Comparison of Attitude Estimation Algorithms With IMU Under External Acceleration

Abstract

For improved flight stability, the flight controller needs to compute precise attitudes of the quadrotor at a fast update rate. This paper provides a comparison between different sensor fusion algorithms for estimating attitudes using an Inertial Measurement Unit (IMU), specifically when the accelerometer gives erroneous readings. Three test cases with an emphasis on the influence of external acceleration on attitudes are selected. For each test case, noise filtered data from the IMU is streamed into four algorithms, namely Complementary, Kalman, Madgwick, and Mahony fusion filters. Furthermore, each algorithm is implemented on ESP32 (XtensaOR 32-bit LX6) microcontroller to benchmark the execution time. The estimated attitudes show that the Madgwick filter mitigates the effects of accelerations the most, while the Kalman filter and Mahony filter are robust to vibrations introduced to the system.