Options
Autonomous navigation of quadruped robots for monitoring and inspection of civil infrastructure
Publikationstyp
Conference Paper
Date Issued
2004-08
Sprache
English
Citation
Proceedings of the 20th International Conference on Computing in Civil and Building Engineering (ICCCBE). Montréal, Canada, 08/25/2024
Contribution to Conference
International Conference on Computing in Civil and Building Engineering, ICCCBE 2024
Mobile robots have increasingly been gaining recognition in monitoring and inspection of civil infrastructure, owing to their potential to improve efficiency and accuracy. Specifically, quadruped robots offer enhanced stability and adaptability, rendering the robots ideal candidates for automated monitoring and inspection. To enable quadruped robots to conduct monitoring and inspection of civil infrastructure, the robots must be capable of autonomous navigation. Existing approaches towards autonomous navigation usually incorporate joint-state information to plan motions with whole-body controllers, representing a non-linear, high-dimensional problem that is computationally expensive to solve and is a particular burden when implemented into the robots for real-time navigation. Addressing the challenge of real-time navigation, this paper presents a robust motion-planning framework, tailored to autonomous operation of quadruped robots in complex and dynamic environments, commonly encountered when monitoring and inspecting civil infrastructure. The motion-planning problem is decoupled into high-level task-space and low-level joint-space components, and the framework, building upon the “Cartographer” simultaneous localization and mapping algorithm, combines the “batch-informed trees” algorithm for global planning and the “timed elastic band” for local planning and obstacle avoidance. For validation, the framework is implemented into quadruped robots, employed for conducting inspection tasks, and validation tests are conducted in indoor office environments to be inspected. As a result, it is demonstrated that the quadruped robots navigate safely and collision-free in real time, accommodating both static and dynamic obstacles. Enhancing the efficiency and accuracy of autonomous navigation of quadruped robots in complex and dynamic environments, the framework is expected to pave the way for future research in robust controller development and 3D-state space planning.