Efficient Relative Coordinate Inference for Dynamic SLAM Exploiting Monocular Cameras

In the development of autonomous driving robots and related technologies, the successful implementation of SLAM (Simultaneous Localization and Mapping) is crucial. To this end, it is very important to estimate the current position and, orientation of the robot while efficiently constructing a map of the environment. Various algorithms have been proposed so far, utilizing LiDAR, graph-based methods, and inertial systems. However, these methods share common issues, such as high computational and resource costs for map construction, as well as limitations when operating in dynamic environments with numerous moving objects, such as inside large shopping mall. To address these challenges, we propose an algorithm for relative coordinate inference using distance measurement/trilateration, which is applicable even with monocular cameras.