Autonomous Model Reconstruction of Marine Structures

Our goal is to develop a system that can autonomously construct 3D model of marine structures, i.e., structures that are partially submerged, such as offshore oil-rigs, sea-caves, and harbours. This project includes development of a robotic system that can autonomously scan the entire part of marine structures, development of software system to construct 3D model of the structures from the scan data, and development of algorithms and software that can autonomously and efficiently scan the structures of interest.

Our work is motivated by two applications: structural inspection and scientific exploration. Structural inspection is a tedious task where even a small oversight may have severe consequences. This is accentuated further for structural inspection of marine structures, as inspectors need to inspect them in uncomfortable positions aboard boats or with diving equipment. To reduce the chances of oversight, we would like to send robots, instead of humans, to autonomously scan the structures and then construct 3D model of the structures. The inspectors can then inspect the model of the structures from the comfort of their office and perform in-field inspection only on a few small parts of the structures that are suspected to be critical. Scientists who are interested in understanding the geological structure of a sea-cave, for instance, can use our system to get a rough idea about which part of the cave would provide the most information, and can focus on these more useful parts of the cave when gathering data during his/her field-trips.

We have developed the system's hardware using off-the-shelf robotic platform and sensors. In particular, we use SCOUT robotic platform, a 3D LiDAR to scan above the waterline part of the structures, and a microbathymetry to scan under the waterline part of the structures. We have successfully tested the robustness of this system to scan a jetty in rough Singapore water, where the water current is 1-2m/s.

We have developed a method and software for constructing 3D models of marine structures. The main problem in constructing the 3D models is the unreliable positioning sensors, as GPS signals are often blocked by the structures themselves. To overcome this problem, we use scan matching approach to develop a method that does not depend on any positioning sensors. The method and software has been tested to construct 3D models of marine structures using the data gathered by our robotic system.

Before the robot can efficiently scan the structures of interest, it must first be able to move from one point to another efficiently. A critical problem in a marine robot navigation around marine structures is that it must be able to take into account uncertainty in its motion, sensing, and environment map during planning, as each of these three sources of uncertainty may easily cause the robot to collide with the structures to be scanned. Our preliminary work in overcoming this problem can be seen here.

We have tested our hardware system to scan a jetty in Pulau Hantu (south-west of Singapore main island). Scanning the jetty.
The jetty we scanned at Pulau Hantu, Singapore. The model constructed from the scan data generated by our hardware system.


H. Kurniawati, J.C. Schulmeister, T. Bandyopadhyay, G. Papadopoulos, F.S. Hover, and N.M. Patrikalakis. Infrastructure for 3d model reconstruction of marine structures. In Proc. Int. Offshore and Polar Engineering Conference, International Society of Offshore and Polar Engineers (ISOPE), 2011.
bib |.pdf ]

G. Papadopoulos, H. Kurniawati, A.S.B.M. Shariff, L.J. Wong, and N.M. Patrikalakis. 3D-surface reconstruction for partially submerged marine structures using an autonomous surface vehicle. In Proc. IEEE/RSJ Int. Conf. on Intelligent Robots & Systems, 2011.
bib |.pdf ]

H. Kurniawati, T. Bandyopadhyay, and N.M. Patrikalakis. Global motion planning under uncertain motion, sensing, and environment map. In Proc. Robotics: Science & Systems, 2011.
bib |.pdf ]

H. Kurniawati, T. Bandyopadhyay, and N.M. Patrikalakis. Global motion planning under uncertain motion, sensing, and environment map. Submitted to Autonomous Robots special issue on selected papers from RSS 2011.
bib |.pdf ]

Last updated: 29 October 2011