Minimal decentralized control in a Drosophila-inspired walking robot

In bio-inspired robotics, rhythmic coordination of limbs can be obtained either by using Central Pattern Generators (CPGs) without the involvement of sensory signals, or with a massive feedback that, mainly through proprioceptive sensors, allows the emergence of walking gaits. The aim of this paper is to propose a minimal approach to develop a decentralized locomotion controller for a Drosophila-inspired simulated robot, based on spiking neurons, with the addition of an attitude controller to improve locomotion stability. Walking gaits can be successively learned to create a CPG that can be used in absence of external stimuli or in order to build an internal locomotor estimator.