AUTONOMOUS EXPLORATION AND ACTIVE MAPPING WITH MOBILE ROBOT

• Main Author: Agung Andika Perkasa, Dionesius
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/51260
• Institution: Institut Teknologi Bandung

Map or knowledge of the environment is a knowledge that a robot must possess. It is absolute for a robot to know the environment in order to operate autonomously. Before taking various roles such as logistical operations, search and rescue missions, and many others, a robot must first be familiar with the map of its environment. Because the knowledge of the map is so important, the ability to map its environment becomes a crucial and fundamental problem in robotics. The development and research on map learning have therefore advanced quite fast in recent decades. There are two approaches in learning and constructing map. The first approach is called passive approach. This approach only utilizes current observation of the environment to build the map. Passive approach needs external commands to move and guide the robot so it can build the complete map of the environment. Active approach, in addition to observing the environment, actively plans the robot’s trajectory. This approach enables the robot to obtain an accurate model of its environment autonomously. A robot with this capability is considered to have met one key requirement for being a fully autonomous robot. The contribution of this thesis is on the development of an active approach to map learning. The proposed active approached is based on the frontier exploration technique. The proposed exploration strategy uses theories from the field of information theory. This technique uses a heuristic to calculate information gain that is based on Kullback-Leibler Divergence. Experiment results show that the proposed strategy was able to increase the efficiency of active mapping tasks in terms of time needed to complete and distance traveled by the robot