This work proposes and evaluates an evolutionary system using genetic algorithms and directed graphs for the optimization of a scheduling problem for ore loading of ships in a harbor. In this kind of problem, some tasks are constrained in such a way that they must be planned or executed before others. For this reason, the use of conventional evolutionary models, such as genetic algorithms with an order-based representation, might generate invalid solutions which can not be penalized, needing to be discarded or corrected, leading to a loss in performance. To overcome this problem, we use a hybrid system, based on directed graphs, to allow better handling of the precedence constraints. Results obtained show performances almost 3 times better than a non-trivial search algorithm.
展开▼
