关键词: 分布式优化, 数据驱动优化, 约束优化, 进化计算, 差分进化算法

Abstract: Distributed data acquisition and processing in ubiquitous computing mode have brought the demand for distributed data-driven optimization. To address the challenges such as distributed data acquisition，asynchronous constraints evaluation and incomplete information，a Distributed Data-Driven Evolutionary Algorithm（DDDEA）framework for multi-constrained optimization was constructed. A series of terminal nodes were responsible for data provision and distributed evaluation，while the server nodes were responsible for global evolutionary optimization. Based on this framework，a specific algorithm instance was implemented，the terminal nodes utilized their local data to construct a Radial Basis Function（RBF）model to assist the differential evolution of the server node. Experimental comparison with three centralized data-driven evolutionary algorithms for multi-constrained optimization on two standard test suites show that，the DDDEA achieves significant optimal results in 68. 4% of test cases and has a success rate of 1. 00 in finding feasible solutions in 84. 2% of test cases. Therefore，the DDDEA has satisfactory global search and convergence abilities.

Key words: distributed optimization, data-driven optimization, constrained optimization, evolutionary computation, differential evolution