Learning scheme of multiple-patterns in quadruped locomotion using CPG model

Quadrupeds show several locomotion motion patterns adapting to environmental conditions. An immediate transition among walk, trot, and gallop implies an existence of the memory for locomotion patterns. In this paper, we postulate that the motor patter learning necessitates the repetitive presentation of the same environmental conditions, and aim at constructing a mathematical model for new pattern learning. The model construction deals with a decerebrate cat experiment where only the left forelimb is driven at the higher speed by the belt on the treadmill. A CPG model that adaptively generates locomotion pattern and qualitatively describes the decerebrate cat behavior has already proposed. Developing this model, we introduce a memory to retain locomotion patterns. Here, the memory is represented as the minimal point of the potential function whose gradient system describes recollecting process, and new minimal point is generated by the bifurcation from already-existed minimal point. The process where two minimal points are generated based on the repetitive presentation of the same environmental condition is described.