Review and Comparison of In-field Calibration Methods of Accelerometer Based on Gravity Vector

Considering the wide applications of accelerometers to determine position and attitude and due to reducing of accuracy of this sensors because of some errors, this paper discusses the calibration of accelerometers. Also because the traditional calibration methods are very time consuming, costly and need precision laboratory equipment, in-field calibration methods are recommended which are simple, fast, cheap and don't need equipment. In these methods, by using of gravity vector a cost function is defined and after optimization of this function, calibration coefficients includes bias and scale factor which are the most important error factors in output model of accelerometer are obtained. Then accelerometer is simulated in MATLAB software and these calibration methods are used to calibrate sensor and they are compared in terms of accuracy, pace, the effect of different noises and the effect of different initial conditions. The fastest and simplest method is least square method but all methods are the same accuracy. In term of effect of noise, four methods of Levenberg-Marquardt, Newton, quasi-Newton and least square are the same and the best and in term of effect of initial condition, the best method because of independence from value of initial-conditions is least-square and after that quasi Newton algorithm and Levenberg-Marquardt algorithm are good.