Researchers from Disney Research, URJC Madrid, and ETH Zurich have developed a system by which a hopefully-optimum sensor network for soft robots can be quickly computed, creating a simple yet low-error system for reconstructing their shape when deformed by external forces.
"Soft robots have applications in safe human-robot interactions, manipulation of fragile objects, and locomotion in challenging and unstructured environments," the researchers claim. "We present a computational method for augmenting soft robots with proprioceptive
"Our method automatically computes a minimal stretch-receptive sensor network to user-provided soft robotic designs, which is optimised to perform well under a set of user-specified deformation-force pairs. The sensorised robots are able to reconstruct their full deformation state, under interaction forces. We cast our sensor design as a sub-selection problem, selecting a minimal set of sensors from a large set of fabricable ones which minimises the error when sensing specified deformation-force pairs. Unique to our approach is the use of an analytical gradient of our reconstruction performance measure with respect to selection variables."sensing capabilities.
The team's system was demonstrated in three example experiments, using an Arduino microcontroller to sense the voltage drop from the sensors. The first two — a bending bar and a pneumatic gripper — were built physically and proven in the lab; the third, a bio-inspired tentacle, was only simulated. All three, however, were able to find an optimum sensor network. From an initial set of 200 sensors, the bar ended up with just five, the gripper with a suggested 10 which was manually reduced to six to reduce complexity, and the simulated tentacle was given 14 sensors.
"We have devised a method that aids the roboticist with the automated sizing of a stretch receptive sensor network, capable of reconstructing the proprioceptive state of soft robots," the researchers conclude. "By introducing a set of selection variables in our sensing objective, we couple the design and sensing problems with a first-order optimality constraint. Intuitively, our design problem measures the performance of our sensing problem under changes to selection variables. Initialising a large set of fabricable sensors, we then use the analytical gradient of our design problem to select a small subset of sensors that results in a good reconstruction performance."
There's still work to be done, though: The team admits that it is difficult to model real-world interaction forces, and that the system as demonstrated can't handle unknown rather than pre-expected interaction behavior; it also doesn't include any guarantees that it will find a global optimum, despite having worked well for the three examples chosen.
The paper is available under open access terms on the Disney Research website.