Litcius/Paper detail

Adaptive Directional Neighbor Discovery Schemes in Wireless Networks

Btissam El Khamlichi, Jamal El Abbadi, Nathaniel W. Rowe, Sunil Kumar

20202020 International Conference on Computing, Networking and Communications (ICNC)16 citationsDOI

Abstract

The use of fully directional links in ad-hoc networks improves the overall network performance but complicates the process of neighbor discovery. In most of the state-of-the-art schemes, nodes switch their direction either randomly or in a predefined sequence, without considering the result of past discovery attempts in each sector. This introduces high discovery latency and severe long tail problem, especially for high density networks and/or narrow beams. To overcome these limitations, we propose two novel decentralized, and low complexity reinforcement learning-based neighbor discovery schemes in this paper. In these schemes, the neighbor discovery is mapped to a stochastic multi-player game, where each node independently adjusts its policy via the Q-learning based scheme to minimize the discovery latency. Effectiveness of the proposed schemes is assessed through numerical results, and it is observed that these proposed schemes are able to achieve a significantly faster network-wide neighbor discovery, while incurring low overhead and computational complexity.

Topics & Concepts

Neighbor Discovery ProtocolComputer scienceBusiness process discoveryLatency (audio)Overhead (engineering)Wireless ad hoc networkDistributed computingReinforcement learningComputer networkNode (physics)WirelessArtificial intelligenceThe InternetWork in processInternet protocol suiteEngineeringOperating systemOperations managementBusiness process modelingBusiness processTelecommunicationsStructural engineeringWorld Wide WebAdvanced MIMO Systems OptimizationMobile Ad Hoc NetworksWireless Networks and Protocols