Finding tradeoffs in muscle types through a robophysical model of the Hill muscle

Muscles are biology's actuators: muscle's ability to generated force and do work ultimately allows organisms to navigate their environments and search for food. Studying the mechanical state properties (i.e., force generation, rate of contraction, power output) of muscle and different muscle types, therefore, is critical in understanding how life persists. Studying muscle in-vivo, however, remains a long-standing challenge: real-time data collection of muscle's mechanical state properties is incredibly invasive and difficult to do. Here, using feedback control and a DC motor as our actuator, we create a robophysical model of the Hill muscle. The Hill muscle describes the nonlinear relationship between a muscle's contraction rate and force generation where the degree of nonlinearity is parametrized by a nondimensional α. With our muscle-mimicking actuator, we can measure real-time mechanical state properties and the energetic inputs required for actuation. Moreover, we can systematically test an array of different muscle types parameterized by α and assign an efficiency of actuation. We establish a scalar power characteristic defined as the area under an actuator's force-velocity curve: it is a property of all actuators and indicates an actuator's ability to do work. Using the power characteristic as a basis, we find a tradeoff in muscle types: muscles that are nearly linear in force-velocity space are optimized for power output whereas highly nonlinear muscles are optimized for energy efficiency.