Bio-mimetic machine learning based on compound control

Abstract

Tacit learning is a new machine learning paradigm that attempts to implement the superb adaptation capability of living organisms to unexpected environmental changes. It emphasizes body/environment interactions and is equipped with some elementary sets of action rules and appropriate initial conditions of the neural states that correspond to elementary survival reflexes. Along this line, we propose a new scheme of neural computation based on compound control which represents a typical feature of biological controls. This scheme is based on a classical neuron model where macroscopic purposeful behavior emerges as the result of the interaction of local rules. This scheme is applied to a bipedal robot and generates the rhythm of walking without any model of robot dynamics and environments.