Sense Less, Generate More: Pre-training LiDAR Perception with Masked Autoencoders for Ultra-Efficient 3D Sensing
Sina Tayebati, Theja Tulabandhula, Amit R. Trivedi
TL;DR
Problem: LiDAR energy consumption limits long-duration autonomy. Approach: Radially Masked Autoencoding (R-MAE) pretraining that masks angular sectors and predicts voxel occupancy with a 3D sparse encoder–decoder and an occupancy loss. The method reduces sensing energy by generating unobserved content, consistent with the energy model $P_{\text{total}} = P_{\text{laser}} + P_{\text{scan}} + P_{\text{signal}} + P_{\text{control}}$ and $P_{\text{laser}} = \frac{E_{\text{pulse}} f_{\text{pulse}}}{\eta_{\text{laser}}}$, where $E_{\text{pulse}}$ scales approximately as $E_{\text{pulse}} \propto R^4$. Findings: across Waymo, nuScenes, and KITTI, R-MAE yields over 5% improvements in mAP and over 4% gains in cross-domain transfer, with up to 4.37% gains for small objects on KITTI and robustness to 90% radial masking on Waymo. Significance: demonstrates practical ultra-efficient LiDAR perception enabling edge autonomy by trading sensing energy with training data.
Abstract
In this work, we propose a disruptively frugal LiDAR perception dataflow that generates rather than senses parts of the environment that are either predictable based on the extensive training of the environment or have limited consequence to the overall prediction accuracy. Therefore, the proposed methodology trades off sensing energy with training data for low-power robotics and autonomous navigation to operate frugally with sensors, extending their lifetime on a single battery charge. Our proposed generative pre-training strategy for this purpose, called as radially masked autoencoding (R-MAE), can also be readily implemented in a typical LiDAR system by selectively activating and controlling the laser power for randomly generated angular regions during on-field operations. Our extensive evaluations show that pre-training with R-MAE enables focusing on the radial segments of the data, thereby capturing spatial relationships and distances between objects more effectively than conventional procedures. Therefore, the proposed methodology not only reduces sensing energy but also improves prediction accuracy. For example, our extensive evaluations on Waymo, nuScenes, and KITTI datasets show that the approach achieves over a 5% average precision improvement in detection tasks across datasets and over a 4% accuracy improvement in transferring domains from Waymo and nuScenes to KITTI. In 3D object detection, it enhances small object detection by up to 4.37% in AP at moderate difficulty levels in the KITTI dataset. Even with 90% radial masking, it surpasses baseline models by up to 5.59% in mAP/mAPH across all object classes in the Waymo dataset. Additionally, our method achieves up to 3.17% and 2.31% improvements in mAP and NDS, respectively, on the nuScenes dataset, demonstrating its effectiveness with both single and fused LiDAR-camera modalities. https://github.com/sinatayebati/Radial_MAE.