PhD:
Exploration of a large underwater area with autonomous vehicles
Subject
This thesis considers a group of n low-cost underwater robots that have to explore a
large area (seabed mapping). Robots are not allowed to surface (except at the beginning
and at the end of the mission) and should organize themselves in order to perform
the exploration without being lost. The main objective of the group is to build a reliable
sonar/optical image (or a mosaic) of the seafloor without any hole in the image.
To build such an image it is important to limit as much as possible the drift generated
by the dead reckoning localization. This drift is particularly important for low-cost robots
which are not equipped with loch-doppler sensors
and accurate gyroscopes.
The drift can theoretically be controlled using SLAM
(Simultaneous Localization And Mapping) techniques. Since the seabed is not structured, it
is particularly difficult to detect automatically reliable seamarks that are needed to
perform the SLAM. What we propose here is to use some of the robots of the swarm as anchors.
The other robots are called explorers. For security reason, the anchors will
not be physically fixed to the bottom but instead, the robots are maintained stationary
thanks to a stabilization of the seabed image collected by the camera of each anchor.
To scan a large area, role of anchors and explorers will alternate with the evolution of
the mission. This strategy of exploration is called a walking strategy, where foot
steps are performed by anchors. In a submarine context, we can also assume that robots can
measure distances between them using the time flight of sound.
With this technique, we hope that we will avoid the localization drift due to time and limit
significantly the drift in space.
The objective of this thesis is to develop estimation and control methods in order to perform
a reliable and accurate image of the environment, using a walking technique.