Scientists build metamaterial that can ‘learn’ to change shape, PgII

Key Highlights:

• European researchers have created a metamaterial that can physically learn and adapt its shape based on external conditions.
• The robotic metamaterial consists of connected units, each with a motor, angle sensor, and microcontroller.
• The material uses contrastive learning to 'learn' specific shapes by comparing free and clamped states.

Detailed Insights:

• Metamaterials are engineered materials whose properties are determined by their structure, not just their composition, allowing for unique functionalities.
• The contrastive learning method involves the metamaterial chain transitioning from a free state to a clamped state, adjusting its stiffness based on the difference between the two states.
• The ability of the metamaterial to exhibit non-reciprocity challenges traditional physics concepts and expands the understanding of 'intelligence' in materials.
• The local decision-making process of the chain, where each unit learns based on its neighbors, eliminates the need for complex data transfer networks.
• The bistable units enable the metamaterial to perform life-like actions, such as gripping objects, by navigating through different stable states.
• Potential future applications of such metamaterial chains include advanced prosthetic limbs and soft robots capable of agile responses to obstacles.
• The research suggests that metamaterials can be an exciting platform for physical learning, opening avenues for designing adaptive materials and robots.

Scientific/Technical Concepts Involved:

• Metamaterials: Materials engineered to have properties not found in nature, often due to their structure.
• Contrastive Learning: A learning method where a system learns by comparing two different states or inputs.
• Non-reciprocity: A property where the response of a system depends on the direction of the input.
• Bistability: The ability of a system to exist in two stable states.