کنفرانس مهندسی برق ایران

صفحه اصلی / سی و دومین کنفرانس بین المللی مهندسی برق

Learning-Based Routing Policy For Wireless Sensor Networks

نویسندگان :

‪Najim Halloum‬‏¹ Yousef Darmani² Ali Ahmadi³

1- دانشگاه صنعتی خواجه نصیرالدین طوسی 2- دانشگاه صنعتی خواجه نصیرالدین طوسی 3- دانشگاه صنعتی خواجه نصیرالدین طوسی

کلمات کلیدی :

IoT،Reinforcement Learning،WSN،Q-Learning،Routing،Reliability،RPL،Load-balancing،Objective Function

چکیده :

The pervasive Internet of Things (IoT) technology applications take the quality of the service provided to another level. Wireless Sensor Network (WSN) is the sensing layer in IoT-based systems that collects data and sends it to a central component. On the other hand, IoT-WNSs have special characteristics in terms of lossy environment and resource-constrained devices. However, the next generation of the Internet of Things networks is expected to be large-scale networks with large amounts of data being transferred. In such IoT networks, the main issues come from the need for an efficient routing protocol that can provide an efficient forwarding mechanism in congestion-based scenarios. Here, the routing protocol for low-power and Lossy networks (RPL) is a standardized routing protocol for IoT-WSNs. With such conditions, RPL could suffer from reliability decreasing, control overhead and latency increasing. Reinforcement learning provides a methodology by which a system can improve its behavior to select upcoming actions efficiently by learning from previous interactions with the environment. In this study, a Reinforcement Learning-based load balancing-aware routing scheme will be proposed to build a reliable path for IoT-WSNs. To do so, a multi-metric reward function based on the Received Signal Strength Indicator and Buffer utilization metrics is designed to form the essential part of the next forwarder selection process using the Q-Learning algorithm. According to an evaluation using the Contiki OS / Cooja simulator in terms of control overhead, network latency, and packet delivery ratio, the proposed approach has demonstrated its efficiency against the counterparts.