Multi-channel Communication in Wireless Networks

SAHRAOUI, Mohamed (2022) Multi-channel Communication in Wireless Networks. Doctoral thesis, Université de mohamed kheider biskra.

Text Thesis.pdf Download (3MB)

Abstract

Multi-channel communication has been developed to overcome some limitations related to the throughput and delivery rate which become necessary for many applications that require sufficient bandwidth to transmit a large amount of data in Wireless Networks (WNs) such as multimedia communication. However, the requirement of frequent negotiation for the channels assignment process incurs extra-large communication overhead and collisions, which results in the reduction of both communication quality and network lifetime. This effect can play an important role in the performance deterioration of certain WNs types, especially the Wireless Sensor Networks (WSNs) since they are characterized by their limited resources. This work addresses the improvement of communication in multi-channel WSNs. Consequently, four protocols are proposed. The first one is the Multi-Channel Scheduling Protocol (MCSP) for wireless personal networks IEEE802.15.4, which focuses on overcoming the collisions problem through a multi-channel scheduling scheme. The second protocol is the Energy-efficient Reinforcement Learning (RL) Multi-channel MAC (ERL MMAC) for WSNs, which bases on the enhancement of the energy consumption in WSNs by reducing collisions and balancing the remaining energy between the nodes using a singleagent RL. The third work is the proposition of a new heuristically accelerated RL protocol named Heuristically Accelerated Reinforcement Learning approach for Channel Assignment (HARL CA) for WSNs to reduce the number of learning iterations in an energy-efficient way taking into account the bandwidth aspect in the scheduling process. Finally, the fourth contribution represents a proposition of a new cooperative multi-agent RL approach for Channel Assignment (CRLCA) in WSNs, which improves cooperative learning using an accelerated learning model, and overcomes the extra communication overhead problem of the cooperative RL using a new method for self-scheduling and energy balancing. The proposed approach is performed through two algorithms SCRLCA and DCRLCA for Static and Dynamic performance respectively. The proposed protocols and techniques have been successfully evaluated and show outperformed results in different cases through several experiments.

Actions (login required)

View Item