Desenvolvimento de um sensor virtual para a velocidade longitudinal de um veículo

• Autor(a) principal: Braz, Pedro Henrique de Araujo
• Data de Publicação: 2018
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Institucional da UFLA
• Texto Completo: http://repositorio.ufla.br/jspui/handle/1/29553
• topic: Veículos inteligentes; Sensores virtuais; Modelos NARX; Intelligent vehicles; Virtual sensors; NARX models; Sistemas de Computação

The worldwide increase in vehicle circulation has led to problems with traffic, environmental pollution and safety, which motivated researchs into the development of better vehicles for the user and the surroundings. One of the ways to overcome these problems was to use electromechanical control systems, such as Advanced Driver Assistance Systems (ADAS) and Active Stability Control (ASC) systems. These systems have assisted the development of intelligent vehicles, considering that they need access to the magnitudes of the environment and the dynamics of the vehicle by means of reliable sensors; physical sensors, however, are susceptible to problems, such as: measurement errors, availability, reliability, measurement delays and high cost. An alternative without high economic costs, to overcome these problems, is the operation of soft sensors or virtual sensors. This work describes the process of identifying a virtual sensor, which is capable of estimating the longitudinal velocity of an intelligent vehicle by means of low cost physical sensors, such as the accelerometer of a smartphone. Data collection was performed regarding the dynamic behavior of the vehicle, using an OBD-II (On-Board Diagnostic) device and a smartphone. In data collection the vehicle’s inclination and mass were varied between the essays, in order to obtain information of the numerous situations in which a vehicle normally travels. From the collected statistics, polynomial NARX (Non-linear AutoRegressive with eXogenous inputs) models, with parameters obtained by least squares (LS) estimator and regressors chosen using the error reduction rate (ERR), were identified. Finally the attainment of robust models to the track’s inclination variations and mass of the vehicle were implemented and analyzed. These models are said to be robust because they describe the longitudinal velocity of the vehicle during the intervals of these variations. The obtained results were satisfactory as to the robustness of the mass variations of the vehicle, althouhgh better identification is needed in order to achieve consistency regarding the slope of the road. This problem can be circumvented by a project that applies a combination of models to be developed in future works.