ABSTRACT
We present a neural network model inspired from the cockroach Periplaneta Americana for leg control. The model consists of a central rhythmic pattern generator and a sensory-motor network. We show that it exhibits various behavioral properties observed in cockroaches: It produces a continuum o f stable gaits ranging from metachronal to tripod. The changes in the duration of protraction retraction phases for various walking speeds follow closely the behavioral data. It is shown that the network is insensitive to the initial position of the legs, in that it can achieve rapidly a coherent phase relationship regardless of the initial conditions. In addition, this transient property also extends to situations where rapid changes occur in the walking speed: The network can reorganize its phase rapidly during fast accelerations as well as fast decelerations.
