Quantum computing shows great potential for solving complex problems but faces challenges due to hardware limitations and noise. In 2D architectures, the nearest-neighbour connectivity requirement causes excessive SWAP gate insertions, increasing latency and noise accumulation, hindering large-scale algorithm implementation. This work introduces an optimized mapping framework for 2D architectures that address qubit placement and dynamic mapping. Our contributions include (1) an interaction-aware initialization strategy, (2) a genetic algorithm for latency optimization, and (3) a multi-process mapping mechanism. Evaluations show significant improvements, reducing SWAP gates by 25.0%–54.54% across benchmarks. Under different quantum architecture scales, the multi-process program execution method achieves 89.24% resource utilization and reduces execution time by up to 48.92%.