Learning the site-specific distribution of the wireless channel within a particular environment of interest is essential to exploit the full potential of machine learning (ML) for wireless communications and radar applications. Generative modeling offers a promising framework to address this problem. However, existing approaches pose unresolved challenges, including the need for high-quality training data, limited generalizability, and a lack of physical interpretability. To address these issues, we combine the physics-related compressibility of wireless channels with generative modeling, in particular, sparse Bayesian generative modeling (SBGM), to learn the distribution of the underlying physical channel parameters. By leveraging the sparsity-inducing characteristics of SBGM, our methods can learn from compressed observations received by an access point (AP) during default online operation. Moreover, they are physically interpretable and generalize over system configurations without requiring retraining.

The potential of Time Interference Alignment is investigated in this work, with particular reference to the attainable degrees of freedom. The K-user interference channel is considered, in which transmitters and receivers are placed randomly in a Euclidean space. A model for long delay networks is introduced and the degrees of freedom for different cases (with and without transmitter delay coordination) are evaluated. It is shown how time interference alignment can provide more degrees of freedom than TDMA when the transmitters jointly coordinate their transmission delay and the number of pairs is K ≥ 5. Closed form expressions are derived for several cases of interest which provide insight and useful predictions. This work is concluded with an investigation of the achievable degrees of freedom for multi-satellite networks, where it is shown that the results obtained under several assumptions do predict accurately the degrees of freedom in a real setting.

The existing studies of RObust Header Compression (ROHC) have provided some understanding for memoryless channel, but the behavior of ROHC for correlated wireless channels is not well investigated in spite of its practical importance. In this paper, the dependence of ROHC against its design parameters for the Gilbert Elliot channel is studied by means of three analytical models. A first more elaborated approach accurately predicts the behavior of the protocol for the single RTP flow profile, while a simpler, analytically tractable model yields clear and insightful mathematical relationships that explain the qualitative trends of ROHC. The results are validated against a real world implementation of this protocol. Moreover, a third model studies also the less conventional yet practically relevant setting of multiple RTP flows.

The Airborne Internet is a vision of a large-scale multihop wireless mesh network consisting of commercial passenger aircraft connected via long-range highly directional air-to-air radio links. We propose a geographic load sharing strategy to fully exploit the total air-to-ground capacity available at any given time. When forwarding packets for a given destination, a node considers not one but a set of next-hop candidates and spreads traffic among them based on queue dynamics. In addition, load balancing is performed among Internet Gateways by using a congestion-aware handover strategy. Our simulations using realistic North Atlantic air traffic demonstrate the ability of such a load sharing mechanism to approach the maximum theoretical throughput in the network.

The necessity of accurate channel estimation for Successive and Parallel Interference Cancellation is well known. Iterative channel estimation and channel decoding (for instance by means of the Expectation-Maximization algorithm) is particularly important for these multiuser detection schemes in the presence of time varying channels, where a high density of pilots is necessary to track the channel. This paper designs a method to analytically derive a weighting factor a, necessary to improve the efficiency of interference cancellation in the presence of poor channel estimates. Moreover, this weighting factor effectively mitigates the presence of incorrect decisions at the output of the channel decoder. The analysis provides insight into the properties of such interference cancellation scheme and the proposed approach significantly increases the effectiveness of Successive Interference Cancellation under the presence of channel estimation errors, which leads to gains of up to 3 dB.

Time Interference Alignment is a flavor of Interference Alignment that increases the network capacity by suitably staggering the transmission delays of the senders. In this work the analysis of the existing literature is generalized and the focus is on the computation of the dof for networks with randomly placed users in a n-dimensional Euclidean space. In the basic case without coordination among the transmitters analytical expressions of the sum dof can be derived. If the transmit delays are coordinated, in 20% of the cases time Interference Alignment yields additional dof with respect to orthogonal access schemes. The potential capacity improvements for satellite networks are also investigated.

