomy applications, yet it remains flexible as a platform to enable adaptation and optimization. The SDIR can be used for applications from autonomous vehicles and trucks to heavy equipment and infrastructure.

Zadar’s vision is to help build a future where the quality of life is radically improved through safer and more effective machine sensing and cognition using radar vision. Seeing the trend toward SDIR, Zadar has developed a platform to realize SDIR’s resolution, flexibility and scalability.

Zadar Labs
San Jose, Calif.