Biomedical Engineering Reference
In-Depth Information
path towards its goal. The technique computes
the potential surface using Dijkstra's algorithm
in a moving window, updating the cost map as it
moves with the information of obstacles and target
obtained by the ultrasonic sensor, responsible of
the navigation module. A Fuzzy Logic Controller
(FLC) controls the wheels of a differential drive
robot to the angle of minimum potential. This
ensures a smooth trajectory towards the objec-
tive. A second FLC controls the average speed
of the platform.
To minimize calculations, a 20 x 20 cm grid
was defined and the cost and potential functions
were computed only considering points on the
grid. A window was defined covering a few cells
on the grid in the neighbourhood of the vehicle.
To compute the Dijkstra's algorithm, the local
goal was established in the intersection of the line
drawn from the actual position to the global goal,
with the border of the active window. The case for
a 13x8 window and a 20x20 grid was analyzed
with the mobile position fixed with respect to the
window. Sometimes the objective may fall in an
obstacle but when the windows moves, this local
objective will change. If the final objective is re-
ally occupied by an obstacle, a trap situation will
occur. Previsions in the algorithm should be taken
to cope with those situations, as well as wrong
trajectories generation due to the application of
such a sub-optimal method.
Later, cost map is placed on the active window
based on the information obtained by the sensors.
Then, with this partial cost map Dijkstra's algo-
rithm is computed in the window. First, the actual
position of the vehicle is taken as the ground node
and later, the local destination is used as the refer-
ence node. Potentials are added to obtain single
potential surfaces that hold the optimum path.
A new iteration follows: the window is moved
one-step in the grid and the previous procedure
is repeated until the local objective matches the
global one. More details can be found in (Calvo,
RodrÃguez & Picos, 1999).
The configuration of the objects was changed
with good results in most of the cases. A com-
parison was also made with the Virtual Force
Field technique (VFF) (Borenstein & Koren,
1991) giving shortest and smoothest trajectories
but with longer computing times. With this tra-
ditional path planning, the robot performs good
enough to face autonomous displacement in real
world. However, the mission planner was static
and the robot's ability to move avoiding obstacles
is constraint to the previous user programming
skills. The navigation scenario was glimpsed
beforehand with no possibility of smart reaction
to a new moving obstacle or a change in the target
point. Also, the robot had no ability to learn from
its own experience, which is a very important
feature in ER.
A second example is the desired trajectory
generation of an
Autonomous Underwater Ve-
hicle
(AUV) devoted to pipeline inspections.
In this case, the path planner is an instance of a
dynamic mission planner (DMP) built from the
experience of remote operated vehicles (ROV)
users compiled as a production system of forward
chained rules. The experimental vehicle was the
Geosub constructed by Subsea7, shown in Figure
2. It is robust enough to allow surveys at depths
of several thousands of meters. A main goal of
this research and development was to evaluate
experimentally if the current technology (by the
year 2004) was able to face autonomous inspec-
tions in deep water, practically with minimum
human intervention. The expert system called
EN4AUV
(Expert Navigator for Autonomous
Underwater Vehicle) (Acosta, Curti, Calvo, &
Mochnacs,2003-2005), showed the possibility
of incremental growth of the knowledge base as
more experience is gained on AUV navigation and
pipeline inspection, although consistency within
it must always be considered in these cases.
When compared to ROV based inspections,
AUV ones allow a smoother navigation and then
a more reliable data acquisition, because there is
not any umbilical cable to any ship or platform. In
Search WWH ::
Custom Search