I. Introduction
Robotics environment modeling aims to generate a proper representation of 3D space, based on which the autonomous navigation and manipulation can be performed. In this letter, we consider the occupancy mapping and terrain modeling for large-scale environments, corresponding to the structured and unstructured scenarios, respectively. Such environment modeling, common in practice, generally suffers from high computation and storage consumption due to the vast amount of captured data. From the practical perspective, these problems have been the major hinders for large-scale environment modeling.