A Systematic Review of Safety-Oriented Model Predictive Control Methods for Autonomous Vehicles

Ensuring safety remains a fundamental challenge in the utilization of autonomous vehi-cles (AVs) in real-world environments. Model Predictive Control (MPC) has emerged as a prominent control strategy for enhancing AV safety because of its ability to handle multi-variable systems, anticipate future events, and incorporate system constraints. This study conducts a systematic literature review (SLR) of MPC-based strategies for autonomous ve-hicle safety, with a focus on existing approaches, identified challenges, and emerging fu-ture directions, following the PRISMA 2020 guidelines. The methodology included prede-fined inclusion/exclusion criteria, structured screening, and thematic classification. Based on an in-depth examination of 33 peer-reviewed studies (2015–2025), a diverse landscape of MPC applications in AV safety has been revealed; particularly in three critical domains: collision avoidance and risk mitigation, trajectory tracking and path following, and inter-section and coordination tasks. Various formulations of the MPC have been explored such as linear, nonlinear, robust, adaptive, and learning-enhanced variants. These formula-tions are often integrated with perception modules, motion prediction, or probabilistic risk models to support real-time decision-making under uncertainty and constraints. Howev-er, it stills a challenge to achieve scalability, computational efficiency, and robustness in complex, high-speed, or uncertain environments. In this study, the need for unified safety metrics, standard testing protocols, and scalable frameworks are highlighted to bridge control and perception. Also, the promising direc-tions have been outlined such as stochastic MPC, data-driven modeling, dual-control frameworks, and integration with vehicle-to-everything (V2X) communications. This study provides a foundation for developing the next generation of MPC schemes that are safer, more adaptable, and better aligned with the demands of fully autonomous driving systems.