A comprehensive analysis of model predictive control for lane keeping assist system

Lane Keeping Assist System (LKAS) enhances comfort and safety while driving. It plays a significant role in the Advanced Driver Assistance System (ADAS) and future Automated Driving (AD). The LKAS solution aims to help the driver keep the vehicle within the road lines, preventing unintentional lane...

ver descrição completa

Detalhes bibliográficos
Autores: Garcia, James, Teixeira, Evandro, Murilo, André, Rodrigues, Rafael
Formato: artículo
Estado:Versión publicada
Fecha de publicación:2023
País:Brasil
Recursos:Universidade Federal de Lavras (UFLA)
Repositorio:Repositório Institucional da UFLA
Idioma:portugués
OAI Identifier:oai:repositorio.ufla.br:1/59504
Acesso em linha:https://repositorio.ufla.br/handle/1/59504
https://ieeexplore.ieee.org/document/10354292
Access Level:acceso abierto
Palavra-chave:Advanced driver assistance systems
Lane Keeping assist system
Model predictive control
MPC parameterization
Descrição
Resumo:Lane Keeping Assist System (LKAS) enhances comfort and safety while driving. It plays a significant role in the Advanced Driver Assistance System (ADAS) and future Automated Driving (AD). The LKAS solution aims to help the driver keep the vehicle within the road lines, preventing unintentional lane departure. Despite LKAS being an important solution for comfortable driving, robust LKAS steering control is still lacking, requiring constant driver intervention or premature LKAS deactivation. LKAS require optimal control solutions with real-time constraints. This paper comprehensively analyzes Model Predictive Control (MPC) for real-time LKAS applications. Classical and parameterized MPC schemes with distinct Quadratic Programming (QP) solvers are combined to evaluate LKAS closed-loop control performance and realtime constraints. A sideslip and lateral speed bicycle modes were used to evaluate classical, trivial, and exponential MPC schemes. Experimental results highlight the three MPC and QP-appropriate solutions with satisfactory reference tracking without steering command and real-time constraints violation.