Original title: Stereo 4D Radar for 3D Object Detection: Integrating Geometric Alignment and Absolute Velocity Estimation
Seung-Hyun Song, Dong-Hee Paek, Woong-Chan Byun, Seung-Hyun Kong · September 2026
Radars help cars spot 3D objects in bad weather. However, standard 4D RadarA sensor that detects objects in three dimensions and measures their speed using radio waves. data has big flaws. It picks up clutter, noise, and fake reflections from roads, guardrails, and other cars. Cleaning out this clutter leaves the data too sparse. Also, a single radar only measures speed toward or away from itself, missing true overall velocity. To fix this, researchers built a stereo 4D Radar framework that uses paired left and right radars. This setup uses the geometric gap between both sensors to figure out an object's absolute velocity. It then merges complementary details from both sides. Testing showed this setup outperformed mono radar systems, gaining 8.82 points in 3D accuracy and 9.0 points in bird's-eye-view accuracy.
Imagine trying to guess how fast a car is passing by while you're standing on the sidewalk. You can only tell if it's coming closer or moving away, but you can't see its exact speed or direction from one spot. This is hard because a single viewpoint gives incomplete information. Here, using two radars is like having a friend watch from another angle. By comparing what you both see, you can figure out the car's real speed and path, not just the part moving toward you.
Self-driving cars need reliable sensors to work in rain or fog, where cameras struggle. This approach makes radar detection more accurate by understanding an object's full motion, which could prevent accidents in bad weather.
The abstract doesn't mention testing on public datasets or real-world trials beyond their in-house data, so how well this works in other scenarios isn't shown.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
Stereo 4D Radar for 3D Object Detection: Integrating Geometric Alignment and Absolute Velocity Estimation — Seung-Hyun Song, Dong-Hee Paek, Woong-Chan Byun et al.
Abstract licence: CC0
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.