Robot Metabolism

Robots that grow by integrating found or scavenged parts

Living creatures are open systems that sustain themselves by absorbing matter from and releasing waste to their environment. As a result, they grow, self-heal, and self-sustain—abilities that robots struggle to emulate. Prior research has explored self-assembly, self-repair, and self-replication in modular systems using standardized building blocks to form complex robots.

Our process can produce new robot organisms (see photo). The Artificial Metabolism absorbs and integrates material from its environment or from other robots to form organisms that are bigger, faster, and capable of overcoming obstacles. To conduct our experiments, we use our newly developed modular robotic platform. While previous systems were limited to few pre-defined attachment points, each robot link in our system can connect up to nine other robot links on both sides from a wide range of angles.

We demonstrated a triangular robot capable of absorbing a three-pointed star robot, then morphing into a tetrahedron. Further, the experiments in this paper show a tetrahedron that doubles its downhill walking speed after integrating a found robot link as a leg. Our findings expand the field of developmental robotics beyond robot controllers, to encompass physical transformation in a robot’s lifetime.

Particle Robotics

Stochastic and distributed amorphous robots

A soft robot comprised of rigid components

We introduce a new soft robotics concept that employs multiple rudimentary components: particles form amorphous bodies capable of controlled motion. Each particle has a single modifiable degree of freedom: its vibration speed. Multiple particles constrained by a passive, inelastic membrane form soft robots composed of rigid parts (see photo). When half of the particles inside the membrane vibrate faster than the other half, the robot moves. The particles bump into each other inside the membrane, causing directional motion.

We analyze how frequency modulation and particle count affects locomotion direction and magnitude. Environment-responsive particle motion modulation enables decentralized control. In other words, each particle adjusts its vibration speed in response to its distance from a light source, so the robot can move towards or away from the light. Without communicating with each other, the particles achieve coordinated motion. In future research, we will study system resilience and adaptiveness, including deactivated (dead) particles and obstacle navigation. Particle robots offer an alternative soft robotics method by achieving flexibility from rigid components.