AntBot: a Fully Insect-Inspired Hexapod Homing like Desert Ants
Résumé
Desert ants Cataglyphis fortis are endowed with an amazing navigational toolkit. While foraging for food, these insects keep tracking their nest entrance by updating their so called path integrator (PI). This localization system requires celestial cues for the estimation of the heading (solar elevation, polarized skylight, color gradient), and odometric cues based on ventral optic flow and stride-counting (Wehner et al., Ethol. Ecol. Evol., 1990). An insectoid robot, called AntBot, was built and equipped with insect-inspired minimlistic sensors: (i) a celestial compass composed of two UV-sensitive photodiodes topped with rotating linear polarizing filters to mimic the dorsal rim area of the compound eye (Dupeyroux et al., Sci. Robot., 2019); (ii) an optic flow sensor comprising only 12 pixels that can auto-adapt to light changes within 7 decades (Mafrica et al., Opt. Express , 2015) to measure the walking speed. AntBot was also made able to count its stride while walking. Those three sources of information were used in a series of 5 PI models tested over 130 homing experiments (Dupeyroux et al., Sci. Robot., 2019). The fully ant-inspired PI allowed the AntBot robot to reach outstanding homing performances under lighting conditions similar to those encountered by ants, with an overall homing error as small as 0.67% of the full trajectory distance (Dupeyroux et al., Sci. Robot., 2019). AntBot is the first robot ever to successfully perform PI with ant-inspired sensing modalities in real world conditions. We hope that biologists will consider our open-source robotic platform to test in real-time various models of ant-inspired strategies including their neural basis that are still difficult to validate in simulated conditions.
Origine : Fichiers produits par l'(les) auteur(s)