3D Dynamic Obstacle Prediction and Obstacle Avoidance Method for Underwater Intelligent Robotic Fish Based on IMM ‐ DWA Algorithm

Aiming at the problems of 3D spatial navigation, dynamic obstacle interaction, and limited computing resources when the Dynamic Window Approach (DWA) algorithm is applied to obstacle avoidance for small underwater devices such as intelligent robotic fish in complex underwater environments, this paper proposes a dynamic obstacle prediction and obstacle avoidance control method based on the fusion of Interacting Multiple Model ‐ Dynamic Window Approach (IMM‐DWA). First, a Z ‐axis Proportion Integration Differential (PID) height control mechanism is introduced to extend the 2D kinematic model of the basic DWA algorithm. Second, by simplifying the IMM algorithm model, a Constant Velocity‐Constant Turn Rate and Velocity (CV‐CTRV) hybrid obstacle prediction model with adaptive weights is designed to improve the trajectory prediction of dynamic obstacles. Meanwhile, the velocity sampling space is optimized into spherical coordinate layered sampling to enhance sampling efficiency. Finally, the evaluation function is optimized to improve decision‐making quality. Simulation experiments show that the proposed IMM‐DWA algorithm significantly outperforms the basic DWA algorithm in terms of path length and running time, with smoother paths and reduced obstacle collision risk. © 2026 Institute of Electrical Engineers of Japan and Wiley Periodicals LLC.