Technical Survey on Sensor-Aided Automatic Parallel Car Parking Systems for Effective Vehicle Navigation

Abstract

Sensor-aided automatic parallel parking systems represent a cornerstone of modern Advanced Driver Assistance Systems (ADAS) and emerging autonomous vehicle technologies. These systems alleviate urban parking challenges by automating space detection, path planning, and precise vehicle control, thereby reducing driver stress, low-speed collisions (by up to 75%), circling time, and associated emissions. Early implementations (2000–2015) relied primarily on ultrasonic sensors for basic reverse aids, evolving into sophisticated multi-modal architectures incorporating radars, cameras, Light Detection and Ranging (LiDAR), infrared, magnetic, and electromagnetic sensors. Sensor fusion strategies leveraging Kalman filters, probabilistic occupancy grids, and deep neural networks address individual sensor limitations such as weather sensitivity, noise, and limited range, achieving detection accuracies exceeding 95% in controlled settings. Recent advancements (2023–2025) integrate Reinforcement Learning (RL), diffusion models, 4D imaging radars, transformers, and end-to-end deep learning for robust performance in dynamic, low-visibility urban environments. Path planning employs geometric (Reeds-Shepp), optimization-based Particle Swarm Optimization (PSO), and Model Predictive Control (MPC) methods, while perception benefits from CNNs (YOLO) and RL for adaptive decision-making. Commercial systems (Tesla Autopark, BMW Parking Assistant Plus, Ford Active Park Assist) vary in their strengths in vision-based autonomy, precision, and reliability. However, challenges persist in adverse weather, computational constraints, sensor interference, regulatory compliance (ISO 26262), and user trust. Real-world benchmarks reveal success rates of 85-99% under ideal conditions but highlight degradation in clutter, rain, or unstructured lots. This review underscores the transition toward fully Autonomous Valet Parking (AVP) and smart-city integration via Vehicle-to-Everything (V2X), while identifying critical needs for weather-resilient fusion, verifiable AI, and enhanced human-machine interfaces to accelerate safe, widespread adoption.