Complex High-Speed Interconnect Design Based on Collaborative Bayesian Optimization

The design of high-speed interconnects in advanced packaging faces challenges due to complex signal paths and tight constraints. This paper proposes a Collaborative Bayesian Optimization (CBO) framework to address these issues efficiently. By introducing problem decomposition, surrogate-assisted modeling, and subspace refinement, CBO enhances scalability and optimization performance. Validation on a 2+2 stacked differential interconnect shows that CBO outperforms conventional Bayesian Optimization in minimizing insertion and return losses, demonstrating its effectiveness in high-dimensional design tasks.