Many mobile games adopt auto-battle systems that the major consideration of players is how to assemble strong teams. The automated team assembly (ATA) then becomes a crucial issue from different standpoints, such as to assist the game designers in performing balance analysis and to direct the players to configure teams. Since ATA is generally a combinatorial optimization problem, this paper exploits the evolutionary optimizers. However, unlike the traditional problems that the evaluation functions are explicit, in ATA, we are unable to evaluate the team strengths straightforwardly. To address the issue, we collect data from the server and build an end-to-end deep learning surrogate for estimating the team strengths. The model has a three-layer architecture of a feature embedding layer, a sequential relation layer, and a regression layer, which is able to characterize the complex dependencies between the sparse input features and the team strengths. The evolutionary algorithms are then guided by the constructed surrogate to seek for the strongest teams. Several adjustments are also made on the evolutionary algorithms to adapt it to the ATA problem with multiple constraints. Simulations on the game Romance of the Three Kingdoms: Strategy Edition validate the good performance of the proposed data-driven evolutionary optimizers.
For computationally intensive problems, data-driven evolutionary algorithms (DDEAs) are advantageous for low computational budgets because they build surrogate models based on historical data to approximate the expensive evaluation. Real-world optimization problems are highly susceptible to noisy data, but most of the existing DDEAs are developed and tested on ideal and clean environments; hence, their performance is uncertain in practice. In order to discover how DDEAs are affected by noisy data, this paper empirically studied the performance of DDEAs in different noisy environments. To fulfill the research purpose, we implemented four representative DDEAs and tested them on common benchmark problems with noise simulations in a systematic manner. Specifically, the simulation of noisy environments considered different levels of noise intensity and probability. The experimental analysis revealed the association relationships among noisy environments, benchmark problems and the performance of DDEAs. The analysis showed that noise will generally cause deterioration of the DDEA’s performance in most cases, but the effects could vary with different types of problem landscapes and different designs of DDEAs.