Attractoring-based locomotion*

Locomotion is the result of the complex interplay of the controlling nervous system, the body, and the environment. From a dynamical systems point of view, however, regular locomotion corresponds to limit cycles formed in the phase space of variables describing the robot's or animal's dynamical states. These limit cycles may be generated either centrally by neural circuits called central pattern generators (CPG) or peripherally via closing the local somatosensory loop.

Here, we introduce the concept of attractoring, a control type for which oscillatory dynamics is generated via the somatosensory feedback loop even without CPGs. Based on the conditions under which limit cycles can emerge in the system, we propose a classification of control types. To illustrate these control schemes, we present a simple model of an actuator controlled by a rate encoding neuron with internal adaption and with inputs representing self-coupling, respectively delayed proprioceptive feedback. Within this control framework, rolling, wheeled, and legged robots interact with their environment in a self-organized manner.