Italian technology researchers recently developed a soft robotic arm inspired by octopuses. By integrating distributed sensing and control in the "suction cup", it achieves local "thinking" and autonomous grasping capabilities similar to the octopus's tentacles. It is designed to be used for exploring complex and unpredictable seabed environments.
This research was led by a team from the Biomimetic Soft Robotics Laboratory of the Italian Institute of Technology (IIT). They drew inspiration from the nervous system structure of the octopus: Although the octopus has only a relatively small central brain, about 60% of its neurons are distributed among its eight tentacles. Each arm can process information on the spot and trigger reflex actions, such as catching prey independently without waiting for instructions from the brain. The research team is trying to replicate this distributed architecture in a robotic system using silicone and electronic components, so that perception and movement are closely integrated in the flexible "body" itself, rather than relying on a single central processor.
The result is a soft robotic arm with a length of about 41 centimeters and a base diameter of about 4 centimeters. Its shape and structure are similar to the tentacles of an octopus. It is equipped with 10 artificial "suction cups" from the base to the end, and the size gradually becomes smaller. The system is designed not to rely on cameras, external computers or centralized control units, but instead decentralizes all core perception and primary decision-making capabilities into the suction cup. The research results have been published in the journal Nature Machine Intelligence, and IIT has also released introductory materials for the public.
Each artificial suction cup integrates three light-emitting diodes and three phototransistors to form a micro-optical sensing system, which is used to measure changes in reflected light, which is equivalent to the local nerve nodes of this "mechanical wrist". When a foreign object contacts the surface of the suction cup, the silicone material deforms and the path of the reflected light changes accordingly. The system determines whether contact occurs, the strength of the contact, and the direction of incidence, thereby forming three key sensing data. Tests show that the system's force sensitivity is about 400 millivolts per Newton, and the force measurement error is about 0.1 Newton, which is roughly equivalent to the weight of several paper clips; the maximum error in direction recognition is less than 18 degrees, and the average error is about 8 degrees, which is close to the angle between adjacent scales of a clock.
In terms of control architecture, this soft robotic arm uses two levels of control: the first level is completely executed locally - each suction cup has an independent circuit and will trigger adsorption immediately once contact is detected without waiting for central instructions; the second level is at a higher level and is responsible for receiving data uploaded by all suction cups, comprehensively analyzing the target position and contact characteristics within a time window of about 4 seconds, and deciding the overall grasping strategy based on this, such as letting the robotic arm bend upward or downward, or rotate, and if necessary, cover the autonomous actions of local suction cups. The research team stated that this solution of integrating sensing and signal processing directly into the suction cup enables the robotic arm to respond to contact in real time and accurately without centralized control, and has good scalability and robustness, and can operate in complex environments including underwater.
All experiments are currently conducted underwater. During the test, this robotic arm was able to detect objects such as glass bottles and glasses during movement, gave an estimate of the weight of the grabbed object at about 72.5 grams, and the actual weight was 85 grams, and was able to manipulate targets at different angles, including an artificial "starfish". In terms of load capacity, the robotic arm can lift objects up to about 500 grams, and its sensing performance remains stable after 300 cycles of repeated use, showing good durability. Since each suction cup only sends refined information such as contact direction to the upper control part instead of complete raw data, the bandwidth requirements of the entire system are greatly reduced, so that it can be easily expanded to more suction cups or even multiple contact wrists without significantly sacrificing response speed.
The research team pointed out that the design has strong modular characteristics, and the number and layout of suction cups can be flexibly adjusted according to different tasks. Potential application scenarios include inspection of underwater infrastructure such as submarine pipelines, cables and platforms, and collection of biological samples in narrow or complex areas that rigid robots cannot reach. With its flexible structure and autonomous decision-making capabilities, this "octopus-like" robotic arm is expected to provide new technological paths in fields such as deep-sea exploration, ocean engineering, and underwater maintenance.
Octopuses have long been an important source of inspiration for bionic designs in the field of robotics. As early as 2017, the German automation company Festo demonstrated the OctopusGripper at the Hannover Messe. This is a silicone wrist-touch gripper driven by compressed air. It uses two rows of suction cups to wrap around the target and complete the grasp when it is inflated, but it still relies heavily on external air pressure control and manual manipulation. In recent years, researchers from the University of Bristol in the UK have approached the situation from another dimension, no longer copying the shape of the tentacles, but studying the mucus secreted by the octopus sucker. They have developed a new type of suction cup composed of a multi-layered soft structure and a bionic fluid system. It can simulate the way the octopus mucus seals gaps on rough curved surfaces, so as to grasp irregular objects such as stones and wood that are difficult to firmly attach to traditional suckers.
Taking a step further, research teams from Peking University, National University of Singapore, Zhejiang University, Beijing Institute of Technology and other institutions jointly designed the OUT-Robot mechanical grasping system to simulate the grasping strategy of cephalopods, allowing it to quickly switch between soft and rigid states to sort and grasp objects of different shapes, different flexibility and different weights. Compared with these previous attempts, the biggest feature of IIT's new design is "autonomy"-it can not only crawl, but can decide how to crawl. The researchers also emphasized that the geometry of the objects selected in the current experiment is relatively simple. The next step includes testing with more complex and diverse shapes and weights, and trying to introduce brain-like neuromorphic computing to make the entire system closer to the neural circuits of real octopuses in terms of information processing.