Abstract
Inspections of aquaculture net pens are regularly performed
using remotely operated vehicles (ROVs), as structural failures can lead
to fish escape incidents with huge environmental and economic impacts.
Net pens are large 3D structures that deform with waves and ocean
currents such that their shape cannot be predetermined. In this paper, we
propose a coverage planning method to automate ROV inspections of net
pens. The net pen is parameterized with azimuth and depth coordinates,
and boustrophedon cellular decomposition is used to compute a coverage
plan using this parameterization. Moreover, the real shape of the net pen
is estimated using range measurements from a forward-looking Doppler
velocity log. Finally, CHOMP trajectory optimization is used to guide the
ROV along the coverage plan, where a functional gradient to maintain a
desired distance to the net is also proposed. The method is evaluated in
simulations.