Information Technology Reference
In-Depth Information
function Crossover(Genome A, Genome B):
location list = locations in Genome B that are not in Genome A
new child Genome = copy of Genome A
for 0 to some number of crossovers:
randomly select a gene in the child Genome
set that gene's location value to the next location in
the location list
return child Genome
For example, if GenomeA = (AA12, [1,1,1]) (BB34, [2,2,2]) and GenomeB = (CC56,
[1,1,1]) (AA12 ,[3,3,3]), the unique locations in Genome B would be simply [3,3,3],
and one possible child genome would be (AA12, [3,3,3]) (BB34, [2,2,2,]). The effect
is to construct a child genome that maintains the move ordering of Genome A, while
incorporating location material from Genome B.
Figure 4 shows results from a single EA run of an example scenario involving the
extraction of 20 containers from a 10
2 yard. The containers are requested accord-
ing to a fixed schedule and the EA is initialised with a population of 200 individuals.
Termination is after a given number of generations. The figure shows the population
converging on a lower (that is, better) fitness value, where the fitness value is propor-
tional to the total distance of Straddle Carrier moves to complete the scenario.
×
3
×
Fig. 4. Preliminary Experimental Results (minimal complexity problem)
7
Discussion and Conclusion
We have presented a port simulation platform ( ContMAS ) which has been implemented
as a multi-agent system. The ContMAS platform can serve as an integrating framework
for solutions to different aspects of the container terminal management and optimisation
problem. ContMAS can be seen as a basic simulation tool which can be equipped with
different optimisation strategies (called module agents). These module agents are easily
replaceable, thus, allowing an easy integration of different optimisation strategies. This
has been illustrated with the description of two specific aspects of the problems that we
have tackled, namely yard allocation and straddle carrier move optimisation, which have
been investigated, implemented, and integrated with ContMAS .Furthermore,wehave
conducted evaluations (some of which are covered in this paper) which have shown that
Search WWH ::




Custom Search