Abstract:With the accelerated development of robot technologies, control becomes one of the central themes of research. In traditional approaches, the controller, by its internal functionality, finds appropriate actions on the basis of specific objectives for the task at hand. While very successful in many applications, self-organized control schemes seem to be favored in large complex systems with unknown dynamics or which are difficult to model. Reasons are the expected scalability, robustness, and resilience of self-organizing systems. The paper presents a self-learning neurocontroller based on extrinsic differential plasticity introduced recently, applying it to an anthropomorphic musculoskeletal robot arm with attached objects of unknown physical dynamics. The central finding of the paper is the following effect: by the mere feedback through the internal dynamics of the object, the robot is learning to relate each of the objects with a very specific sensorimotor pattern. Specifically, an attached pendulum pilots the arm into a circular motion, a half-filled bottle produces axis oriented shaking behavior, a wheel is getting rotated, and wiping patterns emerge automatically in a table-plus-brush setting. By these object-specific dynamical patterns, the robot may be said to recognize the object's identity, or in other words, it discovers dynamical affordances of objects. Furthermore, when including hand coordinates obtained from a camera, a dedicated hand-eye coordination self-organizes spontaneously. These phenomena are discussed from a specific dynamical system perspective. Central is the dedicated working regime at the border to instability with its potentially infinite reservoir of (limit cycle) attractors "waiting" to be excited. Besides converging toward one of these attractors, variate behavior is also arising from a self-induced attractor morphing driven by the learning rule. We claim that experimental investigations with this anthropomorphic, self-learning robot not only generate interesting and potentially useful behaviors, but may also help to better understand what subjective human muscle feelings are, how they can be rooted in sensorimotor patterns, and how these concepts may feed back on robotics.

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

Self-Organized Behavior Generation for Musculoskeletal Robots

Neural self-organization for muscle-driven robots

Closed-loop robots driven by short-term synaptic plasticity: Emergent explorative vs. limit-cycle locomotion

The Sensorimotor Loop as a Dynamical System: How Regular Motion Primitives May Emerge from Self-Organized Limit Cycles

Embodied robots driven by self-organized environmental feedback

Self-organized control for musculoskeletal robots

The Strange Attractor Model of Bipedal Locomotion and its Consequences on Motor Control

Physical synchronization of soft self-oscillating limbs for fast and autonomous locomotion

Surprising simplicities and syntheses in limbless self-propulsion in sand

Self-organized adaptation of a simple neural circuit enables complex robot behaviour

Learning-based Hierarchical Control: Emulating the Central Nervous System for Bio-Inspired Legged Robot Locomotion

Mechanical Intelligence Simplifies Control in Terrestrial Limbless Locomotion

Morphological Properties of Mass-Spring Networks for Optimal Locomotion Learning

Self-organizing Continuous Attractor Networks and Motor Function

Kinetic energy fluctuation-driven locomotor transitions on potential energy landscapes of beam obstacle traversal and self-righting

Intrinsically motivated collective motion

When the goal is to generate a series of activities: A self-organized simulated robot arm

Excitable crawling

Moving in an Uncertain World: Robust and Adaptive Control of Locomotion from Organisms to Machine Intelligence