Fast Marching subjected to a Vector Field - path planning method for Mars rovers
Expert Systems with Applications 78:

planning is an essential tool for the robots that explore the surface of
Mars or other celestial bodies such as dwarf planets, asteroids, or
moons. These vehicles require expert and intelligent systems to adopt
the best decisions in order to survive in a hostile environment. The
planning module has to take into account multiple factors such as the
obstacles, the slope of the terrain, the surface roughness, the type of
ground (presence of sand), or the information uncertainty. This paper
presents a path planning system for rovers based on an improved version
of the Fast Marching (FM) method. Scalar and vectorial properties are
considered when computing the potential field which is the basis of the
proposed technique. Each position in the map of the environment has a
cost value (potential) that is used to include different types of
variables. The scalar properties can be introduced in a component of the
cost function that can represent characteristics such as difficulty,
slowness, viscosity, refraction index, or incertitude. The cost value
can be computed in different ways depending on the information extracted
from the surface and the sensor data of the rover. In this paper, the
surface roughness, the slope of the terrain, and the changes in height
have been chosen according to the available information. When the robot
is navigating sandy terrain with a certain slope, there is a landslide
that has to be considered and corrected in the path calculation. This
landslide is similar to a lateral current or vector field in the
direction of the negative gradient of the surface. Our technique is able
to compensate this vector field by introducing the influence of this
variable in the cost function. Because of this modification, the new
method has been called Fast Marching (subjected to a) Vector Field
(FMVF). Different experiments have been carried out in simulated and
real maps to test the method performance. The proposed approach has been
validated for multiple combinations of the cost function parameters.