Geoscience Reference
In-Depth Information
5 Discussion on Adjusting the Least Risk Path Algorithm
The previous analyses have shown multiple times that only limited differences can
be found in terms of length and risk value between the least risk path algorithm
and the shortest path algorithm. This indicates that both algorithms often return
paths with a similar path choice. For short path lengths, this is to be expected as
the path choice is limited by the limited density of the indoor network. Also, given
the typical network structure with a main corridor connecting various rooms and
the importance of staircases in connecting various floor graphs; often not many
options exist on a short distance to deviate from the shortest path. However, for
paths with a more extensive total path length, we have seen varying results with
sometimes large differences in path choice and sometimes barely any difference.
Also, when there were differences, the least risk path algorithm selected theoreti-
cally less risky paths (when compared to our benchmark parameter set), but evenly
as many times the shortest path would still be preferred to guide unfamiliar users
during their wayfinding endeavours.
As shown, the least risk path algorithm does not return stable results in terms of
selecting the least risky edges in indoor environments. Therefore, we are inclined
to say that at this point the least risk path algorithm indoor calculates alternative
routes between two points, without necessarily reducing navigational complex-
ity. This leads us to believe the least risk path algorithm and its definition of risk
should be investigated in more detail and altered to be more aligned to the specifi-
cities of indoor wayfinding.
5.1 Possible Improvements to the Algorithm
In this final section, we will suggest some other improvements to the original algo-
rithm which will be tested and compared in our future research.
First, the way in which the risk value is defined by only taking into account
the average wrong path length and the intersection complexity (i.e. number of
edges converging) could be one of the reasons for the currently inaccurate results.
Because of its current definition, the algorithm will always try to select the long-
est edge (larger risk value cost if not chosen), which is not necessarily always the
least risky edge (e.g. bumping into complex intersections, less integration and vis-
ibility…). Also, the risk value weights the intersection complexity (i.e. number of
edges) according to an exponential relationship: i.e. the more edges converging,
the less importance to the total number of edges. It should also be noted that up to
this point no aspects denoting the overall individual importance of each edge, apart
from the edge length (e.g. width, number of curves, integration value), are yet
incorporated in the assessment of risk. On intersection level, other aspects like the
directional orientation of each edge, local visibility, etc. that can also influence the
