Modelagem longitudinal e controle de velocidade de um carro autônomo

• Ano de defesa: 2013
• Autor(a) principal: Jullierme Emiliano Alves Dias
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; UFMG
• Palavras-chave em Português: Engenharia Elétrica; Engenharia elétrica; Veículos autônomos
• Link de acesso: http://hdl.handle.net/1843/BUOS-99GG4S

An autonomous car is an intelligent vehicle able to drive itself without human intervention. Autonomous cars are certainly inserted in the future of mobility, once its use brings several benefits to the human society, among them the increasing in safety and best use of public roads. Thus, these vehicles have been the focus of attention of several research centers around the world. The motion control of autonomous cars is usually divided into lateral and longitudinal control. The first one is related to the steering of vehicle to follow previously determined paths; whereas the second one refers to the application of acceleration and braking forces. This work is related to the second, specifically the regulation of the longitudinal velocity. Generally, there are two ways to perform the longitudinal control. One is through imitation of the human driving behavior; usually using fuzzy logic, where there is no need of a thorough understanding of the systems dynamics. The other way, which is used inthis work, is to perform control based on a mathematical model of the system. This master thesis presents the dynamic modeling and longitudinal control processes of the autonomous car developed by the PDVA (Group for Research and Development of Autonomous Vehicles at UFMG). The main elements composing the longitudinal dynamics are the powertrain and the body shell of the car, whose general behavior is nonlinear. The model for the accelerator input was based on the structure defined by the physical equations of the car, whose parameters were estimated through stochastic identification techniques. The model was used on the controller, as a way to compensate the nonlinearities of the car dynamics. Furthermore, the use of a proportional and integrativeactions related to the velocity error had the objective of attenuating the multiple disturbances that affect the plant. The brake and throttle actuations were performed through a switching logic, similar to the one in the human driving process. Also, in this work we propose a way to compensate for abrupt changes in the longitudinal dynamics caused bygear changes that occur when the gearshift lever is in automatic mode.The validation of both the model and controller were conducted by computer simulations and real experiments, starting from simple situations toward a more complete context. Discussions were made at the end of this text, pointing out the successes and shortcomings of models and designed controllers.