- Morphological computation is a way of outsourcing intelligence to the body
- Spider webs provide complex information via vibration
- Similar physical outsourcing may lead to advances in robotics design
A spider crouches at the center of its web, waiting for a hapless fly to blunder against its threads. Though the web appears to be a simple trap, scientists consider it a complex information hub that extends the spider’s intelligence beyond its mind.
Constructed of five different silk materials, the web of a common garden spider (such as the Argiope spider), provides the spider with a home base and an effective way of catching some tasty snacks.
But scientists are beginning to suspect that the intricate structure of the web also transmits valuable information to the spider about the environment, prey, potential mates, and even predators.
Spiders probe their webs by sending out vibrations and receiving the response via mechanoreceptors on their legs. The differences in the resulting pulses locate and categorize events in the spider’s environment, effectively outsourcing pattern recognition to the physical structure of the web.
Scientists see possibilities in this physical outsourcing and are now investigating spider webs to find clues for improving future robot designs.
“We want to understand how vibrating morphology can carry out computational tasks like complex nonlinear signal processing, filtering, and information extraction,” says Helmut Hauser, lecturer in robotics at the University of Bristol. “We can then understand and extract general principles that go beyond spider web structures.”
The body knows
Morphological computation is a concept inspired by observations of nature. It theorizes that the physical bodies of biological systems (animals, plants, cellular structure, etc.) play a crucial role in intelligent behavior.
“In nature, computation does not just happen in the brain, but is partly outsourced to all over the body,” says Hauser.
A human example of this is the way in which the muscles and tendons in our legs react to uneven ground when running, and can adapt without communicating with the brain.
Nature provides more dramatic examples in the form of a trout with a body so well-designed that it can swim in flowing water even when it’s dead. Without brain activity, the body still interacts with its environment.
Or consider the Erodium seed, which literally drills itself into the soil when increased humidity signals favorable conditions for germination.
Hauser and his colleagues believe that studying and expanding upon these kinds of morphological computation could be the key to improved robotics design.
“Currently, robots are designed under the assumption of first making a body that is then controlled centrally by a ‘brain’,” says Hauser. “This works well in predictable environments like the assembly line of a factory, but it fails completely when applied to complex, unpredictable, and dynamic environments like our living and working spaces.”
Hauser suggests, however, that in order to design an intelligent body, the robot’s structure must demonstrate softness or compliance. “Other beneficial properties are nonlinear dynamics, complex bodies, and even noise. Yet these are all properties deliberately suppressed in conventional robots.”
Exciting the web
To examine the computational capabilities of spiders’ webs, Hauser’s study begins by first exciting different points on real spiders’ webs and measuring the vibrations (using a laser Doppler vibrometer) close to the center where the spider usually sits.
“We are looking into the complexity of the signals that are transmitted in the web and how different pathways through the web behave differently when excited,” says Hauser. “We want to understand how the web is transforming the input signal— how it’s filtering, nonlinearly combining, transforming, delaying, and damping the signals.”
Computer simulation and mathematical modeling will enable Hauser's team to extract fundamental design principles for building bio-inspired devices capable of carrying out useful computations through vibration.
The final step is to experiment with different materials, scales, and morphological structures to construct real sensor prototypes using what they have learned from the study. The goal is to build spider-like robot prototypes that can move around and deploy these sensors on demand.
Hauser envisions putting these new technologies to work via maintenance robots that could potentially detect earthquakes or structural failures in buildings and machines.
So, the next time you see a spider sitting calmly in the center of its web, don't be scared. It may be inspiring a future of robots whose intelligence lies in their bodies, not their brains.