Abstract: With the rapid development of autonomous driving technologies, including sensors, data processing, visual perception, vehicle control, decision making, and artificial intelligence, achieving fully automated Level-5 autonomous driving has become a critical direction for the future of the automotive industry. Object detection lies at the core of autonomous driving systems. It analyzes surrounding-environment data obtained from onboard sensors so as to determine the position, category, size, and posture of objects in the road environment. Object information provides critical support for localization, navigation, path planning, and motion control. Of the many environmental perception sensors, light detection and ranging (LiDAR), which provides high-precision three-dimensional point cloud data, plays a vital role in environmental perception of and object detection by autonomous driving systems. Advanced LiDAR-based object detection methods leverage deep learning techniques to extract object features, predict object locations, and classify objects within an end-to-end neural network framework, achieving remarkable accuracy in many general scenarios. However, adverse weather conditions, such as rain, fog, and snow, notably affect the quality and reliability of LiDAR point cloud data. These conditions introduce challenges, such as increased noise in point clouds, occlusion of foreground objects, weakening of point intensity, reduction in sensor confidence, and difficulties in data acquisition, all of which degrade the performance of general object detection methods. Consequently, addressing the robustness of LiDAR-based object detection under adverse weather conditions has become a prominent research focus. Aiming at overcoming these challenges, various methods have been proposed to enhance object detection performance, including strategies that enhance the robustness of object detection networks under complex weather environments. This study comprehensively reviews state-of-the-art object detection methods based on LiDAR point clouds under adverse weather conditions. To establish the research background, we first summarize the basic concepts, tasks, methods, and evaluation metrics of point cloud object detection while analyzing the specific effects of adverse weather on detection performance. Following this, we systematically categorize the existing object detection methods into four main approaches: (1) data enhancement-based ones, which enhance network robustness via augmentation of training datasets to account for weather variations; (2) point cloud denoising-based ones, which mitigate the effect of weather-induced noise via cleaning of raw point cloud data, thereby increasing detection accuracy; (3) domain adaptation-based ones, which use models trained under favorable weather conditions to guide the training of networks designed for adverse weather so as to enhance the model’s generalization ability; and (4) multisensor fusion-based ones, which enhance detection performance via integration of multimodal data from various sources, such as combining LiDAR with camera or radar. Each of these methods is reviewed in detail, with their strengths and limitations analyzed. Furthermore, this study speculates that future research will continue to adopt deep learning methods and innovate in four aspects: data, models, learning strategies, and practical applications. By analyzing in detail the current status of research and identifying the pressing challenges, this review aims to offer valuable insights and inspiration for subsequent studies, ultimately facilitating the development of robust object detection technologies under adverse weather conditions.