Multi-Algorithm Collaborative Path Planning for High Voltage Distribution Room Inspection Robots under Electromagnetic Field Interference

To address the problems associated with the operation of inspection robots in high-voltage distribution rooms, such as the fuzzy confirmation of the detection target, vulnerability to strong electromagnetic field interference, and the path algorithm being prone to falling into local optimal solutions, a multi-algorithm collaborative path planning method is proposed. First, a dual-stream CNN-SVM model is constructed. It integrates two independent feature extraction frameworks to learn multi-dimensional data features, with data fusion performed via a channel concatenation layer. This design ensures accurate identification of partial discharge detection targets. Next, an A* algorithm with adaptive electromagnetic field interference is developed by modifying the traditional A* algorithm. Specifically, it incorporates an electromagnetic interference cost term and a density-adaptive weight factor, while introducing a dual-detection fusion smoothing strategy. These improvements enable the global inspection path to effectively avoid high electromagnetic field areas and enhance search efficiency. Finally, the method integrates with a dynamic window algorithm guided by global nodes. By planning local paths based on key nodes of the global path, it achieves optimized local path planning. Experimental results show the dual-stream CNN-SVM model performs excellently on the test set, with high scores in AUC, Precision, and Recall, confirming its superior classification ability. In two test scenarios, compared with the traditional A* algorithm and its fusion algorithms, the proposed algorithm reduces path length by up to 15.5% and 12.2% respectively. These results demonstrate its excellent performance and strong robustness, making it suitable for meeting operational needs in complex high-voltage distribution room environments.