Intelligent Smart Parking Management Using Hybrid Deep Learning in IoT Cities

Abstract

Urbanization and the Internet of Things (IoT) are on the rise, and managing urban parking has become a difficult challenge for existing smart parking systems, as they have failed to manage traffic effectively, lack security, have low user adaptability, inaccurately predict parking demand, store unsecured data, create unfair pricing, and suffer from improper parking detection. This paper introduces a new hybrid deep learning-based Artificial Intelligence (AI) smart parking management system for the IoT city. We design a simulator environment and add 10 vehicles to simulate real-time data input. User data privacy is secured with Elliptic Curve Cryptography (ECC) and Local Differential Privacy (LDP). Dynamic pricing is formulated as a Stackelberg Leader-Follower Game Framework (SL-FGF) for optimization. At the same time, we propose the Acar Agent Multi-Criteria Decision-Making (AcarA-MCDM) framework for intelligent space allocation. A hybrid CNN-LSTM performs parking slot prediction with the Lion Optimization Algorithm (CNN-LSTM-LOA), and an Intelligent Parking Reservation System (IPRS) is deployed. Mis-parking detection is based on an autoencoder with a GPS Navigation system to ensure proper space utilization. Evaluation of the system on 10 performance metrics yielded the best results, i.e., detection accuracy of 98.5%, precision of 99.2%, recall of 97.3%, F1-score of 98.3%, allocation efficiency of 95.5%, latency of 57 ms, and energy consumption of 99 mJ, establishing significant improvements over state-of-the-art methods