Embodied mechanical sensing through self-generated motion in climbing plants

Abstract

Embodied mechanical intelligence and morphological computation enable adaptive behavior by shifting computation from centralized control to body mechanics. Yet how dynamical material structures encode environmental information remains poorly understood. Here, we address this question using climbing plants as a minimal decentralized system. Combining force measurements with physical modeling, we show that circumnutating shoots generate predictable forces against candidate supports, and that the interaction is equivalent to a rotating cantilever governed by torque balance. Force amplitude is determined by stem stiffness and geometry, while the timescale is set by circumnutation rate. Twining occurs only after reaching a critical torque threshold, corresponding to a deformation that reflects support stability and geometric reach. Motorized perturbations further show that twining time scales inversely with effective circumnutation rate. Together, these results show how self-generated motion extracts actionable information about the environment, while the tapered geometry and spatially varying mechanics of the shoot encode the decision-making threshold for attachment.