An ant-inspired celestial compass applied to autonomous outdoor robot navigation

Abstract

Desert ants use the polarization of skylight and a combination of stride and ventral optic flow integration processes to track the nest and food positions when travelling, achieving outstanding performances. Navigation sensors such as global positioning systems and inertial measurement units still have disadvantages such as their low resolution and drift. Taking our inspiration from ants, we developed a 2-pixel celestial compass which computes the heading angle of a mobile robot in the ultraviolet range. The output signals obtained with this optical compass were investigated under various weather and ultraviolet conditions and compared with those obtained on a magnetometer in the vicinity of our laboratory. After being embedded on-board the robot, the sensor was first used to compensate for random yaw disturbances. We then used the compass to keep the Hexabot robot's heading angle constant in a straight forward walking task over a flat terrain while its walking movements were imposing yaw disturbances. Experiments performed under various meteorological conditions showed the occurrence of steady state heading angle errors ranging from 0.3 • (with a clear sky) to 2.9 • (under changeable sky conditions). The compass was also tested under canopies and showed a strong ability to determine the robot's heading while most of the sky was hidden by the foliage. Lastly, a waterproof, mono-pixel version of the sensor was designed and successfully tested in a preliminary underwater benchmark test. These results suggest this new optical compass shows great precision and reliability in a wide range of outdoor conditions, which makes it highly suitable for autonomous robotic outdoor navigation tasks. A celestial compass and a minimalistic optic flow sensor called M 2 APix (based on Michaelis-Menten Auto-adaptive Pixels) were therefore embedded on-board our latest insectoid robot called AntBot, to complete the previously mentioned ant-like homing navigation processes. First the robot was displaced manually and made to return to its starting-point on the basis of its absolute knowledge of the coordinates of this point. Lastly, AntBot was tested in fully autonomous navigation experiments, in which it explored its environment and then returned to base using the same sensory modes as those on which desert ants rely. AntBot produced robust, precise localization performances with a homing error as small as 0.7% of the entire trajectory.

Bio-inspired celestial compass yields new opportunities for urban localization

“…This method has been previously investigated and resulted in outstanding performances with a strong resilience toward both the meteorological and environmental conditions [26], [28]. The results also suggested that applying a median filter would have an interesting impact on the final heading estimation.…”

“…We developed an ant-inspired sky compass [25], [26] ( Fig. 1B-D) that analyses the polarization pattern of the skylight to extract the AoP around the zenith, corresponding to the heading of the vehicle with respect to the sun.…”

“…1) The AntBot method [25], [26]: the estimation of the heading angle requires to carry out a complete rotation of the polarizers, which lasts for 4 seconds considering an angular resolution of ∆ψ = 360 • /224 = 1.61 • (N = 224). Due to the symmetry of the polarization pattern across the sky dome, the signals measured by the sky compass are 180 • -periodic sine waves with a phase opposition.…”

“…Their method was tested on a 2.52km-long journey, resulting in an orientation RMSE equal to 0.92 • despite foliage. Recently, Dupeyroux et al demonstrated the superiority of the AoP detection in the UV range under foliage, whose measurement depends mainly on meteorological conditions (clear sky: 0.39 • ; covered sky: 0.59 • , using a median filter) [26].…”

“…Section 2 presents the celestial compass and how to use it for heading estimation purposes. Three distinct methods are considered, namely a GPS-based approach, and two methods relying on the celestial compass: the AntBot method [25], [26], and the Hilbert method, introduced in this study. Section 3 describes the results obtained over a 18.6km long journey in town.…”