REVZEN, S.*; KODITSCHEK, D.E.; FULL, R.J.; Univ. of California, Berkeley; Univ. of Pennsylvania, Philadelphia; Univ. of California, Berkeley: Testing Feedforward Control Models In Rapid Running Insects Using Large Perturbations

Sensory feedback dominates limb coordination in walking. By contrast, insects are capable of stable running over rough surfaces with no detectable change in motor output to major leg muscles. Passive, dynamic models suggest self-stabilization. A central pattern generator forcing a mass-spring system can model these observations. RHex, a rapid running, bio-inspired hexapedal robot is a physical realization of these models. The robot can self-stabilize over simple terrain when its leg motors are driven with feedforward �clock� (CPG-like) signals only. However, RHex requires sensory feedback to its clock to attain comparable speeds over complex terrain. With such feedback, traversal of significant obstacles induces phase shifts in motor timing to coordinate legs. We examined high-speed running in cockroaches to determine whether neural feedback shifts leg phase and/or frequency after encountering an obstacle. Cockroaches were tripped with a hip-high hurdle while running at 25 cm/s on a treadmill. High-speed video tracked the body and distal portion of all six legs versus time. Perturbations struck different legs in different locations, but always caused a significant disruption to the mechanical system for the several strides traversing the obstacle. Cockroaches incurred significant phase and frequency changes. When the pattern of leg motions was extrapolated from before the obstacle to after the obstacle, it failed to match the animals’ pattern after recovery in 17 of 22 trials. Results suggest that sensory feedback is likely sent to a CPG-like clock driving leg movements, as required by our physical model. We reject the hypothesis that animals use a feed-forward clock without neural feedback when challenged with large perturbations in rapid running.