Publicação

Endowing intelligent vehicles with the ability to learn user’s habits and preferences with machine learning methods

Ver documento

Detalhes bibliográficos
Resumo:A private vehicle frequently carries the same passengers who routinely take specific objects with them, have their vehicle comfort preferences, and visit the same places at relatively the same time of a given day or day of the week. Thus, developing intelligent vehicles that are able to reduce the cognitive workload of the drivers by learning and adapting to their occupants’ routines is of the highest interest. In this paper, we present two independent models based on machine learning methods, including artificial neural networks and linear and ridge regressions, to learn the habits and preferences of the vehicle’s users. The first model is responsible for predicting the next vehicle trip state, i.e., the departure location and time, and the driver, passenger, and object states. The second model anticipates the comfort setting inside the cockpit - temperature, cockpit mirror, and driver seat poses. The developed models were trained, evaluated, and validated with different datasets in the Portuguese city of Braga. The results prove that the vehicle efficiently learns the routines of several users with varying complexities. Prediction errors happen in cases of an exceptional, one-time deviation from routine behavior.
Autores principais:Barbosa, Paulo
Outros Autores:Ferreira, Flora; Fernandes, Carlos; Erlhagen, Wolfram; Guimarães, Pedro; Wojtak, Weronika; Monteiro, Sérgio; Bicho, Estela
Assunto:Human-centered computing Machine learning Driver routines Learning driver habits Human mobility patterns Time and space prediction Deep learning Intelligent vehicles
Ano:2022
País:Portugal
Tipo de documento:comunicação em conferência
Tipo de acesso:acesso aberto
Instituição associada:Universidade do Minho
Idioma:inglês
Origem:RepositóriUM - Universidade do Minho

Registos relacionados

Registos relacionados