Electronic compass calibration based on Adaptive Differential Evolution Algorithm-Fourier Neural Network

Based on the comparison of several common methods of electronic compass error compensation, this paper presents a new error compensation method based on Adaptive Differential Evolution-Fourier Neural Networks (ADE-FNN) to improve the measurement accuracy of electronic compass. This method uses Fourier neural network to model electronic compass error, and adopts Adaptive Differential Evolution to optimize the weights of neural network, and get more exact error model to compensate measured values. The compensation object is the common electronic compass composed by two-dimensional magnetic resistance sensor. Compared with the compensation effect of Least-square method, BP neural network and Fourier neural networks, It proves that the mode of this method can realize the high precision in the sample space mapping and high non-linear approximation ability, and this method has faster convergence rate, can avoid falling into local minima, reduces the training error, and improves error compensation accuracy. This method decreases the error range from 3.4° ∼ 25.2° before compensation to -0.20° ∼ 0.72°, and the average of the absolute error is 0.30°. Repeatability tests also proved the compensation plan have a good consistency.