Efficient hexapodal locomotion control based on flow-invariant subspaces

In this paper the issue of locomotion control in bio-inspired hexapod structures is considered as a problem of convergence toward flow invariant subspaces in networks of mutually and locally coupled neural units. Since the network topologies used refer to undirected diffusive tree graphs, in this case a unique gain value on the graph connection matrix can be found to guarantee exponential convergence to any desired gait (i.e. phase locked) configuration. Such sufficient condition is exploited for the locomotion control in a hexapod structure. In the paper, after dealing with the sufficient condition above mentioned, both simulation and experimental results on a robotic hexapod structure are reported. Moreover the role of key parameters for insect inspired turning strategies is investigated.