<p>Urban waterlogging poses a significant threat to public safety, necessitating advanced vulnerability assessment methods. Traditional approaches often fail to capture dynamic human–environment interactions or account for interdependencies among influencing factors. An innovative dynamic vulnerability assessment framework is proposed, integrating hydrological-hydrodynamic modeling, multi-agent systems, and the Choquet integral method. This integration enables simulation of evolving inundation patterns and population movement during waterlogging events. The framework constructs a dynamic waterlogging environment by simulating time-series inundation depth and velocity. Agent-based modeling incorporates human behavior, simulating population distribution and movement pathways in response to floods. The Choquet integral is employed for vulnerability indicator fusion, effectively addressing the nonlinear interactions and dependencies between various factors that traditional methods overlook. Applied to a campus area, the framework demonstrates its capability to reveal the spatiotemporal evolution of waterlogging vulnerability, offering a more scientifically-grounded basis for urban disaster management and emergency planning.</p>