Abstract: To improve measurement precision of gyroscope, a novel nonlinear error compensation model is proposed using an algebra algorithm of neural networks. The model has advantages of real-time processing and precision mapping in the sample space because the algorithm transforms the complicated nonlinear optimization problem into linear algebraic equations. Compared to polynomial fitting, the proposed model has better precision. Experiments show that the model can reduce error to 0.1o/s and meet practical control needs.

Key words: gyroscope, neural networks, algebra algorithm, nonlinear error compensation