The Airborne Internet is a vision of a large scale multihop wireless mesh network consisting of commercial passenger aircraft connected via long range highly directional air-to-air radio links. We propose a geographic load sharing strategy to fully exploit the total air-to-ground capacity available at any given time. When forwarding packets for a given destination, a node considers not one but a set of next hop candidates, and spreads traffic among them based on queue dynamics. In addition, load balancing is performed among Internet Gateways by using a congestion-aware handover strategy. Our simulations using realistic North Atlantic air traffic reveal the potential of such a load sharing mechanism to approach the maximum theoretical throughput in the network.

This paper explores the relationship between cooperation and two very different medium access strategies: CSMA (epitomized by IEEE 802.11 DCF) and one instance of TDMA (represented by IEEE 802.15.3). By cooperation we mean basic decode-and-forward relaying as well as more advanced forms thereof based on network coding. The essential features that make each system more or less suitable to support the potentially high performance gains of hybrid ARQ are analyzed and studied by means of extensive simulations. The discussion carried out in this paper shows that centralized systems are able to reap far higher gains from a cooperative paradigm than their distributed counterparts, and investigates in depth the reasons that lead to such different behaviors, prompting also some interesting insights for proper design of advanced hybrid ARQ protocols.

The Airborne Internet is envisioned to be a large scale multihop wireless mesh network of commercial passenger aircraft connected via long range highly directional air-to-air radio links. We propose a localized geographic load sharing technique to mitigate congestion in this network, taking into account the underlying link scheduling constraints with directional antennas. When forwarding packets for a given destination, a node considers not one but a set of next hop candidates, and spreads traffic among them based on queue dynamics. Our simulations show that introducing this flexibility in the routing function can greatly increase a node's ability to satisfy its bandwidth demands during link scheduling, yielding significant performance improvements in terms of network throughput and average packet delay.

Past research has shown that Carrier Sense Multiple Access (CSMA) is not a particularly favorable environment for cooperative networks. This observation suggests that other medium access schemes, like scheduled TDMA, may enable cooperation to yield more substantial improvements over conventional techniques. From this point of view, a hybrid cooperative-network coding protocol (called Phoenix, and so far only proposed for CSMA) has been adapted for Personal Area Network communications. The improvements brought by cooperative-NC techniques in this setting have roughly doubled with respect to the CSMA scenario, with gains of over 20% with respect to standard cooperative systems.

This paper proposes a method to improve the robustness of the carrier sense system for opportunistic MAC (OMAC). This carrier sense method is specific for MIMO and the new version here illustrated is robust to noise and interference. Network simulations prove that this modified MIMO carrier sense correctly estimates whether a new communication would harm other ongoing transmissions for a wide range of system parameters.

We consider a network where each route comprises a backlogged source, a number of relays and a destination at a finite distance. The locations of the sources and the relays are realizations of independent Poisson point processes. Given that the nodes observe a TDMA/ALOHA MAC protocol, our objective is to determine the number of relays and their placement such that the mean end-to-end delay in a typical route of the network is minimized. We first study an idealistic network model where all routes have the same number of hops, the same distance per hop and their own dedicated relays. Combining tools from queueing theory and stochastic geometry, we provide a precise characterization of the mean end-to-end delay. We find that the delay is minimized if the first hop is much longer than the remaining hops and that the optimal number of hops scales sublinearly with the source-destination distance. Simulating the original network scenario reveals that the analytical results are accurate, provided that the density of the relay process is sufficiently large. We conclude that, given the considered MAC protocol, our analysis provides a delay-minimizing routing strategy for random, multihop networks involving a small number of hops.

This paper introduces a novel MAC protocol for wireless networks, called Phoenix, that employs ideas from network coding to enhance decode and forward cooperation. A relay is allowed to code data of its own together with a corrupted packet during a retransmission at no additional cost in bandwidth. Therefore, while in conventional cooperative protocols a node becomes a relay only to assist other terminals, with our proposal a cooperator can also serve its own traffic. We evaluate Phoenix's performance by means of a theoretical model and extensive simulation campaigns. We show that Phoenix is especially beneficial in multihop settings and interesting gains over benchmark protocols can be achieved.

