Motivation

The inherent non-linear properties of a soft material make developing control strategies and modeling motion difficult for soft robotics. Crowdsourcing, a way of obtaining information by enlisting the services of a large number of people, reduces this barrier, allowing human intelligence to perform as a genetic algorithm would. This work differs from other robotics crowdsourcing work in that the human participants are determining the robot’s control strategy, not simply moving the robot through a set of predesigned tasks using a pre- designed strategy.

Soft Robots

The vast majority of robots are hard and inflexible; yet, most animals are soft and flexible, traits that give animals numerous advantages over robots. Invertebrates like octopuses can fit through small holes and pipes, and even vertebrates can fit into small spaces, such as a cat crawling under a couch or a mouse squeezing through a hole in the wall. Like these animals, robots made entirely or partially from soft materials have the capability to interact with a wider number of environments, traverse multiple of terrains and an have increased dexterity. These abilities enable soft robots to achieve higher levels of biomimicry and larger deformations than traditional, hard-bodied robots.

Non-Linear Controls

However, the non-linear elastic properties, e.g. deformation rate, of soft materials makes developing control input strategies for soft robots difficult. Additionally there is no framework for identifying general control solutions for highly deformable moving structures. In order to overcome the difficulties of controlling a soft robot imposed by the robot’s material properties, we have developed a crowdsourcing platform in order to determine different strategies for motion. Crowdsourcing allows human intelligence to act as a sort of genetic algorithm, modifying and building upon what successes they have had. Different from other crowdsourcing work, this work uses human participants to determine strategies for controlling the robot rather than simply guiding the robot through an existing set of tasks using an existing strategy.

Crowdsourcing

Crowdsourcing has proven successful for generating and analyzing large data sets and for solving problems. Wikipedia as a resource and a community exists solely because people want access to a free and quality encyclopedia. Foldit players solved the riddle of an HIV enzyme structure in a matter of weeks, something that had been eluding professional scientists for years. But that wasn’t the only amazing thing Foldit players did, they also discovered and algorithm for protein folding comparable to yet better than one being developed by scientists. And recently, people around the world have rallied to scour satellite images for the missing Malaysian Airlines Flight 370.