A hybrid multi-agent and genetic algorithm framework for multi-objective spatial allocation in urban regeneration

Modern urban regeneration must address increasing complexities due to land‑use conflicts, strict environmental controls, and multi‑actor stakeholder relations. Traditional planning approaches, limited in their capacity for dynamic simulation and systemic integration, frequently fail to balance diverse spatial optimization aims. Here, we propose IMAGO (Integrated Multi‑Agent Genetic Optimization), an original framework that links a Multi‑Agent System (MAS) with a Multi‑Objective Genetic Algorithm (MOGA) to establish reciprocal feedback between stakeholder decisions and transformations of the built environment. Implemented in Wuhan’s core urban zone, IMAGO synthesizes parcel‑level geospatial information, socioeconomic indicators, ecological standards, and regulatory limits. It models the decision‑making of public authorities, developers, and residents, directly incorporating their dynamic negotiations into the genetic search cycle. Empirical evaluation confirms that IMAGO reliably outperforms conventional planning models on key performance indicators: land‑use diversity (0.842 ± 0.025), park service coverage (0.683 ± 0.020), inhabitant satisfaction (0.756 ± 0.022), and developer return on investment (1.842 ± 0.045). These findings emphasize how feedback from agent‑based behavioral simulation increases the feasibility of planning solutions. The generated layout demonstrates smart land‑use transitions: obsolete industrial sites are redeveloped into integrated neighborhoods, underused peripheral areas are converted to ecological buffers or low‑density housing, and fragmented green spaces are merged into continuous systems. These changes correlate with demographic trends and policy priorities, showing that simulated stakeholder actions productively guide spatial reorganization. Consequently, IMAGO offers a transferable computational platform for aiding complex urban renewal decision‑making.