CPG network to generate the swimming motion of the crawl stroke

The objective of this study was to propose a CPG network which can generate the swimming motion of the crawl stroke. First, the CPG network for legs performing a flutter kick was constructed by connecting the neural oscillators for the leg joints. The flutter kick motion was successfully generated by the proposed CPG network. The propulsion by the generated flutter kick motion was confirmed by the simulation of the swimming movement. Next, the CPG network for both the arms and legs were constructed, in which the neural oscillator for the arms initiated the trigger signal to start the prescribed stroke motion. By changing the intrinsic cycle of the neural oscillators for the legs, both six- and two-beat crawls could be realized. It was also found that a stable region with respect to the relationship between the intrinsic cycles of the neural oscillators for the arms and legs certainly existed for the six-beat crawl, although the intrinsic cycles of the arms were three times longer than those of the legs in this case. The propulsion by the generated swimming motion was confirmed by the simulation of the swimming movement both for the six- and two-beat crawls. Finally, the roll angle of the swimmer was fed back into the CPG network in order to restore the balance in the roll direction. Restoring the balance in the roll direction was successfully realized by the proposed feedback algorithm. The resultant motion showed a complicated behavior, such as skipping strokes.