UAV attitude estimation based on the dual filtering methods

Purpose A toy UAV performs tumbling, rolling, racing and other complex activities. It is based on low-cost hardware and hence requires a better algorithm to estimate the attitudes more accurately with low power consumption. The proposed technique based on optimized Madgwick filter and moving average filter (MAF) ensures improved convergence speed in estimating the attitude, achieves higher accuracy and provides robustness and stability of the toy UAV. The paper aims to discuss this issue. Design/methodology/approach Traditional methods are prone to problems such as slow convergence speed and errors in calculation of the attitude angles. These errors cause the vehicle to drift and tremble, thus affecting the overall stability of the vehicle. The proposed method combines the features of optimized Madgwick filter and MAF to provide better accuracy, achieved through the fusion of gyroscope and accelerometer data, and zero correction to eliminate the random drift error of the gyroscope and removal of high-frequency interference by MAF of the accelerometer data. The experimental results on actual flight data showed that the method was better than the conventional Madgwick and Mahony complementary filters. Findings The performance of the proposed method was analyzed by estimating the pitch and roll angles under the static and dynamic condition of the toy UAV. The results were compared with two traditional methods: Madgwick and Mahony complement filter. In the static condition, the variance and average error while estimating the attitudes was comparatively lower than the traditional method. For the dynamic conditions, the convergence time to achieve a prescribed swing angle was again lower than the traditional method. From these two experiments, it can be seen that the proposed method provides better attitude estimation at lower computation time. Originality/value The proposed method combines the optimized Madgwick filter and MAF to accuracy estimate the attitude of toy UAV. The algorithm mainly suits the toy UAVs which are based on low-cost hardware and require better control systems to ensure stability of the vehicle. The experimental results on real flight data illustrate that the method not only improves the convergence speed in estimating the attitude angle for large maneuvers of the toy UAV, but also achieves higher accuracy in the attitude estimation, thus ensuring the robustness and stability of the UAV.