Northwestern University researchers have introduced a groundbreaking development involving modular robots combined with artificial intelligence (AI). The concept behind these robots, named “legged metamachines,” is their capability to be reassembled in the outdoors, recover from damage, and keep functioning under various conditions. These robots are assembled from autonomous, Lego-like modules that interlock in different configurations. Each module acts as a complete robot on its own with a motor, battery, and computer, granting it the ability to roll, turn, and jump.
Unlike traditional robots with fixed body shapes, the engineers decided to leverage AI to generate unique body configurations. Their rationale was that while modern robots may possess speed and agility, their structures are often inflexible, hindering their adaptability to new tasks, environments, or physical harm. For instance, if a robotic dog sustains a broken leg, the entire robot typically becomes inoperative. To address this limitation, the team led by Kriegman employed an evolutionary algorithm mimicking natural selection, with modular legs half a meter in length connected by a central sphere serving as robotic building blocks. Inside the sphere, each module contains essential components such as a circuit board, battery, and motor, functioning collectively as the robot’s ‘nervous system,’ ‘metabolism,’ and ‘muscle.’
The algorithm’s objective was to design a robot with efficient and versatile movement by mixing and matching modules in varied configurations. As it created new body designs, the algorithm simulated the actions and performance of each model, retaining the most successful designs while discarding the less effective ones. Subsequently, superior designs were either combined or modified, allowing modular legs to transform into legs, spines, or tails. To evaluate these designs, Kriegman and his team assembled the most promising three-, four-, and five-legged robots discovered through the evolutionary process.
During outdoor trials, these metamachines successfully navigated challenging terrain like gravel, grass, tree roots, and mud, showcasing abilities to jump, spin, and self-right if overturned, without requiring complex adjustments or retraining. Their key advantage over traditional robots lies in their adaptability, resilience, and survivability. Even in scenarios where a leg breaks off, the metamachine can adapt to the loss and continue moving, while the detached leg can autonomously return and rejoin the group.
These reconfigurable robots exhibit remarkable capabilities such as self-recovery and seamless adaptation to challenges, making them highly suitable for tasks in unpredictable environments. The study detailing the innovative modular robots was published in the Proceedings of the National Academy of Sciences, signifying a significant advancement in robotics technology that leverages AI for enhanced functionality and adaptability.
About the Author
