Research On Dynamic Decoupling For Multi-Axis Sensor

Differential evolution algorithm is very popular in recent years, it is widely used in complex optimization questions because of its easy implementation and has less control parameters. But conventional differential evolution algorithm has a problem of stagnation that can stop the algorithm convergence. Aiming to solve the problem, this paper gives a kind of improved differential evolution algorithm,another ending condition was proposed, also can change the number of population (N),variation coefficient (F) and crossover coefficient (CR) in computing process. Elite tactic was added to the computing process to quicken the convergence speed and make differential evolution algorithm converge to best solutions. The dynamic coupling of a multi-axis sensor is defined as a situation that t he output signal of one direction includes the additional value, which affected by the input in other direction. Obviously,the dynamic coupling error will decrease the sensor measurement precision seriously. A new multi-axis sensor dynamic decoupling method is proposed based on a niche genetic algorithm and genetic programming. The method first gets the calibration data of the multi-sensor, and then uses the genetic programming method and the differential evolution algorithm to calculate and optimize the decoupling network based on the analysis method of transform function matrix. In addition, it also avoids the high precision requirement of the sensor model in the traditional dynamic decoupling method. Finally, the simulation results show the correctness and the affectivity of the proposed method.