This study improves the performance and adaptability of the insulation network laying robot system for power system tasks. Convolutional Neural Network (CNN) is used to enhance path planning by enabling the robot to intelligently select the optimal path based on environmental features and obstacle distribution. This reduces path length, execution time, and improves smoothness. Experimental results show that the ML-based adaptive control strategy outperforms traditional obstacle avoidance algorithms. While the traditional algorithm avoids obstacles 80 times, with 20 collisions and an average avoidance time of 10 s, the ML strategy achieves 95 successful avoidance, 5 collisions, and an average time of 7 s. These results demonstrate that the ML strategy significantly improves obstacle avoidance efficiency, reducing collisions and cost time.