New AI helps soft robot arms learn new tasks without retraining

Soft robot arms can now learn new tasks on the fly, unlike before. This AI system works on different types of soft robots, making them more useful for jobs.
By Newsroom | Technology, Robotics |

Soft Arms Learn to Adapt, Bridging a Gap in Control

A novel artificial intelligence control system has been developed, allowing soft robotic arms to master a range of movements and tasks. Crucially, these systems can then adapt to new situations in real-time without needing to be retrained, maintaining their operational capacity. This development addresses a long-standing hurdle in controlling soft robots, which, unlike their rigid counterparts, are constructed from yielding materials like soft rubber and employ artificial muscles for motion. Their inherent flexibility, while beneficial for delicate or adaptable operations, typically leads to unpredictable shape changes that complicate control.

The system's efficacy has been demonstrated on two distinct physical soft arm setups: one driven by cables and another using shape-memory-alloy actuators. The AI approach has shown cross-platform applicability, meaning it can be implemented across different soft-arm designs. This move toward more adaptable robotic forms, however, raises questions about the increasing autonomy and unforeseen consequences of systems designed to learn and adjust independently.

Read More: Linux users struggle to get older AMD GPUs working with new drivers

Insect-like Bots Gain Agility, Hinting at Mechanical Labor

Streamlined Design Enhances Performance

In parallel, a redesigned robotic insect, boasting increased flight precision and agility, has emerged. This new iteration features a split design, essentially dividing the robot into two halves, and a novel long wing hinge mechanism. This hinge works to lessen the mechanical strain on the artificial wing flexures during their flapping motion. Such modifications allow for quicker maneuvers, greater endurance, and an extended operational lifespan for the robot.

The improved performance per unit is a significant outcome, addressing previous limitations where individual components outperformed the assembled robot. The reduction in wing count, by half, also strategically frees up internal space, potentially accommodating more sophisticated electronics. This development in nimble, energy-efficient robotics, while presented as a step towards tasks like mechanical pollination, opens avenues for diverse, potentially disruptive, applications.

Read More: China's Unitree Robotics plans to make 20,000 humanoid robots by late 2025

Frequently Asked Questions

Q: What new technology allows soft robot arms to learn tasks?
A new artificial intelligence control system has been made. It helps soft robotic arms learn many different movements and jobs. They can also change to new situations quickly without needing to be taught again.
Q: Why is it hard to control soft robots before this new AI?
Soft robots are made of soft things like rubber and move with artificial muscles. Their soft bodies change shape in ways that are hard to control. This makes them unpredictable compared to hard robots.
Q: Can this new AI work on different types of soft robot arms?
Yes, the AI can be used on different soft robot arms. It has been tested on arms that use cables and on arms that use special metal that changes shape with heat. This shows it can work on different designs.
Q: What does the development of adaptable robot arms mean for the future?
This new AI makes robots more able to learn and change on their own. This could lead to more robots doing different jobs. It also brings up questions about how these smart robots will work and what might happen next.