| Author: | Eldemiry, Amr Mostafa Osman Yasin |
| Title: | Exploration strategies for high-quality indoor mapping using autonomous mobile robots (AMR) |
| Advisors: | Chen, Wu (LSGI) Wen, Chih-yung (ME) |
| Degree: | Ph.D. |
| Year: | 2023 |
| Subject: | Mobile robots Autonomous robots Hong Kong Polytechnic University -- Dissertations |
| Department: | Department of Land Surveying and Geo-Informatics |
| Pages: | xvii, 116 pages : color illustrations |
| Language: | English |
| Abstract: | Indoor mapping using autonomous mobile robots (AMR) can be employed in several applications, such as rescue scenarios, utility tunnel monitoring, and indoor 3D modelling. These AMRs rely on simultaneous localization and mapping (SLAM) systems for self-localization and scene perception. In addition, an autonomous exploration strategy is essential to guide the AMR in covering the unknown environment. Full coverage with minimal exploration time is the main objective that attracted many researchers to propose autonomous exploration strategies. However, the state-of-the-art exploration strategies do not consider the mapping quality using low-cost sensors in challenging indoor conditions. RGB-D SLAM system is attractive for its low-cost, low-computational cost, low power consumption and dense real-time colour reconstruction ability. However, RGB-D SLAM constructs low-quality maps, especially in low-texture environments. Therefore, this study proposed novel exploration strategies to fill that gap by optimizing mapping quality and exploration time in low-texture environments using an RGB-D camera. Subsequently, our novel exploration strategies consider the mapping quality factors of the RGB-D SLAM system. Furthermore, path planners can also assist the RGB-D SLAM system by generating trajectories that increase the ability to mitigate the RGB-D SLAM drift and safety in narrow, complex, and cluttered real-world environments. Therefore, this study proposed a Voronoi-based path planner that increases the number of detectable loop closures to mitigate the RGB-D SLAM drift. In addition, for safety considerations, these trajectories offer the maximum clearance for obstacle avoidance. Three exploration strategies are evaluated. The first is the baseline frontier-based exploration (D-strategy), which depends on the Euclidean distances between the robot and the next goal candidates. The second is the proposed RGB-D mapping-based exploration (M-strategy), which considers the mapping quality factors of the RGB-D SLAM system. Finally, the proposed hybrid exploration (M+D-strategy) combines the first two strategies. According to the texture level of the environments (low, moderate, and rich), these three exploration strategies are evaluated in three real-world environments. A significant enhancement is achieved in mapping quality and exploration time using our proposed exploration methods compared to the baseline frontier-based exploration, particularly in a low-texture environment. The proposed novel exploration strategies significantly enhanced mapping quality and exploration time in the low-texture environment. The results show that when using feature-based SLAM, it is beneficial to consider the quantity and distribution of features when selecting the next new goal. The enhancement percentages in point-to-point distances (PTPDs) RMSE (between the generated models and the ground truth model) are more than 40 % for both proposed strategies (M and M+D). Furthermore, total path length and exploration time are enhanced by more than 28 % and 33 %, respectively, for the M-strategy and more than 21 % and 23 % for the hybrid M+D strategy. In this environment, the hybrid strategy was optimizing mapping quality and exploration time, so the robot took more time than the M-strategy to make decisions as the motion cost of each goal is not constant during the exploration, which always changes the goal scores. In the moderate-texture environment, the enhancement percentage in PTPDs RMSE is more than 15 % for the proposed M-strategy, while for the hybrid strategy (M+D-strategy) is more than 13%. On the other hand, total path length and exploration time increased by 29 % and 22%, respectively, for the M-strategy and increased by 11 % and 6%, respectively, for the hybrid M+D strategy. In this environment, the hybrid strategy took more time than the M-strategy to make decisions for the same reason in the low-texture environment, as the goal scores during the exploration are dramatically changing. In the texture-rich environment, the enhancement percentage in PTPDs RMSE is more than 2 % for the proposed M-strategy, while for the hybrid strategy (M+D) is more than 9%. In addition, total path length and exploration time are decreased by more than 26% for the M-strategy and more than 34 % for the hybrid M+D strategy. In this rich-texture environment, the RGB-D SLAM mapping quality is always high due to the sufficient features in all regions used for alignment, regardless of the exploration method. Therefore, the mapping enhancements in this environment were insignificant compared to the low-texture environment. |
| Rights: | All rights reserved |
| Access: | open access |
Copyright Undertaking
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item:
https://theses.lib.polyu.edu.hk/handle/200/12440