A Multi-Scale Agent Urban Community Spatial Planning Algorithm Based on Cellular Automata Modeling

With the rapid acceleration of global urbanization, urban community spatial planning faces increasingly complex multi-objective optimization challenges. Traditional planning methods, which rely heavily on expert experience, often struggle to adapt to the dynamic and nonlinear nature of urban environments, highlighting the urgent need for intelligent solutions. To address this, this study proposes a multi-agent, multi-scale optimization framework based on deep reinforcement learning. The framework decomposes urban spatial planning tasks into three levels—macro, meso, and micro—allowing for coordinated optimization across scales and achieving comprehensive spatial planning from overall layouts to local details. By integrating a grid-based foundational model with cellular automata, the framework provides agents with precise environmental inputs, enabling dynamic simulation of urban functional zone expansions. Experimental results demonstrate significant improvements in land-use efficiency, public service accessibility, and environmental harmony. Compared to traditional methods, the proposed approach exhibits superior flexibility and adaptability, offering robust technological support for the intelligent and automated evolution of urban spatial planning.