The study explores the application of Cellular Automata (CA) in the design of architectural unit combinations to generate innovative architectural forms. By leveraging CA's self-organizing properties, the research aims to systematically address complex design requirements and enhance design diversity and flexibility. The construction of a CA model tailored for architectural design involves defining grid structures, establishing transition rules, and setting initial configurations and boundary conditions. Through iterative simulation and refinement, the model produces various architectural patterns, evaluated based on spatial efficiency, connectivity, and aesthetic quality. Key findings indicate that hybrid rule sets, which combine deterministic and probabilistic elements, offer the most balanced designs, optimizing both functionality and visual appeal. Despite the promising results, the study identifies several limitations, including scalability issues, complexity in rule formulation, and integration challenges with existing design workflows. Future research directions include scaling the model, automating rule generation, integrating with BIM and CAD tools, validating through real-world case studies, and exploring interdisciplinary applications. The study highlights the potential of CA in transforming architectural design by providing innovative, efficient, and aesthetically pleasing solutions.