An optimization approach for team coordination through information sharing

Team coordination and information sharing are important in concurrent engineering (CE), where multiple design teams execute their tasks simultaneously and then share information to update their designs, e.g., through integrated tests. The process then iterates until the global design objective is optimized. When properly controlled and executed, CE can be an effective method to speed up the design process for complex and large-scale projects thanks to its parallel nature. Recently, a coordinate optimization framework is proposed in [1] to model and control the information sharing in CE. It can be shown that under a convexity assumption, CE converges to a globally optimal design. In this paper, we study how the coordinate optimization framework can be applied to CE in a general environment where the objective function is non-convex. We propose a simulation optimization method using a domain space cutting and optimal computing budget allocation to efficiently select the initial points from which the coordinate optimization can be applied under a mild local convexity condition. The proposed approach has broad potentials in decentralized control and optimization of complex and large-scale systems in automation. Numerical experiments show that the optimal selection of the initial points allow coordination optimization to efficiently find the global optimum.