Bio-Inspired Polarization Compass

Many emerging technologies like autonomous driving, robotic agriculture, and drone control promise to improve safety and efficiency in multiple sectors, however they require high-precision navigation systems which are currently insufficient. Existing navigation approaches rely on the global navigational satellite system (GNSS) that is prone to failure when satellite connection is unavailable. GNSS signal loss occurs in dense urban areas, which makes vehicle localization a challenge. Perhaps more problematically, jamming devices that block GNSS signals are becoming increasingly accessible. These inherent problems have pushed researchers to investigate methods of augmenting GNSS navigation with additional sensors to improve its precision and robustness. One such augmentation is inspired by the biology of the desert ant. This ant uses adaptations in its eyes to extract navigational cues from polarized skylight. When light from the sun hits the atmosphere, it generates a uniform polarization pattern that can be used to develop and maintain a navigational plan. Our work begins by constructing a polarization compass (PC) from the biological insights found in the desert ant. We then consider three algorithms for extracting navigational cues from the skylight polarization pattern captured by the PC: Hough transform (HT), support vector machine (SVM), and linear regression (LR). We demonstrate that PC provides a heading estimation that is accurate to less than 1 degree, which is better than existing magnetic compasses. Our work shows that PC is an attractive augmentation for GNSS-based systems which may improve their performance under adverse conditions. In future work, the PC will be integrated with other sensors, including GNSS, to assess its performance under realistic conditions. This will be carried out in both terrestrial and aerial domains.