Self-Exploration of Autonomous Robots Using Attractor-Based Behavior Control and ABC-Learning
Manfred Hild, Matthias Kubisch · Frontiers in artificial intelligence and applications · 2011
An autonomous robot which is equipped with sensorimotor loops and situated within a specific environment can be regarded as a dynamical system. By using Attractor-Based Behavior Control (ABC), the attractors of this dynamical system correspond to energy-efficient behavioral body postures, and the attractor-connecting heteroclinic orbits can be utilized to generate stable motion trajectories. We introduce ABC-Learning and demonstrate how it enables an autonomous robot to self-explore its behavioral capabilities from scratch and without any given body model.