AntBot, which its designers say is the first walking robot that moves without GPS. [Image: Julien Dupeyroux, ISM (CNRS/AMU)]
Safely navigating hostile environments such as the scorching hot desert without a global positioning system (GPS) is no mean feat. How does one endow a robot with this ability? Researchers in France took their cue from ants when designing their aptly named, six-legged AntBot. Inspired by desert ants’ navigation capacities, the walking robot uses an “optical compass” and an optical movement sensor to explore its environment and to return home to its base (Sci Robot., doi: 10.1126/scirobotics.aau0307).
The team believes that the ant-inspired localization system—with a reported mean homing error as small as 0.67 percent—could complement traditional navigation methods in autonomous mobile robots, or even in driverless vehicles.
No GPS, no problem
Robotics development is progressing at lightning speed, and it’s essential not only that navigation system designs keep pace, but that they are reliable, repeatable and robust. Fully autonomous mobile robots could be useful in myriad situations—from exploring extraterrestrial terrain or the aftermath of natural disasters to reconnaissance missions—but such situations often go hand-in-hand with spotty signals for GPS and wireless communication. With this in mind, the Aix-Marseille Université, France, team set out to design a new navigation system that included a robust, precise back-up option for when GPS fails.
To do this, the team borrowed tactics from desert ants. While many ants use pheromone trails to mark resources and find their way home, desert ants’ pheromones would burn up instantly in the extreme heat of their environment. Instead, in a process called path integration (PI), the wily insects observe the sky’s polarized light, the rate at which the ground moves across the eye—that is, optic flow (OF)—and count their strides to navigate. Using PI, desert ants can cross hundreds of meters in direct sunlight and scorching heat and return in a straight line back to the nest, without asking for directions.
Assembling AntBot
AntBot employs a process similar to PI to navigate its surroundings. The French team designed an “optical compass” to help the robot determine its heading via polarized light, and an optical movement sensor directed at the sun measures the distance traveled. AntBot’s ultraviolet (UV)-sensitive optical compass comprises only two pixels, with spectral sensitivity from 270 to 400 nm and a peak transmission at 330 nm, topped by two rotating polarized filters. The team reports that this relatively simple assemblage is equivalent to two arrays of 374 pixels, each tuned to a specific polarization angle. By turning the filters mechanically, the team was able to cut production costs for the compass from over €78,000 to a few hundred euros.
The second sensor uses 12 pixels that adapt autonomously to changes in light intensity to measure OF and calculate distance traveled. With these optical sensors, the fully 3-D-printed, 45-cm-wide robot was able to travel up to 14 meters away and back with high reliability in both cloudy and sunny weather.
Navigation demonstrations
With the general design in place, the researchers put five different PI modes to the test to find which homing method was the best model for use in high-efficiency robotic navigation—the variables being dependence on OF, stride integration and skylight polarization cues. The researchers found that AntBot gave the lowest homing error (0.67 percent) with the PI-Full mode—the fully ant-inspired method based on celestial cues to estimate heading angle, and a custom-designed combination of OF measurements and stride integration to determine distance travelled.
Neither trajectory distance nor trajectory shape nor sky conditions significantly affected homing performance using this model, which the researchers say indicates the system’s suitability for autonomous navigation in rugged, open-air environments. However, several challenges must be overcome before rolling out this bio-inspired navigation system for future applications. In particular, AntBot must walk at night, over longer distances and over steeper ground.