Optimizing multi-access edge computing deployment in urban areas using lstm-based vehicle density detection

Abstract

5G technology offers unprecedented high speeds and low latency, enabling next-generation wireless networks. However, deploying Multi-access Edge Computing (MEC) at antenna sites remains a critical challenge, especially in densely populated urban areas. MEC is essential for delivering real-time services by positioning computing resources close to end users. This study investigates employing the Long Short-Term Memory (LSTM), a deep learning model, for detecting and predicting vehicle trajectories as well as vehicle density to improve urban transportation systems and optimize MEC placement. Using the Cabspotting dataset, which provides GPS co-ordinates of taxis in San Francisco, the data was converted into a time-series format to predict vehicle locations. The LSTM model demonstrated superior prediction accuracy compared to traditional Recurrent Neural Networks (RNNs). To further refine the results, the K-Means algorithm clustered the detected and predicted vehicle positions, identifying optimal zones based on vehicle density for MEC deployment. These findings underscore the potential of LSTM-based vehicle density detection and predictions to enhance strategic MEC placement, advancing smart city infrastructure and sup-porting the rollout of 5G technology.