Research On Real-time Fusion Algorithm Based On MARG UAV Attitude Measurement System

In recent years,UAV has been widely applied in various fields such as scientific research,military and agriculture.Attitude measurement system is one of the core modules of UAVs.With the growing maturity of MEMS technology,the attitude measurement system based on the MARG(Magnetic-Angular Rate-Gravity)sensor has attracted more and more attention in view of the sensitive requirements for the weight,power,volume and manufacturing cost of small and low cost UAV.The attitude estimation of UAV usually uses Kalman filter to estimate the accurate attitude.However,when the UAV is in a maneuverable state or complex environment,the prior parameters of the initial Kalman filter model may not match the actual state,which leads to the reduction of the accuracy of the Kalman filtering and even result in the filtering divergence.What's worse,the conventional adaptive filtering algorithm may also lead to the process of re-convergence of the filter,this series of problems will threaten the safety and stability of UAV flight.Therefore,this paper first compensate for the magnetic measurement information sources of UAVs which in maneuverable state or complex environments.By researching the key technology of carrier magnetic interference compensation,a new algorithm based on the traditional recursive least square method is proposed,which is improved by the forgetting factor and the non-reference estimation,and the real time compensation for the UAV magnetic interference is realized.Secondly,aiming at the problem that the UAV attitude calculation model is inaccurate or the parameter switching leads to the re-convergence of the filter,the Multi Model Adaptive Estimation algorithm is applied to the UAV attitude calculation.With improving the Interactive Multi Model Adaptive Estimation algorithm by the decentralized filtering,the influence of the inaccuracy of the single model parameters is avoided under the condition that the precision is not reduced.Finally,through the analysis of the requirement of the system hardware,the hardware construction system which meets the requirements of low power,small volume and low cost is selected,and the hardware acceleration technology of the Cortex-M4 F kernel is used to realize the real-time measurement of the attitude of the Multi Model algorithm with large computation in the UAV.What's more,in order to verify the effectivenes of magnetic compensation algorithm,a simulation experiment of magnetic interference compensation is carried out in this paper.Experiments show that this algorithm can achieve the effect and accuracy requirement of timevarying carrier's magnetic interference without relying on the angle reference provided by the outside environment.Meanwhile,a semi-physical experiment is designed to verify the hardware acceleration and Multi Model algorithm.The results show that the hardware acceleration technology can improve the calculation efficiency by 56.4%.At the same time,the improved Multi Model algorithm can control the pitch angle and roll angle error within 0.5 degrees under the condition of the presence of variable noise,and the direction angle error is within 1 degree,and the algorithm has good stability and reliability.