Abstract: To address the path planning problem of mobile robots in dynamic complex environments, a hierarchical path planning algorithm based on nonlinear model predictive control and informed RRT* (NMPC-IRRT*) is proposed. The algorithm is divided into upper and lower layers: the upper layer utilizes the informed RRT* algorithm based on obstacle distribution to generate a high-quality global initial trajectory through an elliptical sampling heuristic strategy. The lower layer employs an NMPC local navigation controller, combined with a linear motion model to predict the future trajectories of dynamic obstacles, achieving real-time path tracking and dynamic obstacle avoidance. Simulation comparisons show that in complex dynamic environments, the average running time of the algorithm is 51.47 s, which is approximately 71.5% shorter than that of the DWA-IRRT* algorithm, and the average path length is reduced by 32.9%. In high-density scenarios where the number of dynamic obstacles increases to 30, the planning success rate remains above 90%, and its trajectory curvature variance and angular velocity change rate are the lowest among the compared algorithms. Through the synergy of global topological guidance and local predictive optimization, the algorithm addresses the problems of poor path quality and insufficient real-time performance in dynamic obstacle avoidance of traditional sampling algorithms, achieving autonomous navigation for robots in complex unknown dynamic scenes.