Talk abstract: Steve Epstein & Nancy Kopell, Boston University
Central pattern generators (CPGs) are often observed to drive motor rhythms at frequencies that differ from the endogenous CPG frequency, but lie close to a principal resonance of the musculoskeletal plant. Such CPGs are said to exhibit `resonance tuning.' Owing to changes in posture, load, muscle properties, and so on, resonant frequencies can change significantly over time scales as short as a single period. A CPG must therefore be able to adapt its output quickly and robustly in response to sensory feedback if it is to match the resonant frequency of the plant. Previously proposed solutions to this problem are neither quick nor robust. We present a solution in the case of a very simple, four-dimensional `neuromechanical' model, consisting of a relaxation oscillator reciprocally coupled to a linear oscillator. We show how our feedback scheme works and why some other schemes do not.