Differential evolution based on individual information parameter setting and diversity measurement of aggregated distribution

Differential Evolution (DE) and its variants have been widely applied to numerical optimization and engineering optimization problems owing to their simple operation, excellent optimization capacity, and high robustness. Many DE variants heavily rely on information from individuals to generate candidate offspring and update control parameters, which limits the search capacity during the later stages of evolution. A more appropriate search scheme is required to enhance DE's performance by utilizing information from both individuals and distributions. This paper presents a novel DE algorithm based on individual information and diversity measurement of aggregated distribution, termed IDMDE. First, to effectively regulate the search behavior of individuals, a hybrid parameter generation mechanism based on individual information is proposed. This ensures the algorithm always searches for a promising direction and fully utilizes the individuals' effective information. Second, to avoid the waste of search capacity during parameter updates in many DE variants, a novel parameter update strategy based on individual diversity is proposed, along with anew weighting scheme that utilizes the fitness and position information of individuals. Lastly, to mitigate premature convergence and stagnation during evolution, a diversity measurement mechanism based on aggregated distribution is proposed, which uses the search performance and diversity of individuals to evaluate the evolutionary state. The proposed IDMDE algorithm is evaluated by comparing it with five advanced algorithms on CEC2013, CEC2014, CEC2017, and CEC2022 across different dimensions. Moreover, the experimental results on the truss structure optimization problem confirm its feasibility in real-world optimization.