Autonomous Parking Systems (APSs) are rapidly evolving, promising enhanced convenience, safety, and efficiency. This review critically examines the current strategies in perception, path planning, and vehicle control, alongside system-level aspects like integration, validation, and security. While significant progress has been made, particularly with the advent of deep learning and sophisticated sensor fusion, formidable challenges persist. This paper delves into the inherent trade-offs, such as balancing computational cost with real-time performance demands; unresolved foundational issues, including the verification of non-deterministic AI components; and the profound difficulty of ensuring robust real-world deployment across diverse and unpredictable conditions, ranging from cluttered urban canyons to poorly lit, ambiguously marked parking structures. We also explore the limitations of current technologies, the complexities of safety assurance in dynamic environments, the pervasive impact of cost considerations on system capabilities, and the critical, often underestimated, need for genuine user trust. Future research must address not only these technological gaps with innovative solutions but also the intricate socio-technical dimensions to realize the full potential of APS.

The perception systems for Traffic Sign Recognition (TSR) and Lane Line Recognition (LLR) are foundational pillars for the safe and effective operation of Advanced Driver-Assistance Systems (ADAS) and fully autonomous vehicles. This review provides a comprehensive analysis of the latest academic research in these domains, strictly focusing on literature published from October 2024 to the present. The analysis reveals several key trends shaping the field. In TSR, architectural evolution is characterized by the refinement of Convolutional Neural Networks (CNNs), the specialization of light-weight YOLO-based models for real-time embedded applications, and the emergence of hybrid CNN-Transformer architectures. Concurrently, a significant research thrust is dedicated to enhancing robustness against environmental adversities and a growing spectrum of sophisticated, physically plausible adversarial attacks. In LLR, the paradigm is rapidly shifting from 2D image-plane detection to full 3D spatial localization and topology reasoning, driven by Transformer-based models that excel at capturing global context and long-range dependencies. Cross-cutting themes common to both domains include a relentless drive for computational efficiency, a data-centric approach marked by the creation of new, challenging benchmarks for adverse conditions and 3D perception, and the nascent but transformative integration of multi-task learning and Vision-Language Models (VLMs) to build systems capable of holistic scene reasoning. Despite significant progress, several key challenges persist in the field of domain generalization, particularly in handling long-tail corner cases and developing safety-aware evaluation metrics. Future research is expected to focus on self-supervised learning, stronger integration between perception and control systems, and the advancement of trustworthy AI through improved explainability and robust-ness. These efforts will lay the groundwork for the next generation of intelligent vehicle systems.