Controlling Interferences in Smart Building IoT Networks using Machine Learning

Research output: Contribution to journalJournal articleResearchpeer-review

Abstract

The coexistence of many IoT networks in smart buildings poses a major challenge because they interfere mutually. In most settings this results in a greedy approach where each IoT node optimizes its own performance parameters like increasing transmit power, etc. However, this means that interference levels are increased, battery powers are wasted, and spectrum resources are exhausted in high dense settings. In order to control the interference levels, share the spectrum resources, and lower the overall power-consumption this paper proposes a centralized control scheme based on a nonlinear cost function. This cost function is optimized by using machine learning in form of a binary particle swarm optimization algorithm. It has been found that this approach shares the spectrum in a fair way, it saves power and lowers the interference levels, and it dynamically adapt to network changes.
Close

Details

The coexistence of many IoT networks in smart buildings poses a major challenge because they interfere mutually. In most settings this results in a greedy approach where each IoT node optimizes its own performance parameters like increasing transmit power, etc. However, this means that interference levels are increased, battery powers are wasted, and spectrum resources are exhausted in high dense settings. In order to control the interference levels, share the spectrum resources, and lower the overall power-consumption this paper proposes a centralized control scheme based on a nonlinear cost function. This cost function is optimized by using machine learning in form of a binary particle swarm optimization algorithm. It has been found that this approach shares the spectrum in a fair way, it saves power and lowers the interference levels, and it dynamically adapt to network changes.
Original languageDanish
JournalInternational Journal of Sensor Networks
ISSN1748-1279
Publication statusAccepted/In press - 2018
Publication categoryResearch
Peer-reviewedYes
ID: 290005905