Self-organized attractoring in locomoting animals and robots: an emerging field

TL;DR

This paper reviews how self-organized attractors in sensorimotor loops enable robust, compliant locomotion in animals and robots through emergent control principles based on limit cycles and chaotic attractors.

Contribution

It introduces the concept of attractoring as a unifying principle for self-organized embodiment and demonstrates its effectiveness across various robotic platforms.

Findings

Self-organized attractors serve as motor primitives for higher-level control.

A simple generative principle ensures robust self-organization in locomotion.

Applicable to diverse robots, including wheeled, legged, and shape-changing agents.

Abstract

Locomotion may be induced on three levels. On a classical level, actuators and limbs follow the sequence of open-loop top-down control signals they receive. Limbs may move alternatively on their own, which implies that interlimb coordination must be mediated either by the body or via decentralized inter-limb signaling. In this case, when embodiment is present, two types of controllers are conceivable for the actuators of the limbs, local pacemaker circuits and control principles based on self-organized embodiment. The latter, self-organized control, is based on limit cycles and chaotic attractors that emerge within the feedback loop composed of controller, body, and environment. For this to happen, the sensorimotor loop must be locally closed, e.g. via propriosensation. Here we review the progress made within the framework of self-organized embodiment, with a particular focus on the concept of attractoring. This concept characterizes situations when sets of attractors combining discrete and continuous spectra are available as motor primitives for higher-order control schemes, such as kick control. In particular, we show that a simple generative principle allows for the robust formulation of self-organized embodiment. Based on the recurrent alternation between measuring the actual status of an actuator and providing a target for the actuator to achieve in the next step, we find that the mechanism leads to compliant locomotion for a range of simulated and real-world robots, which include barrel- and sphere-shaped agents, as well as wheeled and legged robots.