AI-enabled Inverse Design of Planar RF Passives Under Arbitrary Footprint Constraints

Designing planar RF passive electromagnetic (EM) structures requires balancing performance, cost, and physical footprint. Traditional workflows rely heavily on expert intuition, slow EM simulations, and iterative ad-hoc tuning. Moreover, fitting a passive structure into a prescribed area is often challenging: even when template-based designs exist, enforcing footprint constraints typically requires laborious geometric bending, repeated re-simulation, and manual adjustment to achieve the desired response. This paper presents an AI-enabled framework that rapidly synthesizes arbitrarily shaped EM structures realizing target scattering parameters (S-parameters) through a size-agnostic neural surrogate model. For the first time, this approach enables the synthesis of RF passive structures explicitly constrained to fit within a specified area. The method generalizes to any passive component with arbitrary S-parameter specifications and arbitrary footprint limits, bounded only by physical electromagnetic constraints. This capability has the potential to significantly streamline EM and RF circuit design, enabling rapid, footprint-aware synthesis of complex passive structures.