Robotics borrows a strategy from skin cells
When you cut your finger, skin cells rush to the wound to close it. By aggregating and moving collectively, the cells travel faster and with more adaptability. Conventional designs for robots, in which individual components are under centralized control, are less flexible and have a higher probability that failure of a given component will debilitate the entire robot.
Now Hod Lipson of Columbia University and his colleagues have drawn inspiration from collective cell migration to build a robot composed of parts governed by statistical mechanics. The robot is made of individual disk-shaped units, shown in the figure and video, that expand and contract between 16 cm and 24 cm in diameter, but they are incapable of motion on their own. With precisely timed successive waves of expansions and contractions, a collection of three or more units can push and pull one another across a smooth surface. That strategy works when one unit’s outward force during expansion and inward force during contraction can overcome its static friction without budging the other two or more units. The inward force that couples the components is provided by magnets along their edges.
The direction of motion is set by the direction of the waves of expansion and contraction. One mode of operation is for each unit to set its expansion phase based on a signal’s intensity—light, in the current case—measured using an array of photocells. With that setting, the robot seeks out the source of the signal. Robotic swarms with up to two dozen units demonstrated that they can also carry objects and maneuver around obstacles. Simulations matched the experiments and gave the researchers confidence that robots with up to 10 000 units would behave in the same ways. Additionally, simulated robots demonstrated robustness to nonfunctional units: Even when up to 20% of the units are dead, the robot still moves at about half its full speed. (See the video for a demonstration.)
The experimenters’ next step is to use smaller unit sizes in the microscale and nanoscale. At those length scales, controlling components individually is tricky. An amorphous robot made of stochastic components, such as Lipson and his colleagues present, would not require direct control of units. (S. Li et al., Nature 567, 361, 2019 .)
