Papers by Keyword: Steepest Descent Algorithm

Abstract: As an optimization which is of very strong robustness and independent of the concrete areas of the problems, the genetic algorithm are discussed and applied in analyzing the signals of the overlapped peaks more and more. However, whether the Gauss-type or the Lorentz-type spectrum lines, the functions of which are usually very complicated. Thus, it is very difficult to obtain a favorable solution using the genetic algorithm, and there are too many generations of calculation. Therefore, in this letter, the steepest descent algorithm is added to the genetic algorithm. And the fast converged favorable solutions are obtained through simulating the analyzing the Gauss-type overlapped peaks.

Abstract: This paper presents a novel procedure to calibrate the strap-down 3-axis MEMS accelerometers for UAV navigat-ion. Firstly, we establish an explicit calibration model with the measurement values of accelerometers, where the calibration is realized via geometric transformations. Secondly, the transfor-mation parameters are calculated through particle swarm optimization (PSO). For the problem of slower convergence rates near the global optimum, the classical PSO algorithm is improved. Based on the numerical optimization idea, the steepest descent method is introduced to PSO. The parameters are searched in the rough by adopting PSO and the precision ones are found by using steepest descent method. Then, the optimal transformation is achieved by the minimum distance function based on this improved PSO(IPSO) algorithm. Finally, the calibration procedure is tested by comparing the attitude produced by the 3-axis accelerometers with that measured by a turntable. The results show that the IPSO algorithm can significantly improve the performance of the classical PSO algorithm, and the maximum attitude error is reduced to 6% of that before calibration. In addition, the proposed procedure does not rely on prior knowledge of the accelerometers and any equipment. So, it is suitable for calibration in field. Such a method is especially useful in UAV applications.