Modeling occupancy grids using EDHMM for dynamic environments

Abstract

Map generation of mobile robots over long periods of working suffers from inconsistencies because of gradually changes in the environment. These changes further cause hindrances in autonomous navigation of mobile robots. This paper presents a novel method to infer such gradual changes and incorporate them in map generation. We model the environment using an occupancy grid structure with the state of each grid cell is determined in an online fashion using an Explicit-state-Duration Hidden Markov Model (EDHMM). Our work presents filtering of occupancy grid into dynamic and static. We tested our method in simulation as well as on a real world dataset. Our results show robust detection of dynamic changes in the grid map, even in the presence of occlusion.