Proper actuator placement is an important constituent of efficient control. In biological systems, muscle routing has been optimized as part of evolution; analysis of the results, via cadaveric data, has brought many insights into the modeling and simulation of skeletal motion. However, muscle routing geometry is not always available (i.e. for models that have no biological counterpart), nor are biological muscle routings necessarily the most efficient for a specific task. We present a framework that simultaneously optimizes the muscle routing geometry and closed-loop control parameters, applied to bipedal loco-motion for a variety of morphologies. This facilitates the exploration of the interconnections between morphology, muscle routing, motion patterns, and motion energetics. The framework produces robust, real-time simulations of muscle-based locomotion for many morphologies at different speeds, with no reliance on reference motion data.
