Long term applications of autonomous robot navigation consist of repeated traversals in real world. Performing navigation in an unknown environment, requires robot to solve Simultaneous Localization and Mapping(SLAM) problem. Traditional SLAM solver assumes static world where no object around mobile robot changes its place. However the real world has people moving in vicinity of mobile robot and also there are furniture items such as chair, table etc. that might be displaced in long term applications. In order for robust navigation, it is crucial to estimate the places occupied by dynamic objects around a mobile robot.
This thesis targets to estimate the position of objects such as table or chair which are displaced in subsequent traversals of robot and are termed as low-dynamic objects. A method for filtering of positions occupied by low-dynamic object from all the positions occupied by objects around a mobile robot is presented in this thesis. The main idea is to use duration for which an object is observed at a place to estimate the low-dynamic or static state. Environment around the robot is represented using an occupancy grid map and each grid cell in the map is modelled as a semi-Markov chain to infer the state of cell. In this thesis, the proposed method is tested on simulation environment as well as in real world environment. The results show that the proposed method robustly tracks positions of low- dynamic objects in the environment for weeks of observations even in the presence of noise or occlusion.