A modular robot architecture capable of learning to move and be automatically reconfigured

Tackling the problem of making a modular robot automatically learn the movements necessary to locomote in different environments is not an easy task. The ability of modular robots to have an arbitrary morphology provides an advantage over usual monolithic robots when moving in different environments...

Full description

Autores:: Moreno García, Rodrigo

Tipo de recurso:: Doctoral thesis

Fecha de publicación:: 2019

Institución:: Universidad Nacional de Colombia

Repositorio:: Universidad Nacional de Colombia

Idioma:: spa

Description
Summary:	Tackling the problem of making a modular robot automatically learn the movements necessary to locomote in different environments is not an easy task. The ability of modular robots to have an arbitrary morphology provides an advantage over usual monolithic robots when moving in different environments. However, being able to reconfigure also has its problems. Movement control for reconfigurable robots is difficult to design and implement. Morphology can also influence the sensing capabilities of a modular robot. Only a few studies include sensor information when adjusting or optimizing controllers for modular robots. The main contribution of this work is the development of an architecture that includes a locomotion training framework that enables a modular robot to move in different environments taking into account sensor information. The framework is composed of four main parts: a control strategy, a configurable environment approach, an adaptation mechanism and a new modular robot platform: the EMERGE modular robot. The EMERGE modular robot platform is designed to be easy to be assembled and can be quickly reconfigured thanks to the magnetic connectors present in its modules. This in turn enables an external agent, like a robot manipulator to reconfigure the robot. Results show that well coordinated movements turn out to be very important for controllers using sensors to improve when being adapted. The mechanisms inside the controller, for example, decision structures, also play a major part in allowing a robot to adapt to move in different environments and be improved. Evaluating robots in reality is a very expensive task and differences between simulation and reality also make robots behave very differently. The magnetic connector makes the assembly of an EMERGE morphology easier but hinders the disassembly process.

A modular robot architecture capable of learning to move and be automatically reconfigured

Publicaciones similares