3D adaptive coverage path planning for autonomous submarine tank inspection robots

Author: Rowan Pivetta

Pivetta, Rowan, 2020 3D adaptive coverage path planning for autonomous submarine tank inspection robots, Flinders University, College of Science and Engineering

Terms of Use: This electronic version is (or will be) made publicly available by Flinders University in accordance with its open access policy for student theses. Copyright in this thesis remains with the author. You may use this material for uses permitted under the Copyright Act 1968. If you are the owner of any included third party copyright material and/or you believe that any material has been made available without permission of the copyright owner please contact copyright@flinders.edu.au with the details.


The manual inspection of the confined spaces within the Australian Collins Class submarine tanks for evidence of corrosion, paint delamination and potential defects, is a hazardous, time consuming, and expensive process. Automating submarine tank inspections will eliminate the prolonged exposure of human inspectors to the hazardous confined spaces of the submarine. The robotic platform suitable for this task is a six-legged robot that possesses electromagnetic adhesion giving it the ability to climb freely throughout the complex steel tank structures of a submarine.

Coverage path planning algorithms can be used to generate inspection plans, however, in complex environments, this a challenging problem. Coverage planning algorithms can be generated offline prior to inspection, however, these plans are unable to adapt to changes in the environment. An adaptive coverage planner that enables the robot to navigate around detected obstacles whilst providing sensory coverage of the newly detected features is preferable. An adaptive coverage planner is developed in this thesis to fulfil this requirement and enable an autonomous platform to perform a comprehensive inspection of submarine tanks.

This thesis extends the capability of an existing offline sampling-based coverage planner, known for generating discrete coverage plans in complex environments, with online path replanning strategies to perform adaptively during execution. Two replanning strategies were explored, a full replan and plan repair, neither of which have previously been applied to adaptively update a current inspection plan to changing conditions using the offline coverage planner.

An examination of the offline sampling-based coverage planner within the representative submarine tank scenario was used to determine its effectiveness for online performance. Key results showed that the offline coverage planner was susceptible to significant variable planning times for large complex planning problems and was sensitive to minor variations in the environment. New methods derived from the analysis were developed to resolve the variability of planning times and sensitivity to the environment, consequently led to the reduction of planning times from 6 hours to under 10 minutes. The results of these improvement indicated that a plan repair strategy was best to adapt the offline coverage planner to the online domain.

Experiments comparing the two replanning strategies revealed that an adaptive sampling-based coverage planner using a plan repair strategy was the faster approach compared to a full replan strategy. The plan repair strategy demonstrated its capability of replanning up to 92% of the existing tour without significant degradation, completing the replanning problems for the representative submarine tank environment 396 times faster, going from 46 hours to 7.5 minutes.

Keywords: 3D coverage path planning, inspection robotics, adaptive coverage planning, path planning, submarnie inspection, autonomous inspection, replanning, plan repair, skeletonisation, sampling-based coverage planner, offline coverage path planning, online coverage path planning, multi-goal planning problem, multi-legged platforms

Subject: Engineering thesis

Thesis type: Doctor of Philosophy
Completed: 2020
School: College of Science and Engineering
Supervisor: Karl Sammut