Efficient Sparse LiDAR Data Reconstruction using Compressed Sensing for Real-Time Perception in Autonomous Vehicles

Abstract

This paper presents a novel approach to real-time perception for autonomous vehicles, introducing a LiDAR data reconstruction method utilizing Curvelet transform-based compressed sensing. The study substantiates the heightened efficacy of the Curvelet transform in adeptly managing complex geometric structures, yielding significant improvements in reconstruction quality and processing time compared to existing methodologies. Employing specialized algorithms tailored for reconstruction enhances sparse LiDAR data, thereby improving accuracy and detail in 3D mappings. Through systematic simulations, we validate the efficiency of the Curvelet transform, rendering the LiDAR system more proficient and faster at decision-making. This research contributes to the optimization of LiDAR data processing, representing a pivotal step towards the development of advanced real-time perception systems with potential applications in fully autonomous navigation. The transformative impact of Curvelet-based compressed sensing underscores its critical role in advancing the field of autonomous vehicle technology.