Toward Mechanical Intelligence in Robotics

Abstract

Conventional robotic systems concentrate sensing, control, and actuation in electronic hardware, creating constraints in harsh environments, energy efficiency, and miniaturization. The physical body of a robot is more than a passive executor of control commands; it is an active substrate that transforms inputs into functional operations through its intrinsic mechanical dynamics. This perspective describes this capability as mechanical intelligence and identifies three core operations: filtering, routing, and switching. These operations address three fundamental questions in physically mediated transformation: what inputs are selectively admitted, where signals propagate, and when the system transitions between states. Together, they enable selective responsiveness, spatial coordination, and state‐dependent behavior without dedicated electronic hardware. Design variables and quantitative metrics are proposed for each operation, and the framework is illustrated in locomotion and manipulation, providing guidance for designing mechanically intelligent robots. Responsive materials extend this framework by transducing environmental stimuli into mechanical signals while also harvesting ambient energy for actuation. Rather than replacing digital computation, mechanical intelligence complements it by embedding part of sensing, processing, and actuation into the physical body, pointing toward hybrid systems in which mechanical and digital intelligence work together.