Scheduling in Multi-Hop Wireless Networks using a Distributed Learning Algorithm

Abstract

This study takes on the challenge of learning and scheduling in a Multi-Hop Wireless Network (MHWN) without having any prior knowledge of connection charges. Earlier scheduling methods needed knowledge of the connection rates, whereas Techniques for training often required a centralized authority and had an exponential performance. These represent a significant barrier to creating a reliable distributed strategy for resource allocation that relies on learning in massive multi-hop networks. Management of multihop control It is a tough task to implement wireless networks in a distributed fashion while meeting end-to-end timing constraints for a variety of flows. Using the notions of Draining Time and Continuous Review, which originate from the notion of fluid boundaries of queues, an algorithm is constructed that meets delay needs for various streams in a system. This study proposes a regulated maximal matching, a completely distributed scheduling algorithm that ensures at least 50% of the performance of a centralized method. The approach employs a distributed optimization procedure called iterative gradient ascent, which is carried out in a cyclic fashion between nodes with little data interchange. The system prioritizes flows using weights that change over time. The effectiveness of the algorithm is analyzed in a network setting where interruption is symbolized by isolated nodes.