Article Preview
Top1. Introduction
While the bandwidth available in the licensed spectrum is relatively insufficient for high volume traffic, the skyrocketing demand for bandwidth-hungry (e.g. multimedia) applications has led to an equivalent surge in the demand for more bandwidth. Therefore, research efforts for efficient schemes to increase the network capacity are relentless (Afifi et al., 2016). As a result, it was recently proposed to extend the operation of LTE to the unlicensed spectrum, which is known as LTE-U. This extension is realized through the deployment of small cells and carrier aggregation where a carrier in the licensed spectrum is aggregated with a carrier or more in the unlicensed spectrum to create a larger data pipe that can support a higher data rate.
A major challenge, however, that faces the implementation of LTE-U is coexistence of LTE, in its current form, with the incumbent Wi-Fi technology. On the one hand, Wi-Fi employs the contention-based Carrier-Sense Multiple Access with Collision Avoidance (CSMA/CA) mechanism, whereby the channel is sensed for a certain period of time, and accessed only if found idle. On the other hand, LTE does not apply such contention-based access mechanism and hence this will lead to severe degradation in the Wi-Fi performance due to LTE's continuous transmission over the channel.
Several researchers have addressed the coexistence problem of LTE-U and proposed some mechanisms with the objective of enabling efficient sharing of the unlicensed spectrum between LTE-U and Wi-Fi. Such coexistence mechanisms can be classified as time-, frequency-, and power-based mechanisms.
While time-based mechanisms focus on the fair division of the channel airtime between LTE-U and Wi-Fi, frequency-based ones focus on switching the transmissions between channels in a way that busy channels with larger number of Wi-Fi transmissions are avoided. Power control-based coexistence mechanisms (see, for example, (Abinader et al., 2014) and (Chaves et al., 2013) attempt to mitigate the interference between LTE-U and Wi-Fi by adaptively adjusting the transmission power of LTE-U nodes according to the conditions of the channel. Listen Before Talk (LBT) (3GPP, 2014), Almost Blank Subframes (ABS) (Zhang et al., 2015), and Carrier Sense Adaptive Transmission (CSAT) (“Qualcomm research LTE in unlicensed spectrum”, 2014) are among the proposed time-based mechanisms while dynamic frequency selection (“Qualcomm research LTE in unlicensed spectrum”, 2014), and learning-based schemes (Sallent et al., 2015) are examples of the frequency-based mechanisms.
Another class of coexistence mechanism is game-based, where each of the LTE-U and Wi-Fi nodes is treated as a game player that competes for resources (Cai et al., 2016; Zhang et al., 2016). Machine learning (ML) techniques have been also proposed to address the coexistence of LTE with Wi-Fi. While this actually took place in different scenarios, for instance ML is considered in the context of resource allocation in (Chen et al., 2017) while adaptation of LBT contention-window size is considered in (Ali et al., 2017). However, to the best of our knowledge, none of the work in the literature has employed ML-based techniques to predict channel state in the context of LTE-U.