Towards increased autonomy of a rail-guided robot for the inspection of ballast water tanks

Trouwborst, C. (2017) Towards increased autonomy of a rail-guided robot for the inspection of ballast water tanks.

Abstract:Ballast Water Tanks (BWTs) of large ships are required by law to be regularly inspected for corrosion, to ensure a sufficiently strong mechanical structure of the hull. Recognizing the health hazards for inspectors of BWTs, a rail-guided inspection platform was proposed by Christensen et al. in the Robots in Tanks project. As part of the SmartBot project, this platform was further developed into RoboShip platform. Since reliable connectivity to the platform cannot be assumed in steel BWTs a certain level of autonomy is required, which is currently not present in the platform. Given that inspection tasks are not fully predictable, but dependent on knowledge gained on the heavily obstructed, partially known environment, introducing full autonomy is not realistic. Aiming for a blend of mainly executive control and aspects of shared control with robot initiative is most promising for the inspection field, since it allows for both a large range of robot activities and explicitly tapping into the human operator’s experience. To answer the question that results from these considerations, How to increase the autonomy of the RoboShip platformto the level of (blended) executive control, given the fundamental task- and environment uncertainty of the inspection domain? key challenges in increasing the autonomy of the RoboShip inspection platform were defined in three categories: challenging aspects in the environment, challenges stemming from the inspection task and challenges stemming from the status of the RoboShip platform itself. All challenges were translated to specific requirements for the control architecture or the system as a whole. Based on the defined requirements a design for the control architecturewas proposed, inwhich a discrete planner is combined with a number of continuous control components. This thesis shows that discrete planning systems can be very effective in introducing autonomy to an inspection robot. By altering the existing TFD/Mplanner to allow for changes in the goal state requirements as a result of elementary sensing actions, new information on the environment can successfully be taken into account. Additionally, it has been shown that by using a discrete planning system that supports modules, the semantic gap can be bridged and relevant details on the continuous domain can be incorporated into the highest, most abstract layer of planning and control. While not a component of autonomy, significant efforts have been made in localization and armcontrol, to advance the state of these two important requirements. The simulator created as part of this research proved to be a very useful tool in researching and validating abstract, high-level control logic. We believe that the proposed architecture and accompanying tools forma solid foundation for robustly introducing autonomous features to the RoboShip platform.
Item Type:Essay (Master)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:52 mechanical engineering, 53 electrotechnology, 54 computer science
Programme:Systems and Control MSc (60359)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page