Mobile robot´s electronic compass calibration based on modified Fourier Neural Network

In order to improve the precision of the azimuth measured by mobile robot´s electronic compass, this paper proposes a new calibration method based on Fourier Neural Network trained by Modified Particle Swarm Optimization (MPSO-FNN). This method makes use of Fourier Neural Network (FNN) to establish the error compensation model of electronic compass´s azimuth, and introduces Modified Particle Swarm Optimization (MPSO) algorithm to optimize the weights of neural network. Thus the comparatively accurate error model of azimuth is obtained to compensate the output of electronic compass. This method not only has strong nonlinear approximation capability, but also overcomes the neural networks´ shortcomings which are too slow convergence speed, oscillation, and easy to fall into local optimum and sensitive to the initial values. Experimental results demonstrate that after calibrated by this method, the range of azimuth error reduces to -0.35°~0.70° from -3.4°~25.2°, and the average value of absolute error is only 0.30°.