Programa del congreso

The video game sector is one of the fastest-growing industries in recent times and one with the most solid expectations to continue expanding in the following years. One key element in the future of these applications is the adoption of the cloud gaming paradigm, which allowing vent user equipment. Nonetheless, this approach establishes hugely challenging requirements on the network side in order to support high amount of data exchanged with the lowest latency possible. Here, the present work describes the main metrics associated with the performance and requirements of this novel service. Finally, a comparison between different technologies such as Ethernet, WiFi, LTE and 5G is given, showing the performance of each technology in the provision of these kind of services in terms of lag and smoothness.

Traffic forecasting with high time resolution (i.e., hour) is key to manage network slicing in 5G networks. However, the high dynamism of actual cellular networks makes predicting future traffic fluctuations an incredibly arduous task. In this context, amazing results obtained from other fields of research have put deep-learning models into the spotlight. This work presents a comparative study of the performance of different deep-learning models to forecast hourly cell traffic in both Downlink (DL) and Uplink (UL). For this purpose, a dataset from a live LTE network is collected for 2 months. Both classical and multi-tasking deep learning approaches have been considered. Results show that multi-tasking models combining recurrent and convolutional layers show the highest accuracy, revealing that the information of neighbor cells is useful when forecasting traffic in cells serving social events.

This work describes an 8-pixel imaging optical receiver for multiple-input multiple-output (MIMO) indoor visible light communications (VLC) based on space-division multiple access (SDMA). The communication system also uses an adaptive orthogonal frequency-division multiplexing (OFDM) modulation scheme, enabling multi-user transmission with constrained inter-user interference. The simulation results show that the proposed optical receiver offers a smoother system behavior irrespective of its orientation when receptor moves throughout a 4-lamp room as compared with using a 4-pixel imaging optical receiver, which suits better with realistic indoor scenarios where receiver orientation with respect to light lamps cannot be easily controlled.

Free space optical (FSO) communications have been considered a competitive alternative to the radio frequency systems (RF) due to its large bandwidth. Analytical close-form expressions for the probability error are derived for a variety of atmospheric conditions assuming intensity modulation/direct detection (IM/DD) link using on-off keying (OOK). The gamma-gamma distribution will be used to model the random variation of irradiance due to its validity for any turbulence condition. In addition, a spatial diversity technique in the receiver side is proposed to improve the link performance and the impact of the correlation among channels will be studied. The analytical results, numerically verified by Monte-Carlo simulations, will give us an idea of the feasibility of using this technology in a near future.

In this paper, we study the performance of a drone-to-ground free-space optical (FSO) link with energy harvesting in the presence of both atmospheric gamma-gamma turbulence and pointing errors. Furthermore, we assume that the direct current component of the FSO signal, which is normally filtered out, can be employed for energy harvesting at the drone, increasing its autonomy. To this end, a variant of the classic OOK modulation is considered. Analytical closed-form expressions are derived for a variety of scenarios. Monte Carlo simulations are also carried out to verify our analytical results.