Two-dimensional Photoelectric Compensation based on BP Neural Network

To improve the accuracy of heading angle measured by electronic compass, a new error compensation method is proposed based on adaptive differential evolution algorithm and BP neural network. In the method, the 3-layer BP neural network is used to model heading angle error, and adaptive differential evolution algorithm is adopted to train the weights of network, thus obtaining a more exact error model, and compensating the heading angle error measured by electronic compass. This method is compared with eight position least square, BP neural network, differential evolution algorithm, BP neural network optimization and other compensation methods. The accuracy of error compensation is improved obviously, and it has the characteristics of strong global optimization ability, fast convergence speed and good stability. The experimental results show that after compensation, the error range of heading angle is decreased from -16° ~30.7° to -0.22° ~0.2°, which satisfies the needs of higher accuracy navigation systems.