Genetic algorithm-based optimal design for fluidic artificial muscle (FAM) bundles

In this paper, we present a design optimization framework for a fluidic artificial muscle (FAM) bundle subject to geometric constraints. The architecture of natural skeletal muscles allows for compact actuation packaging by distributing a substantial number of actuation elements or muscle fiber motor units, which are to be arranged, oriented, and sized in various formations. Many researchers have drawn inspiration from these natural muscle architectures to assist in designing soft robotic systems safe for human-robot interaction. Although there are known tradeoffs identified between different muscle architectures, this optimization framework offers a method to map these architectural tradeoffs to soft actuator designs. The actuation elements selected for this study are FAMs or McKibben muscles due to their inherent compliance, cheap construction, high force-to-weight ratio, and muscle-like force-contraction behavior. Preceding studies analytically modeled the behavior of arranging FAMs in parallel, asymmetrical unipennate, and symmetrical bipennate topologies inspired by the fiber architectures found in human muscle tissues. A more recent study examined the implications of spatial constraints on bipennate FAM bundle actuation and found that by careful design, a bipennate FAM bundle can produce more force, contraction, stiffness, and work output than that of a parallel FAM bundle under equivalent spatial bounds. This multi-objective genetic algorithm-based optimization framework is used to realize desirable topological properties of a FAM bundle for maximum force and stroke for a given spatial envelope. The results help identify tradeoffs to inform design decisions based on the force and stroke demand from the desired operating task. This study further demonstrates how the desirable topological properties of the optimized FAM bundle change with different spatial bounds.