Review on Multisensor SLAM Datasets for Advanced Perception and Mapping Technologies

Abstract

Abstract. SLAM (Simultaneous Localization and Mapping) is a technique in robotics and computer vision used to build a map of an unknown environment while simultaneously tracking the location of a robot or vehicle within that environment. The primary goal is to enable autonomous systems to navigate and understand their surroundings without prior knowledge of the environment. It has evolved significantly with the integration of diverse sensor modalities which initially used either a single LIDAR (light detection and ranging, or laser imaging, detection, and ranging) or visual sensor to perform the dual tasks of mapping an environment and localizing the device within it. These systems had limitations in accuracy and robustness due to their reliance on a single type of data input. Over time, the field has advanced to incorporate multiple sensor modalities, including LIDAR, visual cameras, Inertial Measurement Units (IMUs), ultrasonic sensors, and GPS. This multi-sensor fusion approach has dramatically enhanced the precision and reliability of SLAM systems. This paper reviews the state-of-the-art datasets that combine data from infrared cameras, depth cameras, LiDAR, and 4D millimeter-wave radar, focusing on their contributions to advancing SLAM technologies. The study analyzes the advantages and limitations of each sensor type, the challenges associated with data fusion, and the impact on perception and mapping accuracy. This review aims to provide a comprehensive understanding of how these multisensor datasets enhance SLAM systems and highlight areas for future research.