Around the inspection robot electromagnetic navigation，an analysis of the characteristics of strong nonlinear mapping between the electromagnetic sensor array information and the robot pose is made，and a method using the back propagation（BP）artificial neural network fusion method for the data fusion array of electromagnetic sensors on the robot is proposed.By making use of the neural network nonlinear fitting approximation characteristics and through offline learning training，a nonlinear relationship model is built between the electromagnetic sensor arrays information and robot arm spatial position relative to the power line，to achieve the autonomous navigation of inspection robot.The method can effectively reduce the influence of the detection error of single electromagnetic sensor information on the robot navigation control and improve the robot's line grasping control accuracy.Finally，the feasibility and effectiveness of the method is verified by experiment.