LiDAR scene flow is the task of estimating point-wise motion from sequential point clouds. It provides critical information about moving objects and their trajectories, making it essential for autonomous driving.

The paper introduces CorrelationFlow, a training-free geometric approach for this task. Unlike learning-based methods, it does not require training, which may simplify deployment and improve generalization.

The work is presented on arXiv, but the abstract does not include experimental details. The authors position the method as a geometric alternative to existing approaches.