by Günter Kiechle
The use of Geographic Information Systems (GIS) for preparing
input data for optimisation algorithms improves the practical
applicability of such algorithms in the field of transportation
planning. This promising combination of technologies is the subject of a
collaboration between the University of Vienna and Salzburg Research.
Vehicle Routing Problems (VRPs) are a widely investigated class of
problems in combinatorial optimisation, and include many transportation
tasks (eg parcel services). In general, a VRP consists of a set of
customers that must be served via a fleet of vehicles, each of which
leaves from and returns to a central depot. The type of VRP determines
whether customers have goods delivered to them, are transported from one
location to another, or are served in some other way.
Using GIS for Real-World Input Data
In research, most solution techniques for this class of problem are designed and tested by means of synthetic problem structures. However, the tackling of real-world VRPs requires a thoroughly elaborated data basis in order to provide reasonable outcomes. If this is not the case, even the best solution techniques are of no use for practical applications.
In research, most solution techniques for this class of problem are designed and tested by means of synthetic problem structures. However, the tackling of real-world VRPs requires a thoroughly elaborated data basis in order to provide reasonable outcomes. If this is not the case, even the best solution techniques are of no use for practical applications.
Essential input data for real-world VRPs is gathered by using Geographic Information Systems (GIS). Whereas most researchers use
Euclidean distances between customers and depots for their optimisation
algorithms, a GIS can provide real distance information derived from a
digital road network.
Experiences in former projects showed that using distance data of
limited quality in optimisation algorithms leads to results of even more
limited quality. In the worst case, a valid solution for a given input
dataset might actually be unfeasible in reality.
more about GIS applications:
No comments:
Post a Comment