Attitude Determination in Space with Ambient Light Sensors using Machine Learning for Solar Cell Characterization

Exploration of novel thin-film solar cell technologies outreaches for their application in space. For extraterrestrial tests, irradiance conditions must be well determined to extract quantitative solar cell performances. Here, a new method for solar position determination is presented, based on parallelized ambient light sensor measurements is presented obtained from the sounding rocket experiment Organic and Hybrid Solar Cells In Space during the MAPHEUS-8 mission. The solar position evolution is optimized using stochastic and gradient-based methods in a Bayesian approach. Comparison with independent positioning estimates shows compelling agreement, lying mostly within 5° deviation. The inclusion of a simple Earth irradiation component mitigates a small systematic offset. Further, solution uncertainties are estimated with Monte-Carlo Markov-chain sampling. The point-source irradiation model's accuracy can compete with that of a camera-based trajectory. During equatorial Sun positions, the method's precision appears even higher––the 1σ uncertainty of the derived solar position is as small as 3° for the effective angular deviation. This simple sensor array triangulation method being complementary to other attitude determination methods shows reasonable accuracies and allows implementation in systems of limited computational capabilities to determine the solar position or irradiance conditions for space or terrestrial solar cell applications.