Skip to main content

A Soft Artificial Muscle Driven Robot with Reinforcement Learning

Penelitian - Researchers of the Zhejiang University in Hangzhou presents a soft artificial muscle driven robot mimicking cuttlefish with a fully integrated on-board system including power supply and wireless communication system.

Without any motors, the movements of the cuttlefish robot are solely actuated by dielectric elastomer which exhibits muscle-like properties including large deformation and high energy density.

Penelitian A Soft Artificial Muscle Driven Robot with Reinforcement Learning

Conventional robots are made of rigid components to provide large output force, high precision, and ease of controllability. When operating in complex environments, bio-inspired soft robots possess unique advantages. Natural creatures are adaptive and resilient to environment.

Fabricated with soft and deformable polymers, bio-inspired robots that mimic natural creatures have drawn a growing interest in recent years. Ultimately, soft robots can perform various tasks beyond the limits of conventional robots, achieving instinctive characteristics in terms of safe for humans, geometric adaptation4, and tunable camouflage.

Currently, the most widely used approaches to control the robot are assuming the mechanical structure as rigid body, but those approaches are not applicable on soft robots. In general, the actuation of soft robot itself is difficult to be modelled. Even for a simple task, it usually requires complicated mechanical analysis.

To-date, there are significant contributions by experts of artificial intelligence and robotics in trying to surmount the modeling and learn to perform specific tasks for soft gripper based on imitating and reinforcement learning. Reinforcement learning is an adaptive control strategy that serves as a potential solution to the control of soft robots.

Unlike fishes that acquire thrust most often by wave-like movements of the fish’s body, fins and tail, cuttlefish and jellyfish move by jet propulsion. In detail, these cephalopods draw water into their body, then expel the jet of water from a rear orifice to generate a series of vortex rings and hence thrust. This mechanism has been studied in-depth by researchers.

Inspired by the structure and propulsion mechanism of cuttlefish, Tiefeng Li of the Zhejiang University and team have designed a biomimetic cuttlefish robot with DE membranes (3 M VHB) as the artificial muscles.

“The cuttlefish robot uses surrounding open water as the electric ground, which makes it more robust when faced with the insurmountable challenge of high voltage to actuate the artificial muscle (DE membrane),” said Li.

“To make the cuttlefish robot more compact, a highly compact electric system (Epod) is selected for both remote control and voltage boosting,” Li said.

Other than design of the overall on-board system, reinforcement learning is implemented to optimize the strategy toward the actuation of the cuttlefish. The cuttlefish robot reaches to a swimming speed of 21 mm/s, much better than the one without learning.

Journal : Tao Yang et al. A soft artificial muscle driven robot with reinforcement learning, Scientific Reports, 28 September 2018, DOI:10.1038/s41598-018-32757-9



An Evolutionarily-conserved Wnt3/β-catenin/Sp5 Feedback Loop Restricts Head Organizer Activity in Hydra

Capit Udang Secepat Kilat Menciptakan Gelombang Kejut

Quantifying the Holocaust: Hyperintense Kill Rates During the Nazi Genocide

Kecepatan Perjalanan Cahaya Tidak Selamanya Konstan

Superplasticity in An Organic Crystal