Azimuth response
Cut with elevation at 0°
Half-power beam width
Meet Euclid.
The radar for
frontier autonomy.
High-resolution radar with a digital-twin.
On your vehicle and in your world-model.
- Antenna array
- 8Tx × 8Rx
- Maximum range
- 300 m
ON THE ROAD
IN YOUR WORLD MODEL
Explore camera scenes alongside their synthetic radar response. Change the conditions and compare what the radar sees.
SYNTHETIC ENVIRONMENTS
Digital-twin also in CARLA.
Build with Euclid radar. Train with its digital twin.
MEASURE THE GAP
Fidelity you can
put a number on.
Compare generated radar with measured radar. Inspect the difference in signal strength, spatial fidelity, motion, and odometry—with the evaluation conditions in view.
Left: compare the number of detections per frame; a generated-to-real ratio of 1 means equal counts. Right: compare the distance from static radar points to lidar; lower is better. Together, these measure how many returns are generated and where they land.
Read the evaluation context
These research figures compare measured and generated radar. Material Properties shows SNR differences, not a direct material-property measurement: real-to-real compares successive measurements after motion compensation; real-to-synthetic compares generated and real returns.
Spatial Fidelity compares average detections per frame and directed distance from static radar points to lidar. Doppler compares wrapped Doppler error for matched moving-object clusters, grouped by range. Moving Objects compares radar point counts inside boxes from a lidar object detector, restricted to objects observed by the real radar and grouped by range. Radar Odometry compares radar-derived speed with LiDAR SLAM, grouped by vehicle speed. Lower distance and odometry error are better; matching counts alone does not establish positional or velocity accuracy.
The popup animations are schematic illustrations. Their point counts and wave patterns are not evaluation data.
The Riemann recordings demonstrate radar generation. Paired real-world measurements establish sim-to-real fidelity.
One radar.
On your vehicle and in your world model.
Euclid connects physical sensing to its digital twin. Collect raw radar data on your vehicle. Generate synthetic radar data with Riemann in your simulation stack.
- Azimuth steering span
- 120°
- Elevation steering span
- 32°
- Virtual channels
- 64
- Frequency band
- 76–81GHz
Loading antenna response…
Elevation response
Cut with azimuth at 0°
Half-power beam width
BRING RADAR INTO YOUR END-TO-END MODEL
Train inside your world-model.
Deploy reliably on the road.
Build with Euclid and its digital-twin. Train camera and radar together, end to end. Target lidar-level autonomy performance with lower sensor costs.