Overview
Facing applications that require the resolution of optimization problems
of unceasingly increasing size and this within increasingly short times, even
in real time, only the jointly implementation of advanced methods resulting
from combinatorial optimization in operational research, from the decision in
artificial intelligence, and from the use of parallelism and distribution would
allow to lead to satisfactory solutions.
After having modelled the problem by graphs, or mathematical programming, the problem can be solved, according to its complexity, either by an exact method (to get an optimal solution) or by an approximate method (to get a good solution).
Exact methods use dynamic programming, optimization in graphs or the arborescent research.
Among recent approximate methods one may cite metaheuristics of local search like simulated annealing, tabu search, or metaheuristics based on populations like genetic algorithms, ant colonies, and dispersed seek.
Parallelism and distribution allow to accelerate the search and to solve problems of large size until then inaccessible by the sequential algorithmic. Parallelism is not only a source of computing power. The parallel cooperation between algorithms can improve quality of the obtained solutions.