Daun, Kevin (2024)
Localization, Mapping and Exploration with Mobile Ground Robots in Disaster Environments.
Technische Universität Darmstadt
doi: 10.26083/tuprints-00028912
Ph.D. Thesis, Primary publication, Publisher's Version
Text
phdthesis_daun.pdf Copyright Information: CC BY 4.0 International - Creative Commons, Attribution. Download (10MB) |
Item Type: | Ph.D. Thesis | ||||
---|---|---|---|---|---|
Type of entry: | Primary publication | ||||
Title: | Localization, Mapping and Exploration with Mobile Ground Robots in Disaster Environments | ||||
Language: | English | ||||
Referees: | Stryk, Prof. Dr. Oskar von ; Nüchter, Prof. Dr. Andreas | ||||
Date: | 17 December 2024 | ||||
Place of Publication: | Darmstadt | ||||
Collation: | xv, 95 Seiten | ||||
Date of oral examination: | 8 April 2024 | ||||
DOI: | 10.26083/tuprints-00028912 | ||||
Abstract: | Responding to disasters and threatening situations is a major challenge for first responders, authorities, and the public. The use of rescue and response robots can help to overcome the challenges by improving overall response capabilities, e.g., by providing valuable insights on dangerous areas through gathering data and creating 3D maps of the environment or performing remote physical actions while enabling first responders to maintain a safe distance from potential dangers. However, the operating conditions for robots in disasters and threatening situations such as fire, flooding, collapse, or CBRNE are very difficult. Environmental conditions are usually very harsh, with challenging ground characteristics, versatile and irregular obstacles, and potentially disturbed visual conditions due to smoke, fog, and dust. Moreover, compared to industrial robot applications, missions, and environments have large variations and low repeatability and offer little prior knowledge and lead time, making applying common methods and approaches from mobile robot autonomy and artificial intelligence (AI) particularly challenging. At the same time, these conditions pose a challenge for remote teleoperation, as they increase the likelihood of fatal errors and mission failures for human operators. This thesis focuses on developing mobile rescue robots with assistance functions motivated by advancing disaster response efficiency and safety, e.g., by contributing to autonomous robot exploration, that account for the specific requirements and challenges to support first responders and civil forces. Therefore, this work presents specific approaches for localization, mapping, and exploration with mobile ground robots in disaster response, addressing crucial challenges in three distinct areas. Firstly, understanding the full range of specific requirements for (autonomous) assistance functions in rescue robots is crucial for research and development towards practical applicability. Previous analyses have primarily focused on general aspects, leaving a gap in the specific understanding of requirements for (autonomous) assistance abilities. We address this gap by deriving a novel model for an integrated function capability from established models for technology acceptance and derive a comprehensive, evidence-driven analysis of application requirements and research challenges for (autonomous) assistance abilities. Secondly, sufficiently accurate and robust simultaneous localization and mapping (SLAM) in unknown environments without relying on GNSS support are essential for (semi-)autonomous operation. In particular, traversing uneven ground can lead to abrupt robot motions that existing SLAM methods cannot model accurately or efficiently enough. Furthermore, relevant environments are often unstructured and potentially visually degraded by smoke, dust, or fog. Therefore, we investigate new methods for robustly registering lidar scans, accurately estimating the trajectory in rough terrain, and efficiently mapping large-scale environments online on a mobile rescue robot system. The proposed approach gains accuracy and robustness by registering lidar data in a multi-resolution Truncated Signed Distance Function (TSDF) with a continuous-time trajectory representation. It enables the efficient mapping of large-scale environments by transferring a branch-and-bound-based loop closure detection approach for TSDF. Furthermore, we investigate extensions of the approach for the operation in visually degraded conditions with radar. Thirdly, in response missions, robots might need to fulfill various tasks in a single mission. In such dynamic and versatile environments, first responders often have prior knowledge and better high-level decision-making skills than AI methods for the perception and reasoning of autonomous mobile robots. However, an operator’s cognitive load is limited, and direct operator control is potentially error-prone, often inefficient, and not always possible. Therefore, we investigate a new, efficient, and flexible method for multi-goal exploration that combines AI methods for perception with operator capabilities by extending a hierarchical planning approach for multi-goal scenarios and facilitating flexible operator assistance with an actionable environment representation based on affordances. The innovations, methods, and implementations presented in this work have been successfully evaluated in various complex simulated and real-world robot experiments, demonstrating accuracy, robustness, and efficiency. Parts of the real-world evaluation are performed under the conditions of various international robotics competitions (RoboCup Rescue Robot League, EnRicH, World Robot Summit), demonstrating better accuracy and robustness than related approaches. In addition, the results from this thesis were used for their application in real missions and as input for two German consortium standards (DIN SPEC), which underline their impact in the field of disaster robotics. |
||||
Alternative Abstract: |
|
||||
Status: | Publisher's Version | ||||
URN: | urn:nbn:de:tuda-tuprints-289121 | ||||
Classification DDC: | 000 Generalities, computers, information > 004 Computer science | ||||
Divisions: | 20 Department of Computer Science > Simulation, Systems Optimization and Robotics Group | ||||
Date Deposited: | 17 Dec 2024 10:26 | ||||
Last Modified: | 19 Dec 2024 08:39 | ||||
URI: | https://tuprints.ulb.tu-darmstadt.de/id/eprint/28912 | ||||
PPN: | 524705569 | ||||
Export: |
View Item |