Evolving Neural Controllers for Collective Robotic Inspection

Alcherio Martinoli
2006
Article de conférence

Résumé

In this paper, an automatic synthesis methodology based on evolutionary computation is applied to evolve neural controllers for a homogeneous team of miniature autonomous mobile robots. Both feed-forward and recurrent neural networks can be evolved with fixed or variable network topologies. The efficacy of the evolutionary methodology is demonstrated in the framework of a realistic case study on collective robotic inspection of regular structures, where the robots are only equipped with limited local on-board sensing and actuating capabilities. The neural controller solutions generated during evolutions are evaluated in a sensorbased embodied simulation environment with realistic noise. It is shown that the evolutionary algorithms are able to successfully synthesize a variety of novel neural controllers that could achieve performances comparable to a carefully hand-tuned, rule-based controller in terms of both average performance and robustness to noise.

Official source

https://infoscience.epfl.ch/record/83630?ln=fr

À propos de ce résultat

Cette page est générée automatiquement et peut contenir des informations qui ne sont pas correctes, complètes, à jour ou pertinentes par rapport à votre recherche. Il en va de même pour toutes les autres pages de ce site. Veillez à vérifier les informations auprès des sources officielles de l'EPFL.

Concepts associés

Chargement

Publications associées

Chargement

This paper addresses the problem of cooperative transport of an object by a group of two simple autonomous mobile robots called s-bots. S-bots are able to establish physical connections between each other and with an object called the prey. The environment consists of a flat ground, the prey, and a light-emitting beacon. The task is to move the prey as far as possible in an arbitrary direction by pulling and/or pushing it. The object cannot be moved without cooperation. There is no explicit communication among s-bots; moreover, the s-bots cannot sense each other. All experiments are carried out using a 3D physics simulator. The s-bots are controlled by recurrent neural networks that are created by an evolutionary algorithm. Evolved solutions attained a satisfactory level of performance and some of the exhibit remarkably low fluctuations under different conditions. Many solutions found can be applied to larger group sizes, making it possible to move bigger objects.

Springer Verlag2004

Chargement

Towards Group Transport by Swarms of Robots

Roderich Gross

We examine the ability of a swarm robotic system to transport cooperatively objects of different shapes and sizes. We simulate a group of autonomous mobile robots that can physically connect to each other and to the transported object. Controllers — artificial neural networks — are synthesised by an evolutionary algorithm. They are trained to let the robots self-assemble, that is, organise into collective physical structures, and transport the object towards a target location. We quantify the performance and the behaviour of the group. We show that the group can cope fairly well with objects of different geometries as well as with sudden changes in the target location. Moreover, we show that larger groups, which are made of up to 16 robots, make possible the transport of heavier objects. Finally, we discuss the limitations of the system in terms of task complexity, scalability, and fault tolerance, and point out potential directions for future research.

2009

Chargement

In this paper we describe the evolution of a discrete-time recurrent neural network to control a real mobile robot. In all our experiments the evolutionary procedure is carried out entirely on the physical robot without human intervention. We show that the autonomous development of a set of behaviors for locating a battery charger and periodically returning to it can be achieved by lifting constraints in the design of the robot/environment interactions that were employed in a preliminary experiment. The emergent homing behavior is based on the autonomous development of an internal neural topographic map (which is not pre-designed) that allows the robot to choose the appropriate trajectory as function of location and remaining energy.

1996

Chargement