Autonomous navigation requires multi-sensors data fusion provided either by global navigation satellite systems (GNSS) devices, inertial measurement units, radars and cameras to achieve accurate localization. Each technological solution features advantages but suffers also from drawbacks. Data fusion aims at maintaining a strong level of accuracy and robustness to make autonomous navigation systems reliable enough to be embedded on board any autonomous vehicles. However, there are still environmental contexts in which most sensors drift or even fail to provide correct estimates. In this study, we discuss the opportunity to use a celestial compass inspired by the desert ants' visual system which is able to extract heading information from the polarization pattern of the skylight in the ultraviolet (UV) spectrum. This new sensing mode has been mounted on-top a car and tested outdoor on a 18.6km-long journey in town and compared with GNSS estimates. The UV celestial compass yielded promising performances regarding its low complexity and the root mean square error of the orientation error was only 0.55°. Our results suggest the suitability of such parsimonious insect-inspired solutions for robotic purposes in urban field like the last mile delivery.