Edging along fences and creeping up walls, climbing plants send out tendrils in search of the sunniest spots in the garden.
In the lab, researchers have replicated the movements of nature countless times. Robots can walk, run and jump. They have even learned how to swim. Now, a team of mechanical engineers from Stanford University have taken inspiration for their latest robot from climbing plants. Following the lead of creepers such as ivy, the soft robot shoots out a tendril to ‘grow’ itself forwards.
The concept behind the idea is very simple and uses a process called ‘eversion’. The robot itself is a tube of soft plastic, folded inside itself. (Think of those slippery ‘water snake’ toys from the 90s!). As pressurised air fills the tube, the folded material turns the right way out, propelling the tip forwards.
Steering soft robots
The team see their robot nosing its way through the debris of disasters in search of trapped people. A camera at the front end guides rescuers, while the tip can deliver sensors for monitoring or drop off supplies to survivors.
In some iterations of the robot, certain parts can inflate independently, guiding the growing tip up, down, left or right. While the light-weight body has its advantages, the team admit driving the robot can be tricky.
Controlling a robot requires a sound understanding of how they move, something that is difficult to predict in a soft, flexible robot. The team also relies on images fed back from a camera in the robot’s growing tip to guide them.
Joey Greer, a graduate student in Stanford’s Okamura lab and co-author on the paper, explained. “Using a camera to guide the robot to a target is a difficult problem because the camera imagery need to be processed at the rate it is produced,” he said. “A lot of work went into designing algorithms that both ran fast and produced results that were accurate enough for controlling the soft robot.”
Back to boot camp
Putting the robot through its paces, the team constructed a gruelling obstacle course, complete with sticky glue, a bed of nails and an ice wall. The robot completed the course, despite picking up some punctures. Remaining stationary allowed punctures to self-seal around the nail, keeping the air inside the robot’s inflated body.
“The applications we’re focussing on are those where the robot moves through a difficult environment, where the features are unpredictable and there are unknown spaces,” said Laura Blumenschein, a co-author on the study.
In a series of other tests, the robot lifted a 100kg crate, squeezed through a gap ten times smaller than itself, and spiralled upwards into a free-standing structure to emit a radio signal. It also manoeuvred through a ceiling cavity, pulling a cable after itself and offering a new way to route wires in tight spaces.
The team hope future versions will see robots propelled by liquid, delivering water to trapped people or putting out fires. They are also testing tougher materials such as Kevlar and rip-stop nylon.
Going forwards, the plan is to trial both an up-scaled and down-sized version of the robot. Here the team can explore applications from disaster recovery to keyhole surgery.
Latest posts by QEPrize Admin (see all)
- Machine learning and AI – ensuring fairness in smart cities - October 11, 2018
- Create the Trophy reopens to young designers around the world - October 5, 2018
- Cleaning our oceans with AI-powered robot microscopes - September 24, 2018