The octopus is a natural escape artist. It can squeeze its soft body into impossibly tight spaces and often baffles aquarium workers with its ability to break out of tanks. These abilities could be very useful in an underwater robot, which is why the OCTOPUS Project, a consortium of European robotics labs, is attempting to reverse engineer it in all its tentacled glory. Now researchers from the Foundation for Research and Technology (FORTH), in Hellas, Greece are learning how the robot might use its tentacles to swim.

Back in 2011 we got a glimpse of the project's first stab at an artificial tentacle, which could have far-reaching implications for soft-bodied robots of the future. However, it's when you combine multiple tentacles that you can begin to explore how they help the octopus propels itself underwater. The FORTH team – Michael Sfakiotakis, Asimina Kazakidi, Nikolaos Pateromichelakis, and Dimitris P. Tsakiris – are experimenting with various methods of sculling.

Most of the time an octopus crawls to get around, but they swim quickly to avoid predation using a built-in water jet. Added propulsion comes from undulating all eight tentacles in unison, but the researchers have found that the robot can also move quite smoothly by fluttering its arms independently (something the real animal does not do).

Besides creating a dynamic simulation (which takes fluid drag into account), they also experimented with an actual prototype. You can see how these strategies affect locomotion in the following video:

This work was presented at the IEEE International Conference on Robotics and Automation (ICRA) 2013 last month. Moving forward, the team plans to experiment with other gaits and will add a stretchy webbing around the base of the tentacles to see how these affect its swimming. Eventually, the robot will be given a pump jet motor to simulate how a real octopus expels water.

Assuming all goes swimmingly, the OCTOPUS Project hopes to "open up new scenarios for marine exploration and underwater rescue," though admittedly that still seems a little far off.

Source: OCTOPUS Project via IEEE Spectrum