A Sucker for Embodied Intelligence
Inspired by octopuses, scientists built a soft robot that senses, thinks, and acts by touch alone.
The responsiveness, privacy, and efficiency of on-sensor processing has made this technology a hot area of research in recent years. Applications in edge computing, IoT devices, and robotics have particularly benefited from advancements in this field, enabling smarter, more autonomous systems that can operate effectively without constant connectivity to the cloud or a central server. However, the pace of progress is not as fast as most people would like, because integrating processing units into tiny sensors is a challenging technical problem.
University of Bristol researchers have taken inspiration from the octopus to develop a new type of on-device sensor that may overcome some of the issues with existing technologies. The team has developed a soft robot that mimics the decentralized intelligence of the octopus. By using fluidic flows of air or water, this robot coordinates suction and movement without the need for a central computer, allowing it to interact with its environment in real time. The technology enables the robot to grip objects, sense textures, and respond adaptively — much like an octopus's own suckers.
The robot uses what the researchers call “embodied suction intelligence,” which integrates sensing and decision-making directly into the suction system. This means it can detect the type of surface it is touching, classify environmental conditions, and even estimate pulling forces — all without conventional hardware solutions.
The system functions at two levels. A low-level embodied intelligence enables adaptive gripping and object encapsulation, while a high-level perception interprets pressure data to understand environmental context. This dual capability allows the robot to interact with unknown or delicate objects safely and intelligently.
Soft robots built with this technology could be deployed in a range of fields, from agriculture and manufacturing to medical devices. They would be appropriate anywhere that gentle and adaptable manipulation is needed. They also offer new possibilities for wearable technology and human-safe robotics.
The researchers are now working on miniaturizing and strengthening the system for real-world applications. They also aim to integrate smart materials and artificial intelligence to further enhance adaptability in complex environments.