An octopus arm, rebuilt in silicone

An octopus arm is a robotics problem without a good rigid solution. It has no skeleton, infinite bending degrees of freedom, suckers along its full length, and a nervous system that runs mostly in the arm itself rather than the brain. Building a rigid manipulator that can reach into a crevice, wrap around an unknown object, and sense its own deformation is exactly the problem soft robotics exists to solve. On 29 November 2023 a team led by Zhexin Xie, with senior authors Li Wen of Beihang University, Yihui Zhang of Tsinghua University and Cecilia Laschi of the National University of Singapore, published in Science Robotics an electronics-integrated soft octopus arm, E-SOAM, that takes the biological model seriously rather than as decoration.

The control borrows the octopus’s move. Real octopuses do not aim the whole arm; they trigger a local bend that travels outward from the base, a bend propagation, and the travelling wave does the reaching and grasping with a small number of muscle groups. E-SOAM implements a bending-elongation propagation model on a soft arm: a wave of curvature is sent down the silicone body, the distal tip becomes the gripper, and the arm performs a reach-grasp-withdraw cycle that extends up to 1.5 times its original length. The control inputs are few, because the biological strategy is few-input by design.

The harder engineering problem is the wiring. A soft arm that stretches must carry its own sensors and electronics, and conventional copper trace cracks the first time it is strained. E-SOAM’s distal tip integrates a stretchable liquid-metal circuit that survives uniaxial stretching of 710 per cent and biaxial stretching of 270 per cent, and it processes three streams of sensory information at once: bending, suction and temperature. The tip works under highly deformed states, which is the regime a rigid sensor would simply leave. The arm can then reach, sense and grasp autonomously in a confined space, rather than being a passive appendage moved by an external stage.

The human interface is as telling as the mechanics. A human operator wears a single finger glove that produces suction sensations; one finger controls the arm’s in-plane and out-of-plane reaching and grasping, both in air and underwater. The design logic is that a soft arm’s value is in environments where a rigid arm cannot go, under water or through rubble, and that the operator should pilot it the way an octopus moves its own arm, by intention rather than by joint-by-joint commands.

The limitations are the usual soft-robotics ones. The arm is slow and weak compared with an industrial manipulator, its load capacity is modest, and liquid-metal circuits are not yet a commodity manufacturing technology. None of that is the point. The point is that a soft, highly deformable body can now carry integrated electronics, sense its own deformation, and be controlled with a small number of intuitive commands, in the water as well as in the air.

Soft robotics spent its first decade proving that compliant bodies could move. This generation of work proves that they can also perceive, and that perception can be packaged into the body itself rather than bolted on. An octopus arm rebuilt in silicone, with liquid metal for nerves, is a design study for the soft machines that will one day inspect pipelines and underwater reefs, and it arrives with a control strategy borrowed, unapologetically, from an animal with no bones and more neurons in its arms than in its head.