A Dual-System Variable-Grain Cooperative Coevolutionary Algorithm: Satellite-Module Layout Design

The layout design of complex engineering systems (such as satellite-module layout design) is very difficult to solve in polynomial time. This is not only a complex coupled system design problem but also a special combinatorial problem. The fitness function for this problem is characterized as multimodal because of interference constraints among layout components (objects), etc. This characteristic can easily result in premature convergence when solving this problem using evolutionary algorithms. To deal with the above two problems simultaneously, we propose a dual-system framework based on the cooperative coevolutionary algorithm (CCEA, e.g., cooperative coevolutionary genetic algorithm) like multidisciplinary design optimization. The proposed algorithm has the characteristic of solving the complex coupled system problem, increasing the diversity of population, and decreasing the premature convergence. The basis for the proposed algorithm is as follows. The original coupled system P is decomposed into several subsystems according to its physical structure. The system P is duplicated as systems A and B, respectively. The A system is solved on a global level (all-in-one), whereas the solving of B system is realized through the computation of its subsystems in parallel. The individual migration between A and B is implemented through the individual migration between their corresponding subsystems. To reduce the computational complexity produced additionally by the dual-systems A and B, we employ a variable-grain model of design variables. During the process of optimization, the two systems A and B gradually approximate to the original system P, respectively. The above-proposed algorithm is called the dual-system variable-grain cooperative coevolution algorithm (DVGCCEA) or Oboe-CCEA. The numerical experimental results of a simplified satellite-module layout design case show that the proposed algorithm can obtain better robustness and trade-off between computational prec- - ision and computational efficiency.