Key points are not available for this paper at this time.
In recent years, the rapid evolution of artificial intelligence has propelled autonomous driving (AD) to the forefront of technological applications in the transportation sector, drawing significant research attention. Perception tasks have received extensive attention due to the strides in sensing technologies like cameras and Light Detection and Ranging (LiDAR). These advancements have significantly improved the accuracy and robustness of perception systems. However, the inherent limitations of single-modal sensors have necessitated a shift towards multi-modal sensor fusion. This approach leverages the unique advantages of diverse sensor systems integrated into autonomous vehicles (AVs), significantly improving perception accuracy. This paper reviews recent developments in multi-sensor fusion for AD, emphasizing its role in overcoming the limitations of single-modal sensors. It examines major open-source datasets, evaluates single-sensor perception performance, and categorizes state-of-the-art fusion algorithms from both methodological and task-oriented perspectives. The review identifies current research gaps and proposes future directions to enhance AD perception systems.
Jiang et al. (Fri,) studied this question.