LSMCL: Long-term Static Mapping and Cloning Localization for autonomous robot navigation using 3D LiDAR in dynamic environments

Abstract

One of the challenges in autonomous robot navigation applications is recognizing the exact location in a dynamic environment in which the location of surrounding objects changes frequently. This study proposes a long-term static mapping and cloning localization (LSMCL) method for estimating real-time accurate location using only natural landmarks even in a dynamic environment using a 3D LiDAR sensor. LSMCL comprises longterm static mapping (LSM) and cloning localization (CL). A LSM creates a 2D grid map and a 3D geometric feature map for objects whose positions do not change in space. A CL uses the generated 2D grid map and particle filter to estimate the 2D global position at the initial stage. After cloning the 2D global location to 3D space, the location is tracked through a 3D feature map and map matching. To verify the LSMCL's usability in a real dynamic environment, a robot navigation experiment was conducted in a highly dynamic parking lot. The experimental results, analyzed in terms of initial localization success rate, location estimation accuracy and precision, processing time, and congested space application confirmed the LSMCL's real-world applicability.