Attitude determination for nano-satellites -- II. Dead reckoning with a multiplicative extended Kalman filter

TL;DR

This study presents a simulation-based evaluation of a novel attitude determination method for nano-satellites using thermal imaging sensors and a multiplicative extended Kalman filter, achieving sub-degree accuracy in ideal conditions.

Contribution

It introduces a simulation model testing a new attitude determination approach for nano-satellites, combining thermal imaging sensors with a multiplicative extended Kalman filter.

Findings

Attitude determination error is about 30 arcminutes with gyroscope drift.

Error can reach up to 25 degrees during Sun occultation.

Attitude information recovers immediately after Sun reappears.

Abstract

This paper is the second part of a series of studies discussing a novel attitude determination method for nano-satellites. Our approach is based on the utilization of thermal imaging sensors to determine the direction of the Sun and the nadir with respect to the satellite with sub-degree accuracy. The proposed method is planned to be applied during the Cubesats Applied for MEasuring and LOcalising Transients (CAMELOT) mission aimed at detecting and localizing gamma-ray bursts with an efficiency and accuracy comparable to large gamma-ray space observatories. In this paper we introduce a simulation model aimed at testing the applicability of our attitude determination approach. Its first part simulates the orbit and rotation of a satellite with arbitrary initial conditions while its second part applies our attitude determination algorithm which is based on a multiplicative extended Kalman filter. The simulated satellite is assumed to be equipped with a GPS system, MEMS gyroscopes and the infrasensors. These instruments provide the required data input for the Kalman filter. We demonstrate the applicability of our attitude determination algorithm by simulating the motion of a nano-satellite on Low Earth Orbit. Our results show that the attitude determination may have a 1$\sigma$ error of $\sim30'$ even with a large gyroscope drift during the orbital periods when the infrasensors provide both the direction of the Sun and the Earth (the nadir). This accuracy is an improvement on the point source detection accuracy of the infrasensors. However, the attitude determination error can get as high as 25$^{\circ}$ during periods when the Sun is occulted by the Earth. We show that following an occultation period the attitude information is immediately recovered by the Kalman filter once the Sun is observed again.