Robust Autonomous Navigation of the Mobile Robot `ROSWITHA` by Sophisticated Sensing and Bio-Inspired Algorithms
- Autores: Sudeep Sharan
- Directores de la Tesis: Juan José Domínguez Jiménez (dir. tes.)
, Peter Nauth (codir. tes.) - Lectura: En la Universidad de Cádiz ( España ) en 2026
- Idioma: español
- Tribunal Calificador de la Tesis: Andreas Hermann Pech (presid.) , Francisco Palomo Lozano (secret.) , Pavika Sharma (voc.)
- Enlaces
- Resumen
Navigation is an important and challenging robotic task. It becomes more effective when applied to different mobile systems like assistive service robots, autonomous automobiles, aerial systems, and autonomous underwater vehicles. Despite the fact that large amounts of research have been continually done for some decades and in spite of advances in these themes, the problem of moving robotic systems in complex and dynamic environments in practical scenarios is still not solved. Therefore, the key problem of robot navigation is to have enough intelligence to navigate in a complex and dynamic environment. Robotic TESIS: Hospital Real Plaza Falla, 8 | 11003 Cádiz Tel. 956 015353 http://www.uca.es posgrado@uca.es Escuelas Doctorales researchers developed and researched different technologies and algorithms for navigation.
To understand the recent improvement of the methodologies and algorithms for robot motion, the research conducted in this PhD thesis first provides a comprehensive literature review detailing existing technologies in navigation, localization, path planning, and obstacle tracking.
Building upon this foundation, the thesis then details the design and implementation of a robust navigation system, which involved both critical hardware and software integration. This comprehensive system leverages diverse sensor data (LiDAR and vision) for improved perception and accurate Simultaneous Localization and Mapping (SLAM). The system integrates advanced path planning techniques, including optimized Ant Colony Optimization (ACO) and Dijkstra algorithm, enhanced with Bezier curves for smoother trajectories. It incorporates deep learning-based semantic segmentation for precise obstacle detection and classification, enabling differentiation between static and dynamic objects.
The developed system was rigorously evaluated in real-world scenarios, demonstrating exceptional capabilities in localization, mapping, and adaptive navigation. Experiments confirmed the system’s ability to maintain precise and robust navigation while effectively avoiding both static and dynamic obstacles, achieving a high success rate.
This thesis provides a comprehensive and validated framework for intelligent autonomous navigation, significantly enhancing the real-world embodied system ROSWITHA’s capabilities for safe and reliable operation in human-centric environments.
