Collaborative Research: Autonomous Design of Complex Systems through Contextualized Generative Design Policies and Learned Coordination

Modern engineering systems are becoming increasingly complex, requiring designers to consider thousands of interacting decisions, competing objectives, and evolving requirements. While artificial intelligence has shown promise in accelerating engineering design, current approaches often rely on large collections of existing designs and struggle to generalize to new problems or coordinate decisions across complex systems. The goal of this project is to develop new AI methods that enable increasingly autonomous design of complex engineered systems while preserving the rigor and reliability of established engineering optimization techniques. The research will create a new generation of AI-based design methods that learn to synthesize engineering designs by combining modern generative AI with classical engineering optimization and systems engineering principles. These methods will take into account a rich design context ranging from design requirements and previous design decisions to design guidelines to make more informed design choices under a wide range of design problems. The project will develop algorithms that learn reusable design knowledge from related engineering problems, intelligently combine learned strategies with established optimization methods, and scale to large systems through coordinated multi-agent decision making. Chiplet-based computer processors and spacecraft systems will serve as representative application domains, demonstrating the broad applicability of the proposed methods to complex engineering design problems. The project will advance AI-enabled engineering design by reducing design time, improving design quality, and enabling more efficient exploration of innovative engineering solutions. These advances have the potential to strengthen U.S. competitiveness in strategically important areas including semiconductor technologies and space systems. The project will also develop open-source software and workforce training modules for AI-enabled design space exploration. By making these resources openly available, the project will help prepare the next generation of engineers to effectively apply AI in the design of future complex systems. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.