Analysis and application of a displacement CPG-based method on articulated frames

Abstract

The large evolution of robotics in the last 20 years has been developed with the great contribution of new techniques from computational intelligence, inspired in living things. They have changed the design of articulated artificial systems. The Central Pattern Generators were revealed in the 90´s as regulators of autonomous and rhythmic movements on fish, reptiles, birds and mammals. In this work, through recurrent and dynamical neural networks for the simulation and physical assembly of a quadruped robot with three joints per leg, the concept of Central Pattern Generators (CPG) is applied. A distributed autonomous control architecture based on modular and hierarchical CPG is designed and embedded in software systems. Five recurrent neural networks, organized in two layers, are simultaneously managed to generate signals, synchronize and execute the movement of each joint from each leg, and for the total movement production of different gaits. Successful autonomous decision-making results found for different gaits are shown.