Insects in general are unfailingly impressive with how intelligent and capable they are, with an absolute minimum of sensing and computing power. Where things start to get really interesting is when insects have to get clever in order to manage particularly challenging environments. Desert ants are a great example of this: While most ants rely on pheromone trails to navigate (they retrace their smell trails to get back to the nest), the heat of the desert means that pheromones don’t last very long. Instead, desert ants rely on a variety of techniques, including step counting, optic flow, landmarks, and most notably solar navigation.
These techniques seem like they could come in handy for small, inexpensive robots exploring out in the solar system, where GPS isn’t available and sophisticated sensors come with a mass and power budget to match. Today in Science Robotics, researchers describe how they built a robot with desert ant-inspired navigation tools, and were able to get it to wander around a little bit and still find its way home without GPS, SLAM, or anything more complex than some slightly fancy eyeballs.
The first thing to understand is how desert ants navigate. In general, the system that they use is called “path integration,” which is essentially the same as what we would call “dead reckoning.” By keeping track of distances and directions traveled over time, the ants can calculate the most direct path back to where they started. Basically, if the ant heads north for a while, and then east for twice as long, it knows that by traveling south and west (twice as long), it’ll end up pretty close to its starting position, and once it’s close, it can visually recognize landmarks to get exactly back to its nest.
Desert ants are remarkably good at this, as the figure below shows. After a wandering outbound trip of nearly 600 meters over about 20 minutes, the centimeter-long ant is able to plot a more or less exactly straight line directly back to its nest in just six minutes.
A desert ant C. fortis uses “path integration” to find its way back to its nest. The thin line shows the outbound trajectory (592.1 meters), with small black dots representing time marks (every 60 seconds). The ant went straight back to its nest (thick line, 140.5 m long). The small circle (lower right) marks the nest entrance, and the large black one shows the feeding location (top center).Image: Aix Marseille University/CNRS/ISM
For path integration to work, the ant has to track two separate things: distance and direction. Distance is the easier one of these by far, since the ant can use a (very familiar to robots) combination of stride counting and optical flow. Direction is the tricky one—it’s well known that ants and other insects can use the Sun to navigate, tracking its location in the sky and correcting for the rotation of the Earth and the consequent apparent motion of the Sun over time. This would only work when it’s actually sunny, except that the ants’ eyes have photoreceptors that are sensitive to polarized light, which can indicate the direction of the Sun even if it’s overcast. The ants are also UV sensitive, helping them see the Sun through cloud cover and foliage.
AntBot is powered by a Raspberry Pi 2B board and its sensors include a celestial compass, IMU, and optical flow sensor.Image: Aix Marseille University/CNRS/ISM
AntBot is an attempt to replicate the sensing systems of desert ants to see how well an autonomous system could use them for ant-inspired navigation. AntBot is a 2.3-kilogram hexapod, the specific physical specs of which are not really all that important for the purposes of this research. What is important are AntBot’s sensors, which include a bioinspired optic flow sensor, and an “an insect-inspired celestial compass” consisting of a pair of UV light sensors with rotating linear polarizers. The compass analyzes the the log-ratio between the data from these two sensors to determine the angle of polarization of the incoming light, which it uses to determine where the Sun is, and therefore which direction it’s pointing in. AntBot can do this very accurately: The median error was just 0.02° when the sky was slightly cloudy, 0.59° under an overcast sky.
By combining optical-flow distance tracking, step counting, and celestial navigation just like the desert ant, it probably won’t surprise you that AntBot was able to repeatedly wander around randomly over a distance of about 14 meters, and then successfully return to its starting point. This is good, but AntBot still has some work to do to prove that it’s as talented as an ant, as the researchers point out:
In its present form, AntBot has a diameter of 45 cm and walked at a speed of about 10 cm/s during the experiments, whereas C. fortis desert ants are only 1 cm long. As shown in Fig. 1A, the ant’s trajectory measured 732.6 m. AntBot should therefore have covered more than 32 km to be properly compared with ants’ navigation performances. Although AntBot can walk at speeds of up to 90 cm/s, very large scale navigation will require improving the hexapod robot’s actuators and power supply. These improvements will make it possible to test the PI-Full mode in more natural contexts, such as rugged terrains in a cluttered environment (forests) where the view of the sky is often barred by the presence of branches and leaves in the visual field of the celestial compass.
Insects may have been the first to figure out this polarized light trick, but it’s possible that humans have been using a similar technique to help them navigate for centuries. There is some evidence to suggest that the Vikings (as well as later seafaring cultures who probably got the idea from the Vikings) could have relied on polarized light to find the location of the sun under an overcast sky using a sunstone—one of a small number of minerals that are birefringent. Birefringent minerals are polarizers, and when light enters them, it gets split into two rays that take different paths through the stone depending on where the light source is relative to the stone. By looking through the stone at the sky, it’s possible to use birefringence to figure out where the sun is within a few degrees, even if it’s completely overcast, or if the sun is below the horizon. All you need is a little bit of sunlight, and a sunstone will work.
The most common birefringent mineral is calcite, and the Vikings would have had access to that. A few Viking legends refer directly to sunstones, and simulations have shown that using a sunstone could have potentially made a huge impact on the ability of Vikings to make extended voyages across the open ocean. Viking ships and burial sites haven’t yielded much calcite, but it’s rather fragile as minerals go and wouldn’t necessarily last all that long underwater or in the ground. And if they didn’t end up using something like this to navigate, well, they really should have, because both the ants and the robots are getting great results with it.