A magnetic-field-prediction-driven method for low-magnetic-exposure path planning of underwater platforms

Objectives To address the low computational efficiency of magnetic-field calculation and the difficulty in characterizing magnetic anomaly exposure risk in covert maneuvering path planning for underwater platforms, a magnetic-field-prediction-driven three-dimensional path planning method for reducing magnetic exposure is investigated. Methods A multi-medium three-dimensional finite element magnetic field model was constructed using COMSOL. Through parametric scanning, the near-field magnetic field distributions of the underwater platform at different spatial positions were obtained, and a local three-dimensional magnetic field dataset was established. A 16×16×16 regular sampling grid was used to extract the three magnetic field components around the underwater platform, forming a magnetic field tensor suitable for three-dimensional convolutional learning. A magnetic field prediction model integrating a multi-layer perceptron and a three-dimensional convolutional neural network (MLP–3D CNN) was then established to map the platform position to the near-field magnetic field tensor. On this basis, the predicted magnetic field strength and magnetic field gradient were incorporated into an improved A* algorithm, and a composite cost function including path length, magnetic field strength and magnetic field gradient was constructed. The probability of path detection was further evaluated using the cumulative magnetic exposure index, a two-layer detection model and the Monte Carlo method. Results The test results show that, in the main task, supplementary task A and supplementary task B, the cumulative magnetic exposure values of the A* paths are 37.787720, 26.224430 and 38.459623, respectively, all of which are lower than those of the corresponding straight-line and random paths. In the two high-risk tasks, the detection probabilities of the straight-line paths are 23.15% and 22.05%, respectively, while those of the A* paths are reduced to 2.70% and 4.30%, respectively. Conclusions The proposed method can reduce cumulative magnetic exposure and detection probability while maintaining path feasibility, providing a feasible engineering approach for covert maneuvering path planning of underwater platforms under the threat of magnetic anomaly detection.