The Multimodal fusion and 3D perception have become pivotal in modern deep multi-object tracking (MOT). This review summarizes a large number of recent works that integrate modalities such as RGB camera, LiDAR, depth sensors, thermal imaging, and language into MOT pipelines and exploit 3D representations (point clouds, bird’s eye view, stereo, UAV and satellite imagery). Developments are synthesized across several categories: tracking-by-detection methods that extend detectors with multimodal or 3D inputs, SOT-based approaches that adapt single-object trackers to multi-object scenarios (especially in point clouds and hierarchical RGB-X domains), and joint detection-and-tracking frameworks embedding 3D reasoning or cross-modal fusion. A dedicated section surveys purely multimodal and 3D MOT techniques, including LiDAR–camera fusion, RGB–depth/thermal/language integration, BEV-based methods, aerial and satellite tracking, and cross-camera or cooperative systems. Additionally, this review also examines key datasets and metrics (e.g. KITTI, Waymo, UAVDT, DIVOTrack, SatSOT) and discuss challenges like occlusion, sensor misalignment, and real-time processing, as well as future research directions.