Beamforming is regarded as a key element for multiantenna ad hoc networks. However, it cannot simultaneously provide the omnidirectional and long-range coverage required by broadcast transmissions, a problem known as the Asymmetry in Gain. We propose a scheme for control packet exchange over an extended radio coverage based on a Space Time Code technique. This solution is shown to solve the Asymmetry in Gain issue, and is proposed as part of a MAC protocol for MIMO ad hoc networks, whose performance benefits include increased throughput and reduced delay and energy consumption.

There is wide consensus that properly taking into account wireless interference is necessary to design high performance link scheduling algorithms for multi-hop networks. However, most approaches in the literature use simplified models, which significantly abstract from the physical behavior of wireless links. Indeed, the main problem in representing the wireless propagation conditions with a proper level of detail is the very large amount of information needed, that includes the wireless link gains between all node pairs. In this paper, we propose a greedy, centralized scheduler which is based on the physical interference model but aims at exploiting local information available at each node, in order to reduce global information exchange and therefore the overhead as well as the computational complexity of the algorithm. We prove the effectiveness of our approach by extensive simulation results. We also show that our system outperforms the most up-to-date benchmark in realistic interference aware schedulers for wireless multi-hop networks.

In this paper, we investigate how to apply MIMO_NC in order to increase the efficiency of cooperative communications. MIMO_NC is a recent strategy to combine, in an integrated system, MIMO detection techniques and network coding. MIMO_NC is robust to channel errors, as it can exploit the spatial diversity provided by multiple transmissions of the same information, and throughput effective as it is based on network coding. In this paper we design a cooperative MAC protocol based on MIMO_NC which can outperform classical cooperative schemes. We also prove its effectiveness by means of analytical modelling and simulations in different network configurations.

Physical carrier sense for wireless multihop networks has attracted a certain degree of attention since it is a simple yet important mechanism for distributed channel access. Some work has modelled interference as a Gaussian process, but in this paper we show that for IEEE 802.11 this assumption is not accurate. We propose an alternative analytical model for the optimization of the carrier sense threshold for these networks that correctly predicts the best threshold value.

We propose the novel concept of cooperation for deafness prevention in directional antenna ad hoc networks, along with a low-complexity multiuser detector specialised for these systems. We call a generic MAC protocol that uses these techniques a cooperative-MAC (CMAC) and we define an upper and a lower bound on the performance of this class of protocols. Several simulations compare these bounds against the circular RTS/CTS protocol (CRCM), which is one of the finest deafness- suppression systems proposed so far, and show that our combined techniques achieve significant gains over CRCM in terms of throughput and energy consumption.

A well-known concept to adapt the transmitted pay- load to the current channel capacity is to perforin segmentation of the uncoded user payload. A new cross layer approach to hybrid ARQ was introduced that performs first channel coding and then segmentation of coded bits. In this paper, we denote the traditional approach as radio centric, while the latter is denoted as packet centric. By changing the order of coding and segmentation, it is possible to overcome the conflicting first requirements of link adaptation and hybrid ARQ in terms of packet size. Although the packet centric concept is not introduced here for the first time, a detailed performance analysis has been lacking so far. This paper presents a study that explores the characteristics of the packet centric approach in more detail and provides performance results and a comparison with the traditional approach. The new method is compared to the radio centric approach by means of a theoretical model and a cellular network simulator. It is shown that users with a weak channel can benefit from the former, leading to some gain in capacity.

In this paper, the problem of broadcast in ad hoc networks with smart antennas in a general propagation scenario is investigated. The conventional models proposed in the literature are surveyed and their limiting assumptions on the array at the receiver are identified. The connection between this point and the gain asymmetry for the broadcast of control packets is explored. As a first result, the impossibility to solve this problem in a conventional setting is proved. Then, a solution is mathematically deduced based on the consideration that a true MIMO (rather than MISO) system is required. The scheme can simultaneously send these packets farther, omnidirectionally and with very limited delay. Finally, a comparison with the Circular RTS scheme [1] is carried out with interesting results.

The problem of gain asymmetry in MIMO ad-hoc networks is outlined and the solutions proposed in literature are surveyed. The approach in R. Korakis et al. (2003) is analyzed and its delay quantified. A new solution, based on the ideas in F. Rossetto and M. Zorzi (2006), is studied. Its performance advantages over . Korakis et al. (2003) are quantified by physical layer and network level simulations