Ultralight artificial muscles for agile soft robots

Soft robots powered by artificial muscles promise safe, adaptable interaction with the world. However, several key challenges remain. One of the most important is how to make them both powerful and lightweight at the same time.

Electrohydraulic actuators already demonstrate impressive performance across multiple metrics, even approaching biological muscle performance in some aspects. They leverage strong electrostatic forces combined with hydraulic amplification. However, their energy density remains significantly lower than that of biological muscles (8 times lower). In practical terms, it means that approximately 240 kg of artificial muscle would be required to replicate a 70 kg human with 30 kg of muscle. Improving this limitation is essential for building agile robotic systems.

From heavier liquids to ultralight gases
 

A key observation is that these artificial muscles rely on liquid components that contribute most of the actuator mass. We therefore asked whether it is possible to maintain performance while removing this weight.

In this work, we explore a simple but powerful idea by introducing a gas phase into electrohydraulic actuators. Conventional designs rely on a solid—liquid architecture, where a liquid dielectric plays a central role in transmitting forces and sustaining high electric fields. However, this liquid can account for more than 90% of the actuator mass.

In the proposed architecture, most of the liquid is replaced with gas, which transmits forces while substantially reducing mass. The key question then becomes whether this replacement can preserve actuator performance.

Advances and trade-offs
 

The answer turns out to be subtle. Replacing liquid with gas dramatically reduces weight, but it also introduces a new limitation. Electrical breakdown in the gas phase, governed by Paschen’s law, sets a fundamental constraint.

This limitation creates a clear design trade-off. Increasing the gas fraction leads to lighter and faster actuators, while excessive gas causes dielectric breakdown and prevents actuation. By combining experiments and modeling, we identify how much gas can be used safely, and show that actuator performance is ultimately bounded by this physical principle.

With this new solid–liquid–gas architecture, we achieve up to nine-fold higher specific energy (51.4 J/kg), up to eleven-fold higher specific power (1600 W/kg), and substantially faster actuation due to reduced inertia. Notably, even ambient air already enables substantial improvements, while tailored gas mixtures (C4F7N/CO2 mixture) with higher dielectric strength can push the performance even further. As a result, the required artificial muscle to match 30 kg of biological muscle performance can be reduced from approximately 240 kg to around 24 kg.

Significance and impact in soft robots

To demonstrate the practical implications of this improvement, we build a jumping soft robot with identical geometry but different internal dielectric architectures.
The robot based on the solid—liquid—gas architecture jumps 60% higher while reaching take-off 32% faster its solid—liquid counterpart. This result highlights how reducing actuator mass directly improves robotic agility.

Beyond this specific system, our work indicates that this architectural design at the material level can strongly influence system-level performance. This approach opens new opportunities for applications where lightweight actuation is crucial, including wearable robotics, prosthetics, and mobile robotic systems. As soft robotics continues to evolve, such material-level innovations may become a key driver of agile and efficient robotic systems.