Simulation and optimization of pedestrian environments in urban residential neighborhoods based on agent-based modeling

Despite extensive research on pedestrian environments, simulation-oriented frameworks that explicitly link planning, behavioral simulation, and spatial optimization remain limited. This study develops an agent-based simulation framework for urban walking spaces by integrating an Attraction Evaluation Model (AEM) with the Slime Mould Algorithm (SMA), enabling the explicit modeling of interactions between built-environment characteristics and pedestrian route choice. The model is validated using empirical pedestrian data from ten representative residential neighborhoods across five block types in Beijing. Results show that the baseline model achieves strong predictive performance, with a coefficient of determination of 0.778, which further improves to 0.796 after incorporating population heterogeneity. Distinct built-environment mechanisms are observed across neighborhood types: RDB and RCB are primarily influenced by functional diversity and accessibility, REB is more strongly associated with safety and walking continuity, while RB and RGB are mainly shaped by visual landscape quality and environmental comfort. Moreover, distinct age-specific preferences are observed, with adults favoring functional mix and service accessibility, retirees showing stronger reliance on green environmental quality and walking comfort, and adolescents exhibiting weaker but stable responses related to safety and public space accessibility. By linking simulation outputs with modifiable spatial elements, the framework enables targeted optimization for pedestrian-oriented urban regeneration.