A DELAY AWARE Q-LEARNING APPROACH FOR HYBRID CLOCK SYNCHRONIZATION PROTOCOL IN WSNS

Abstract

Achieving reliable and energy-efficient clock synchronization in Wireless Sensor Networks (WSNs) is a critical challenge due to clock drift, communication delays, and node failures. In particular, synchronization protocols such as reference and consensus-based techniques, need trade-offs between convergence speed, synchronization accuracy, and robustness. To address these issues, this paper proposes a unified synchronization framework called Q-learning-based Hybrid Time Synchronization Protocol (QHTSP), which combines the strengths of Partial Timestamp Synchronization (PTS) and hybrid synchronization methodologies. The QHTSP protocol uses Q-learning to dynamically switch between reference and consensus-based modes which results in fast convergence and fault tolerance. It also includes Gaussian and exponential stochastic delays, which provide reliable estimations for clock skew and offset using Maximum Likelihood Estimation and Best Linear Unbiased Estimator, respectively. Specifically, the proposed protocol offers an accurate synchronization network situations like queuing-induced and multi-source interference delays. By integrating statistical signal processing with learning-based dynamic adaptation, QHTSP delivers a scalable and fault-tolerant approach to synchronize the nodes. Simulation results show that QHTSP provides fast and energy-efficient reliable synchronization making it ideal for large-scale WSNs deployed in IoT applications. Simulations of various network topologies with nodes ranging from 100 to 1000 reveal that QHTSP outperforms some existing protocols such as Hybrid Time Synchronization Protocol (HTSP), Flooding Time Synchronization Protocol (FTSP), and Enhanced-Flooding Time Synchronization Protocol (E-FTSP) in terms of convergence time, synchronization error, and resilience to node failures.