Towards Real-Time and Adaptable LiDAR Scene Completion

2026-08-17

AI Summary

This study addresses the limitations of slow inference and manual hyperparameter tuning in existing LiDAR completion methods by proposing RapidLiDAR. The proposed framework formulates initialization as a learnable component and replaces conventional point-neighborhood operators with voxel and Bird’s-Eye-View (BEV) features. By integrating adaptive displacement prediction with multi-scale feature querying, RapidLiDAR achieves real-time, high-resolution scene completion without requiring manual parameter adjustment. Experimental evaluations on the SemanticKITTI dataset demonstrate that the model attains state-of-the-art performance with an inference time of only 0.1 seconds per frame, representing a 2.3× speedup over prior methods. These results confirm that RapidLiDAR effectively satisfies the stringent 10 Hz real-time processing requirements essential for autonomous driving applications.

Abstract

LiDAR scene completion is a key component of 3D perception in autonomous driving, where the scene must be completed in real time to be usable in downstream tasks. Existing approaches typically follow an initialize-and-refine paradigm, in which a coarse initialization of the scene is first constructed, then refined into complete 3D geometry. Generative models are slower because they iteratively refine random Gaussian noise into the scene, while non-generative methods perturb the partial scene with a fixed noise scale, which limits coverage of large gaps and occluded regions and requires manual recalibration for each new sensor configuration. We present RapidLiDAR, a LiDAR scene completion method that treats the initialization itself as a learned, data-driven component. We propose an adaptive initialization module that predicts a spatially varying displacement for each partial input point, expanding the partial observations into a coarse scene initialization adapted to the local geometry, without requiring manual noise tuning. To refine this coarse initialization into a complete and coherent scene, we additionally propose a multi-scale reconstruction module that further refines point positions by querying multi-scale 3D voxel and 2D BEV feature maps constructed from the input scan. By replacing point-neighborhood operators such as farthest point sampling and $k$-nearest neighbor search with voxel- and BEV-based feature extraction, our architecture is faster and can handle different input resolutions by design. Experiments on SemanticKITTI and KITTI-360 show that our method achieves completion performance on par with the state of the art while completing a full scene in 0.1 seconds, which is 2.3 times faster than the fastest prior method. This matches the 10 Hz acquisition rate of typical automotive LiDAR sensors, taking a step toward real-time LiDAR scene completion.

Problem

• LiDAR scene completion
• real-time processing
• autonomous driving
• sensor adaptability
• occlusion handling

Innovation

• Adaptive Initialization
• Multi-scale Reconstruction
• Real-time LiDAR Completion
• Voxel and BEV Features
• Data-driven Displacement