Dual resource allocation for enhancing vehicle tracking using deep neural network and Bayesian LoRa

Improper transmission parameter settings result in power and data loss, contributing to delays within the real-time tracking involving static gateways (GWs) and dynamic nodes (DNs) in a LoRa-based intelligent transportation system. Additionally, transitioning between different transmission parameters causes interruptions that lead to latency issues in the tracking system. This paper proposes a hybrid technique that combines a deep neural network (DNN), Bayesian Game Parameter Selection (BGPS), and dual resource allocation to optimize tracking performance. The BGPS manages power control to minimize inter-GW power loss, and dual resource allocation assigns the DNs to fixed clusters for stable and adaptive clusters for dynamic scenarios, determining efficient bidirectional communications and accurate trajectories of vehicles. The DNN decodes vehicle trajectories and communication patterns using labeled data in offline training, with adaptation in the online mode. The integrated use of the proposed model enables dynamic decision-making in response to changing environmental and traffic conditions, providing higher tracking precision, reduced latency, and more economical energy consumption. Experimental results demonstrate 96 % accuracy, a 63.8 % latency reduction, and 66.9 % lower energy consumption with a 4.3 % packet rejection rate (PRR), which achieves higher tracking precision and robustness for the proposed model compared to existing methods.