Ultra-Fast Nonlinear Model Predictive Control for Motion Control of Autonomous Light Motor Vehicles

• Published in: World electric vehicle journal 2024-07, Vol.15 (7), p.299
• Main Authors: Patne, Vaishali, Ubare, Pramod, Maggo, Shreya, Sahu, Manish, Rao, G. Srinivasa, Ingole, Deepak, Sonawane, Dayaram
• Format: Article
• Language: English
• Subjects: Accuracy; Advanced Driver Assistance System (ADAS); Advanced driver assistance systems; Algorithms; Automobile industry; Complexity; Control systems; Degrees of freedom; Dynamical systems; embedded implementation; Error reduction; Extended Kalman filter; Hardware-in-the-loop simulation; Human error; Human motion; Linearization; Mathematical models; model predictive control; Motion control; Motor vehicles; Nonlinear control; Nonlinear dynamics; Nonlinear systems; Optimal control; Optimization techniques; Predictive control; Real time; real-time vehicle control; Safety critical; Vehicles
• ISSN: 2032-6653; 2032-6653
• DOI: 10.3390/wevj15070299

Advanced Driver Assistance System (ADAS) is the latest buzzword in the automotive industry aimed at reducing human errors and enhancing safety. In ADAS systems, the choice of control strategy is not straightforward due to the highly complex nonlinear dynamics, control objectives, and safety critical constraints. Nonlinear Model Predictive Control (NMPC) has evolved as a favorite option for optimal control due to its ability to handle such constrained, Multi-Input Multi-Output (MIMO) systems efficiently. However, NMPC suffers from a bottleneck of high computational complexity, making it unsuitable for fast real-time applications. This paper presents a generic framework using Successive Online Linearization-based NMPC (SOL-NMPC) for for the control in ADAS. The nonlinear system is linearized and solved using Linear Model Predictive Control every iteration. Furthermore, offset-free MPC is developed with the Extended Kalman Filter for reducing model mismatch. The developed SOL-NMPC is validated using the 14-Degrees-of-Freedom (DoF) model of a D-class light motor vehicle. The performance is simulated in matlab/Simulink and validated using the CarSim® software (Version 2016). The real-time implementation of the proposed strategy is tested in the Hardware-In-the-Loop (HIL) co-simulation using the STM32-Nucleo-144 development board. The detailed performance analysis is presented along with time profiling. It can be seen that the loss of accuracy can be counteracted by the fast response of the proposed framework.