An Multi-objective Evolutionary Algorithm with Lower-Dimensional Crossover and Dense Control

As it is hard to control the trade-off between fast convergence and the diversity of the population in designing an EA, this paper proposes an algorithm that can have a good control of both. The algorithm not only adopts the Lowerdimensional Crossover algorithm to accelerate the convergence but also proposes two good methods to keep the wide population distribution for global search. Also a new method is put forward as an algorithm of repulsing mechanism, that every solution repulses other solutions in its neighborhood with a value r t as its radius. With its control of the wide distribution of solutions and the diversity as well, this algorithm can prevent the solutions from falling into the local optimization and thus prompt global search for the optimal solutions. Then another method is used based on the concept of decomposition for the cutoff operator. As it can choose the solutions better than others in the non-dominated sets, and sort the solutions with well distributed, it is used to keep diversity of the population too.