Federico Chiariotti

Online learning adaptation strategy for DASH clients

In this work, we propose an online adaptation logic for Dynamic Adaptive Streaming over HTTP (DASH) clients, where each client selects the representation that maximize the long term expected reward. The latter is defined as a combination of the decoded quality, the quality fluctuations and the rebuffering events experienced by the user during the playback. To solve this problem, we cast a Markov Decision Process (MDP) optimization for the selection of the optimal representations. System dynamics required in the MDP model are a priori unknown and are therefore learned through a Reinforcement Learning (RL) technique. The developed learning process exploits a parallel learning technique that improves the learning rate and limits sub-optimal choices, leading to a fast and yet accurate learning process that quickly converges to high and stable rewards. Therefore, the efficiency of our controller is not sacrificed for fast convergence. Simulation results show that our algorithm achieves a higher QoE than existing RL algorithms in the literature as well as heuristic solutions, as it is able to increase average QoE and reduce quality fluctuations.

SymbioCity: Smart cities for smarter networks

The “Smart City” concept revolves around the idea of embodying cutting-edge information and communication technology solutions in the very fabric of future cities to offer new and better services to citizens while lowering the city management costs in monetary, social, and environmental terms. In this framework, communication technologies are perceived as subservient to the Smart City services, providing the means to collect and process the data needed to make the services function. In this paper, we propose a new vision in which technology and Smart City services are designed to take advantage of each other in a symbiotic manner. According to this new paradigm, which we call “SymbioCity”, Smart City services can indeed be exploited to improve the performance of the same communication systems that provide them with data. Suggestive examples of this symbiotic ecosystem are discussed in this paper. The discussion is then substantiated in a proof-of-concept case study, where we show how the traffic monitoring service provided by the London Smart City initiative can be used to predict the density of users in a certain zone and optimize the cellular service in that area.

D-DASH: A Deep Q-Learning Framework for DASH Video Streaming

The ever-increasing demand for seamless high-definition video streaming, along with the widespread adoption of the dynamic adaptive streaming over HTTP (DASH) standard, has been a major driver of the large amount of research on bitrate adaptation algorithms. The complexity and variability of the video content and of the mobile wireless channel make this an ideal application for learning approaches. Here, we present D-DASH, a framework that combines deep learning and reinforcement learning techniques to optimize the quality of experience (QoE) of DASH. Different learning architectures are proposed and assessed, combining feed-forward and recurrent deep neural networks with advanced strategies. D-DASH designs are thoroughly evaluated against prominent algorithms from the state-of-the-art, both heuristic and learning-based, evaluating performance indicators such as image quality across video segments and freezing/rebuffering events. Our numerical results are obtained on real and simulated channel traces and show the superiority of D-DASH in nearly all the considered quality metrics. Besides yielding a considerably higher QoE, the D-DASH framework exhibits faster convergence to the rate-selection strategy than the other learning algorithms considered in the study. This makes it possible to shorten the training phase, making D-DASH a good candidate for client-side runtime learning.

Using Smart City Data in 5G Self-Organizing Networks

So far, research on Smart Cities and self-organizing networking techniques for fifth-generation (5G) cellular systems has been one-sided: a Smart City relies on 5G to support massive machine-to-machine (M2M) communications, but the actual network is unaware of the information flowing through it. However, a greater synergy between the two would make the relationship mutual, since the insights provided by the massive amount of data gathered by sensors can be exploited to improve the communication performance. In this paper, we concentrate on self-organization techniques to improve handover efficiency using vehicular traffic data gathered in London. Our algorithms exploit mobility patterns between cell coverage areas and road traffic congestion levels to optimize the handover bias in heterogeneous networks and dynamically manage mobility management entity (MME) loads to reduce handover completion times.

A Dynamic Approach to Rebalancing Bike-Sharing Systems

Bike-sharing services are flourishing in Smart Cities worldwide. They provide a low-cost and environment-friendly transportation alternative and help reduce traffic congestion. However, these new services are still under development, and several challenges need to be solved. A major problem is the management of rebalancing trucks in order to ensure that bikes and stalls in the docking stations are always available when needed, despite the fluctuations in the service demand. In this work, we propose a dynamic rebalancing strategy that exploits historical data to predict the network conditions and promptly act in case of necessity. We use Birth-Death Processes to model the stations’ occupancy and decide when to redistribute bikes, and graph theory to select the rebalancing path and the stations involved. We validate the proposed framework on the data provided by New York City’s bike-sharing system. The numerical simulations show that a dynamic strategy able to adapt to the fluctuating nature of the network outperforms rebalancing schemes based on a static schedule.

Learning methods for long-term channel gain prediction in wireless networks

Efficiently allocating resources and predicting cell handovers is essential in modern wireless networks; however, this is only possible if there is an efficient way to estimate the future state of the network. In order to accomplish this, we investigate two learning techniques to predict the long-term channel gains in a wireless network. Previous works in the literature found efficient methods to perform this prediction with the aid of a GPS signal: in this work, we predict the future channel gains using only past channel samples, without any geographical information.

QoE-aware Video Rate Adaptation algorithms in multi-user IEEE 802.11 wireless networks

The spreading of video streaming services in the last few years is presenting new challenges in wireless networking; Video Rate Adaptation (VRA) is a technique that optimizes the bandwidth usage by adapting video quality as network conditions change. We propose two Quality of Experience (QoE) aware algorithms that perform VRA while guaranteeing user satisfaction.

Mobility-aware handover strategies in smart cities

Supporting the Internet of Things and Smart City applications is one of the most important goals in the ongoing design process of 5G cellular systems. Another trend is an increasing focus on data-driven optimization and Self-Organized Networking, in order to automate network deployments and increase performance and efficiency. This approach, however, does not fully take advantage of the data generated by the Smart City. In this work, we propose to process and use the information flowing through the network from the city sensors to increase the awareness of the network itself, improving the communication performance. We exploit vehicular traffic data from the Traffic for London (TfL) sensor network to infer mobility patterns and improve the efficiency of LTE handovers.

Cell traffic prediction using joint spatio-temporal information

In future cellular networks, the ability to predict network parameters such as cell load will be a key enabler of several proposed adaptation and resource allocation techniques. In this study, we consider a joint exploitation of spatio-temporal data to improve the prediction accuracy of standard regression methods. We test several such methods from the literature on a publicly available dataset and document the advantages of the proposed approach.

A game-theoretic analysis of energy-depleting jamming attacks

Jamming may be a serious issue in Internet of Things networks with battery-powered nodes, as an attacker can not only disrupt packet delivery, but also reduce the lifetime of energy-constrained nodes. In this work, we consider a malicious attacker with the dual objective of preventing communication and depleting the battery of a targeted node. We model this scenario as a multistage game, derive optimal strategies for both sides, and evaluate their consequences on network performance. Our work highlights the trade-off between node lifetime and communication reliability and how the presence of a jammer affects both these aspects.