Geoscience Reference
In-Depth Information
edge choice for continuation of the path, are also not considered. For example, the
sight of several small corridors and a single large corridor at an intersection will
highly influence path choice and comfort when selecting the widest corridor and
not the smallest variant. Experiments with defining various risk value definition
with more parameters, individually weighted, should be considered in future work.
Related to this topic is the fact that the risk value of a decision point is currently
calculated based on the assumption that the wayfinder recognizes his mistake at the
first adjacent node and returns from there to the previous node. A question could be
raised whether it is actually realistic that people already notice at the first intersec-
tion that they have been going wrong. An increasing compounding function could be
suggested taking into account the possibility of going further in the wrong direction.
Second, in the current implementation of the least risk path algorithm, both the
length of the path as well as the sum of the risk values at intermediate decision
points have an equal weight in the calculation of the total risk value. Varying the
individual weight of both parameters might results in a more cognitively correct
calculation of the indoor least risk paths. Three different weighing adjustments
can be proposed: (1) geometric weighing by changing the length versus risk value
ratio; (2) semantic weighing by classifying corridor and outdoor areas differently
than rooms (resemblance with hierarchical network structure); (3) topological
weighing by taking the number and complexity of intersections into the definition.
The further elaboration on all three adjustments is subject for further research.
Third, the least risk path algorithm indoor was tested using a Geometric
Network structure as defined by Lee (
2004
), which each corridor being subdivided
in many hallway intersections in front of each doorway connected by short edges.
We have shown that this particular network structure can lead to increased risk
value calculations, deviations from the main corridor and misperceptions for the
wayfinder. Therefore, in the second stage of this research, various other network
structures (e.g. visibility based networks, networks without centreline transforma-
tions, cell decomposition, dynamic hierarchical networks …) will be examined in
order to quantify the dependency of the performance of cognitive algorithms on
various network topologies. Also, the dataset could be improved by classifying
edges in a hierarchical way to be in line with user's hierarchical spatial reasoning.
The main question here is which hierarchical structure should be used and how
should it be defined. In this case, a natural hierarchy similar to the road classifica-
tion hierarchy employed in outdoor navigational research has to be defined.
Fourth, staircases have been demonstrated in our analyses to be key elements in
the path choice and are typically one of the main reasons for getting-lost episodes in a
three-dimensional indoor environment (Hölscher et al.
2012
). The fact that you have
to walk up and down staircases could be naturally having a greater weight because
taking a wrong decision might result in walking up and down the stairs twice. On the
other hand, chances of taking a wrong decision by changing floors are likely to be
slimmer given the effort for vertical movement and a changed cognitive thinking.
In line with this last point, wayfinding research (Hölscher et al.
2009
) showed
the strategy choices people make when navigating in (un)familiar buildings, which
has proven to vary depending on the navigation tasks. The main strategies for
