Active SLAM Based on Geometry Rules and Forward Simulation in Exploration Space

Abstract

Active simultaneous localization and mapping (Active SLAM) is a key problem of robot autonomous exploration in unknown environment. The most popular algorithm is the frontier-based algorithm. However, it is usually time-consuming when used in a large structured environment due to its computation complexity. This paper proposes a forward simulation method based on geometry rules for the active SLAM (G-FS). Utilizing geometry rules in exploration space, a Sequential Monte Carlo (SMC) based method is used to select the next goal. This proposed method can increase the length of a single motion step, and solve the local jammed problem in the forward simulation, so as to increase the efficiency. Comparison experiments are performed among the G-FS method, forward simulation method and frontier-based method. The results demonstrate that the efficiency of the active SLAM is greatly improved by using the proposed G-FS algorithm.