Vachhani, Birju. Autonomous mobile robot navigation with GPU acceleration for unmanned UV-C based decontamination applications. Retrieved from https://doi.org/doi:10.7282/t3-hc0y-xj46
DescriptionNavigation is a complex robotic problem solving which makes the mobile robot intelligent for decision making in dynamic environments. The objective of this thesis is to achieve autonomous mapping and navigation of a 2D environment for unmanned decontamination and sterilization of rooms in facilities such as hospitals and hotels, where UV-based degermination system is carried by a four-wheeled robot during the navigation. The challenge arises from the intensive online computation needed in the navigation process, as a Light Detection and Ranging (LiDAR) system is employed for Simultaneous Localization and Mapping (SLAM) and obstacle detection, and occupancy grid is employed as the data structure to represent surrounding environments in robotics. However, path planning on an occupancy grid is computationally intensive. For example, path planning on a 25m x 25m grid can involve processing of 250,000 grid cells on a 0.05m resolution grid. In this project, we developed a parallel computation framework to substantially reduce the processing times, and thereby, achieve dynamic obstacle avoidance and autonomous exploration. Specifically, a obstacle inflation module was created using parallel computing. Building on that, a graph based path planner was also developed for autonomous as well as user instructed navigation. The system was thoroughly tested in various static and dynamic indoor environments.