Learning-based proximity detection algorithm for device-to-device communications

Abstract

The popularity of mobile services that make use of user location has increased in recent years. Proximity-based services is a type of location-based services that determine when a pair of users is in proximity to each other. The mechanism sends a trigger once the users are close enough to each other to start communication. Proximity detection enables the cellular traffic offloading onto direct D2D links that may improve quality of service for the users, save the energy of the device and reduce the network load. However, continuous tracking of the user’s position results in considerable loss in device battery life and negatively affects network capacity.

This thesis presents a novel learning-based algorithm for proximity detection that uses an intelligent polling policy. Several existing proximity detection strategies are considered. It is demonstrated that the implemented optimization approach may significantly reduce the number of position updates and prolong the battery life of the device.