University of Twente Student Theses


A spatial cognitive exploration algorithm for autonomous mapping of unknown indoor environments

Hari, Atul (2020) A spatial cognitive exploration algorithm for autonomous mapping of unknown indoor environments.

[img] PDF
Abstract:Autonomous exploration has gained remarkable attention among robotics researchers because of its application in automating various tasks such as search and rescue, mapping of underground tunnels, and space exploration. Similarly, exploration can be used to automate the process of indoor mapping, which includes three-dimensional reconstruction of the houses and virtual reality imaging of each room. However, traditional exploration algorithms based on 2D LIDAR do not explicitly determine locations within an environment that is suitable for scanning the surrounding. Inspired by humans, A novel algorithm capable of modeling spatial relation of points in a twodimensional plane to its boundaries is developed in this research. This design models the boundaries by a polygon map approximated from the two-dimensional point cloud created by a laser range scanner. Then it relates the polygon to the points inside it, to develop a twodimensional function representing the visibility of the environment. To develop this algorithm, a Gaussian process model is actively trained with the spatial relation between the polygon map and the points within. The trained model is then used to detect the points to visit while performing mapping. To develop an exploration framework around point detection, the research proposes two approaches. The first hypothesis, a novel exploration strategy by remembering the past visited regions, and the second adapts a traditional approach of exploration by locating the unknown regions. Further, for optimizing the sequence of visiting these points, a traveling salesman framework, is used along with an ant colony optimization algorithm. On verification, it was seen that the algorithm was capable of predicting the points to visit in an unknown indoor environment with an average prediction error of 1m, using both the proposed methods. Further, on comparing the algorithm with traditional approaches, an equivalent performance was observed by the proposed exploration framework that maintains a memory. However, both the algorithms were capable of exploring 90% of the area on an average in the five test environment.
Item Type:Essay (Master)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:31 mathematics, 54 computer science
Programme:Systems and Control MSc (60359)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page