‘Alien-like’ AI robot can evolve and repair itself after being chopped into pieces

Engineers in the United States have created AI-powered robots that can evolve and reconfigure themselves, as well as recover from damage.

The alien-like machines, described as ‘legged metamachines’, are built from autonomous modules that can be snapped together in a variety of forms. Researchers say the approach could lead to robots that are more resilient and adaptable in real-world environments.

The work was carried out by engineers at Northwestern University and published on Thursday in the journal Proceedings of the National Academy of Sciences.

Each individual module functions as a small robot in its own right, equipped with a motor, battery and computer. On its own, a module can roll, turn and jump. However, when several are combined they form larger robotic systems capable of far more agile movement.

Researchers used artificial intelligence to design the most effective combinations. Rather than copying familiar animal or human forms, the AI generated unusual body shapes that human engineers might not have considered.

The resulting configurations allowed the machines to move in striking ways. Depending on how the modules were assembled, the robots could undulate like seals, bound like lizards or spring like kangaroos.

The metamachines were also able to flip themselves upright after being overturned, hop over obstacles and perform acrobatic manoeuvres such as spinning in mid-air. Because each system is made up of multiple self-contained robots, damaged parts can separate and continue operating instead of becoming dead weight.

Researchers say this combination of modular hardware and AI-driven design could point towards a new class of adaptable robots that are able to cope with unpredictable environments.

‘These are the first robots to set foot outdoors after evolving inside of a computer,’ said Sam Kriegman, who led the study.

‘They are rapidly assembled and then quite literally hit the ground running. They can move freely in the wild and easily recover from major injuries that would be fatal to every other wild robot. If flipped upside down, they instinctively bring themselves upright and continue their journey.

‘They can survive being chopped in half or cut up into many pieces. When separated, every module within the metamachine can become an individual agent.’

Dr Kriegman, an expert in biorobotics and AI, is an assistant professor of computer science, mechanical engineering and chemical and biological engineering at Northwestern University McCormick School of Engineering and a member of the Center for Robotics and Biosystems.

While many modern robots are capable of fast and agile movement, their physical design is often fixed. If a component fails – such as a leg on a robotic dog – the machine may no longer function.

To overcome this limitation, Dr Kriegman’s team used an evolutionary algorithm designed to mimic natural selection.

The process began with simple building blocks: half-metre-long modular legs resembling two sticks connected by a central sphere.

‘Inside the sphere, the robot has everything it needs to survive: a ‘nervous system,’ a ‘metabolism’ and ‘muscle,’ Kriegman said. ‘By that, I mean a circuit board, a battery and a motor. The modules are mechanically simple. They can only rotate around a single axis, but they are surprisingly athletic and smart.’

The algorithm was then tasked with designing robots capable of efficient and versatile movement. It produced numerous body configurations, simulated their performance and kept the most successful designs while discarding weaker ones.

Over successive iterations, designs were combined and modified, with modular legs evolving to function as legs, spines or tails depending on the overall body structure.

‘We simulated the Darwinian process of mutation and selection within a virtual, physical environment,’ Kriegman said. ‘This is survival of the fittest — accelerated by computers and made real by athletic modular building blocks.’

To test the approach, the team built several of the most successful designs, including three-, four- and five-legged robots.

In outdoor trials, the metamachines moved across rough terrain including gravel, grass, tree roots, leaves, sand, mud and uneven brick surfaces. The robots were able to jump, spin and right themselves after being flipped over without requiring additional programming or retraining.

Unlike conventional robots that fail when a single component breaks, the modular systems continued to function even after losing a limb. Detached modules could roll back and reconnect with the rest of the machine.

‘It can sense its surroundings, move from place to place, compute and learn,’ Kriegman said.

‘Metamachines can be rapidly assembled, repaired, redesigned and recombined. Once assembled, they immediately move themselves across a wide array of unstructured environments.’

The research builds on earlier work from Dr Kriegman’s laboratory in which an AI system was used to design simple robots from scratch in a matter of seconds.

‘Our previously evolved robots couldn’t sense their own bodies or coordinate themselves,’ Kriegman said ‘But they still taught us a lot about how evolution works and how to distill those lessons into useful technologies.

‘Evolution can reveal new designs that are different from or even beyond what humans were previously capable of imagining. So, we really wanted to study how and why it works. The best way – or at least the most fun way – is to evolve structures in realistic conditions.’

The study, titled Agile legged locomotion in reconfigurable modular robots, was supported by Schmidt Sciences and the National Science Foundation.