Logo: to the web site of Uppsala University

Suppose sender-receiver transmission links in a downlink network at a given data rate are subject to fading, path loss, and inter -cell interference, and that transmissions either pass, suffer loss, or incur retransmission delay. We introduce a method to obtain the average activity level of the system required for handling the buffered work and from this derive the resulting coverage probability and key performance measures. The technique involves a family of stationary buffer distributions which is used to solve iteratively a nonlinear balance equation for the unknown busy -link probability and then identify throughput, loss probability, and delay. The results allow for a straightforward numerical investigation of performance indicators, are in special cases explicit and may be easily used to study the trade-off between reliability, latency, and data rate.

We consider a cellular network equipped with aretransmission mechanism where the signal transmissions in asingle cell depends on interference with simultaneous traffic insurrounding cells. In this framework where failed signals areeither retransmitted or lost we study the channel capacity per-formance of a single-tier model with downlink or uplink traffic.For this purpose, a tractable model that allows for a precisetheoretical analysis of coverage probability and coverage rate isdeveloped further. Specifically, we obtain the Shannon capacity inthe model and introduce relevant performance measures to guidein the identification of those systems which in a precise sense areable to process all incoming work. To emphasize the genericpatterns that arise we extend and simplify the results under ascaling regime of balanced densification, which highlights thatperformance essentially falls into three categories, for pure-loss,buffered, and no-loss systems.

Multi-tier cellular networks able to handle multiple classes of base stations with varying transmission power and coverage area offer a promising approach to meet increasing traffic demands by properly balancing the input load. This paper investigates the performance of buffered multi-tier networks with randomly placed transmission nodes . Under simplifying model assumptions of Rayleigh fading and unbounded attenuation, we propose an approach to derive the multi-tier coverage probability and Shannon capacity of typical cells. The goal is to study the trade-off between base station density and interference effects in multitier compared to single-tier networks by assessing the system performance under varying policies. We also use stochastic simulation to verify the theoretical results and visualize the system behavior under different parameter settings.

In this work we consider an uplink cellular network with the focus on a typical cell rather than the whole network. The base stations (BSs) and the users are distributed according to Poisson point processes (PPP) and the signals are transmitted at random power. The BSs’ serving area is formed according to the Voronoi diagram and the users are associated with a serving BS based on the shortest distance. One of the features of the system is that we primarily take into account the interference inside a d-dimensional ball of the average size of a typical Voronoi cell. In this work we mainly focus on the system stability and discuss a necessary stability condition, which is then studied by using stochastic simulation. We also discuss some properties of the network that can affect the stability and appear to be interesting and promising for the performance analysis of the system.

We consider an uplink cellular network with static users and unlimited retransmissions. The users are assigned to the base stations (BSs) using the shortest distance association policy. The network cells are formed according to the Voronoi tessellation, and we study stability of this model with focus on a single cell. In particular, we consider a model with non-homogeneous users where the buffer size of each user depends on the number and locations of the active users at each time slot. We obtain a basic relation between input and output rate (coverage probability) of each user in steady-state regime. Moreover, we use stochastic simulation to verify sufficient stability conditions (obtained in the paper [8] for a more general system) which are reformulated in terms of the model under consideration. In particular, we find that these conditions turn out to be quite close to stability criteria in the most realistic case of the heavily loaded cell. In this regard and because of analytical unavailability of some metrics, we empirically study the convergence of the stability zone of the lightly-loaded cell to the zone defined by the sufficient stability conditions, when the load increases.

In this work, we investigate the stability of a class of n-dimensional discrete-time Markov chains with state space Nn and monotonic drift conditions. While exact stability criteria are well understood for one- and two-dimensional cases, extending these results to higher dimensions remains an open problem. To address this, we present an analytical approach to establish upper and lower bounds for the stability region for n ≥ 2. A key feature of the considered Markov chain is the monotonicity of the drift vector, which not only reduces the analytical complexity but also reflects realistic dynamics observed in practical applications, such as communication networks. Our analysis uses the Foster–Lyapunov criterion and generalizes the existing stability results to higher-dimensional settings. To illustrate the bounds obtained, we also provide numerical examples. 

This paper proposes a mean-field multi-agent reinforcement learning (MARL) framework for optimizing transmission control in buffered cellular networks. Each base station is modeled as an autonomous agent with finite queuing capacity, interacting with neighboring stations through interference on a Voronoi-based network topology. To address scalability issues in dense networks, a mean-field approximation is used so that agents respond to the average behavior of their neighbors rather than to full global states. A mean-field Q-learning algorithm and a corresponding reward function are derived to jointly balance Shannon capacity, buffer occupancy, and delay. Convergence of the learning dynamics is formally proved, and performance is evaluated via greedy,  tabular, and deep Q-network (DQN) approaches. Simulation results show that the proposed implementation significantly lowers delays and signal losses, and hence achieves better overall performance.