Semantic Mapping in ROS
Year: 2013
Type: dissertation
Abstract: In the last few years robots are becoming more popular in our daily lives. We can see them guiding people in museums, helping surgeons in hospitals and autonomously cleaning houses. With the aim of enabling robots to cooperate with humans and to perform human-like tasks we need to provide them with the capability of understanding human environments and representing the extracted knowledge in such a way that humans can interpret. Semantic mapping can be defined as the process of building a representation of the environment, incorporating semantic knowledge obtained from sensory information. Semantic properties can be extracted from various sources such as objects, topology of the environment, size and shape of rooms and room appearance. This thesis proposes an implementation of semantic mapping for mobile robots which is integrated in a framework called Robot Operating System (ROS). The system extracts spatial properties like rooms, objects and topological information and combines them with commonsense knowledge into a probabilistic framework which is capable of inferring room categories. The system is tested in simulations and in real-world scenarios and the results show how the system explores an unknown environment, creates an accurate map, detects objects, infers room categories and represents the results in a map where each room is labelled according to its functionality.
Author: Xavier Gallart Del Burgo
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W130147540
