LSTM Based Real-time Smart Parking System

In the current scenario, population and vehicles in cities have been increasing exponentially while the spatial resources to handle them, be it roads or parking spaces, have almost remained the same. Due to which, traffic and parking management have received a growing amount of coverage in recent years. Several solutions are being considered in various countries, most of which include mathematical modelling, parametric, and non-parametric methods. Moreover, some solutions such as an IoT network of vehicular sensors, although being very efficient, are considered a privacy concern in most countries. Thus, there is a need for less intrusive and easily implementable traffic management solutions worldwide and especially in developing countries like India. In this paper, LSTM based real-time smart parking system (SPS) has been proposed for parking space availability prediction. An IoT based network has been used for the multidisciplinary approach to solve this problem. Proposed system also contains a companion web application to show the parking predictions, real-time availability status. It will help drivers to find available parking spaces according to their convenience with 6.43% mean relative error and 0.82 R2 score in prediction. Proposed system helps to minimize excessive energy demand and CO2 emission of roaming vehicles. So, the proposed system is highly configurable and efficient to provide better quality of service to solve parking problems in smart cities.