Applying multi-objective optimization to a ventilation duct for an automotive vehicle

The Archive-based Micro Genetic Algorithm is an evolutionary optimization algorithm that relies on genetic operators for creating new solutions. The generation scheme deployed in AMGA can be classified as generational since, during each generation, only solutions created before that iteration take part in the selection process. However, the algorithm generates a very small number of new solutions (it can work with just two solutions) at every iteration.
Therefore, it can also be classified as an almost steady-state genetic algorithm. This attribute helps the algorithm achieve faster convergence than most generational genetic algorithms.
AMGA owes its name to the fact that it works with a very small population size and uses an archive to maintain its search history. The external archive is updated at each iteration and provides the information about the search history of the algorithm, which is then exploited during the selection process to create the parent population.

The size of the archive determines the computational complexity of the proposed algorithm. However, for computationally expensive optimization problems, the actual time taken by the algorithm is negligible. The parent population is updated using the archive and binary tournament selection is subsequently performed (to create the mating pool).

AMGA borrows several ideas from NSGA-II, SPEA2, and FastPGA algorithms. The parent population is created from the archive using an environmental selection-based strategy borrowed from SPEA2. However, the size of the parent population is influenced from the dynamic population sizing approach used in the FastPGA. The creation of the mating pool uses binary tournament selection and is the same as used in NSGA-II. The strategy used to update the elite population relies on the domination level, diversity of the solution, and the current size of the archive, and it is based on the non-dominated sorting concept borrowed from NSGA-II.

AMGA uses efficient diversity preservation techniques. For the case of two objectives, AMGA uses the crowding distance (borrowed from NSGA-II). For the case of more than two objectives, AMGA relies on Efficient Nearest Neighbor Search (used in vector quantization). The genetic variation operators used by AMGA are Simulated Binary Crossover (SBX) and polynomial mutation.
AMGA is effectively a micro-GA combining the best characteristics of well-known algorithms. As a matter of fact both SPEA2 and NSGA-II can be emulated by AMGA by adjusting parameters appropriately.