Performance Comparison of Particle Filter in Small Satellite Attitude Estimation

The drive towards miniaturization, coupled with the latest advances in onboard processing, has given rise to small satellite missions’ ability to use more complex attitude estimation algorithms to fit their progressive mission requirements. Earth observation missions typically require higher satellite attitude pointing accuracies to precisely control the satellite orientation. Hence, to provide greater confidence in the attitude estimation accuracies, new advanced algorithms are continuously being developed. Satellite attitude estimation must be performed autonomously in real-time whilst optimizing computational resources such as time and memory. Small satellite missions with higher complexities tend to demand more sophisticated requirements, which push the limits of classical attitude estimation methods. The Particle Filter is an advanced Bayesian estimation technique that has shown significant improvements in satellite attitude estimation. This work describes the Particle Filter and its implementation to the attitude and angular rate estimation for a 3U CubeSat in Low Earth Orbit, whilst comparing attitude estimation performance in two different settings: with three-axis magnetometer measurements; and with combined measurements from a three-axis magnetometer and sun sensors. This work further reports that for attitude determination in small satellites, the Particle Filter is a more accurate attitude estimator than the widely used Extended Kalman Filter. The Particle Filter yields attitude estimation accuracy of ±0.01°, while the Extended Kalman Filter attitude estimation accuracy is ±1°. Moreover, the results indicate that the use of an additional sensor improves the attitude estimation accuracy of the Particle Filter by 17%. It is essential to consider different sensor combinations as it helps select the most suitable sensor suite and attitude estimator for an individual small satellite mission.