Dynamic gridmaps: comparing building techniques

Mobile robots need to represent obstacles in their surroundings, even
moving ones, to make right movement decisions. For higher autonomy the
robot should automatically build such representation from its sensory input.
This paper compares the dynamic character of several gridmap building techniques:
probabilistic, fuzzy, theory of evidence and histogramic. Two criteria
are defined to rank such dynamism in the representation: time to show a new
obstacle and time to show a new hole. The update rules for first three such
techniques hold associative property which confers them static character, inconvenient
for dynamic environments. Major contribution of this paper is the
introduction of two new approaches are presented to improve the perception
of mobile obstacles: one uses a differential equation to update the map and
another uses majority voting in a limited memory per cell. Their dynamisms
are also evaluated and the results presented.