Localización de robots móviles en entornos cerrados mediante características de audio

When GPS is not operative in closed environments, mobile robots use different types of sensor in order to locate their position. However, the use of audiofrequency signals sensors is significative lower than other type of sensors. Audio characteristics present interesting properties and robots can o...

Descripción completa

Detalles Bibliográficos
Autor: Manzanares Brotons, Manuel|||0000-0002-2852-2500
Tipo de recurso: tesis doctoral
Fecha de publicación:2016
País:España
Institución:Universitat Politècnica de Catalunya (UPC)
Repositorio:UPCommons. Portal del coneixement obert de la UPC
Idioma:español
OAI Identifier:oai:upcommons.upc.edu:2117/96256
Acceso en línea:https://hdl.handle.net/2117/96256
https://dx.doi.org/10.5821/dissertation-2117-96256
Access Level:acceso abierto
Palabra clave:Control automàtic
Robots -- Moviment
Àrees temàtiques de la UPC::Informàtica
Descripción
Sumario:When GPS is not operative in closed environments, mobile robots use different types of sensor in order to locate their position. However, the use of audiofrequency signals sensors is significative lower than other type of sensors. Audio characteristics present interesting properties and robots can obtain great advantages if there are able to process acoustics signals. The main objective of this PhD tesis is to propose and investigate acoustic models and localization systems based in the treatment of audio signals thought for mobile robots self-localization in closed enviroments.This work will help to cover a great gap in this field. In this thesis, a first experimental approach is presented for self-location systems based in proposed models to test their validity by real tests. In this work also new models with simplified hypothesis are proposed which are specially conceived for their application in location systems. First a theoretical model, based in the physical description of sound propagation is proposed, allowing a robot to locate determining the magnitude of gains for different transfer function obtained in different robot location inside an environment. Second, parametrical identification models have been also proposed which allow a robot to locate by means of a system based in the observation of the parameters variations of the models in the different robot locations and the subsequent application of a transformation function. In this research, also an audio signal filtering systems is proposed, near the Schroeder frequency that allow reducing the complexity of the proposed models. The uncertainty of the models is also studied and treated in basis of supplied information by signals acquired in a small set of labeled sampled in the working robot environment that allow the possibility of SLAM techniques usage. The validity of researched location systems is checked with the realization of experimental tests where a comparative study among them is carried out. Strong and weak points in location systems are studied. The outcome of such a study show similar results for proposed systems, but it enhances the lesser degree of uncertainty in the results yielded by the location systems based in the theoretical model than those systems based in parametric identification models. Other relevant conclusions in this research work are, apart from the validation of presented location systems, the need to detemine the orientation angle of the robot respect a known reference point due the high directionality present in the sound pressure wave that is propagated in closed environments. It is also remarkable the excellent results of the proposed functions that allow to obtaining the robot position in a closed environment through the processing of different acoustic signals features acquired by the robot. Another important conclusion to remark for the theoretical model definition is the determination of some mathematical expressions that relate the acoustic response of the environments (obtained from the Sabine's statistical model) with some parameters of such theoretical model making more practicable their application.