Mechanically Intelligent Membrane-bound Active Ensemble

Designing collective systems that leverage mechanical intelligence over centralized computation-heavy control facilitates adaptive and emergent behaviors. In previous work, we studied a gas-like collective of puck-shaped robots (BOBbots) that used magnetic attraction to transition between dispersed and aggregated states, enabling the group to transport objects heavier than individuals. Here, we remove the need for such attraction by enclosing the BOBbots within a flexible, length-adjustable belt, creating a membrane-bound 2D active collective system (“Membot”). Each wheel-driven BOBbot carries a microcontroller and phototransistors that induce photophobic run-and-tumble motion in response to light from a central controller attached to the belt, whose statistics probabilistically govern the mean orientation. The membrane length sets density and morphology: shortening the belt yields a solid-like, jammed state that exerts collective pressure, whereas lengthening produces a fluid-like phase that freely reconfigures. Experiments and simulations show that an 8% increase in belt length from the jammed state triggers the solid-to-fluid transition. Cyclic modulation of belt length further allows the collective to traverse obstacle-rich terrains, pushing obstacles aside when compact and flowing through narrow gaps when expanded. These results demonstrate that effective mechanical work can emerge from morphology and local interactions, with minimal centralized control.