Distributed Average Consensus in Wireless Multi-Agent Systems with Over-the-Air Aggregation The work of T. Charalambous is partly funded by MINERVA, which received funding from the European Research Council (ERC) under the European Union’s Horizon 2022 research and innovation programme (Grant Agreem

Themistoklis Charalambous 1 , Zheng Chen 2 , and Christoforos N. Hadjicostis 1 · 2025

Read Distributed Average Consensus in Wireless Multi-Agent Systems with Over-the-Air Aggregation The work of T. Charalambous is partly funded by MINERVA, which received funding from the European Research Council (ERC) under the European Union’s Horizon 2022 research and innovation programme (Grant Agreem by Themistoklis Charalambous 1 , Zheng Chen 2 , and Christoforos N. Hadjicostis 1 (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2507-22648.txt.

Introduction

Themistoklis Charalambous, Zheng Chen, and Christoforos Hadjicostis let wireless agents agree on an average without a leader in this 2025 paper. Groups of sensors or robots often need the mean of their readings, computed while talking over the air. Sending every value to a hub wastes power and spectrum.

Their algorithm exploits the wireless channel itself: signals sent at once add up in the air. A modified ratio consensus, with a normalizing step to cancel channel quirks, turns that sum into agreement. Proofs show convergence for steady and changing channels when noise is low. Simulations back the math. The trick fits swarms that must stay in constant, cheap contact.

Catalog record

CREATORThemistoklis Charalambous 1 , Zheng Chen 2 , and Christoforos N. Hadjicostis 1
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE25 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2507-22648

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Distributed Average Consensus in Wireless Multi-Agent Systems with Over-the-Air Aggregation The work of T. Charalambous is partly funded by MINERVA, which received funding from the European Research Council (ERC) under the European Union’s Horizon 2022 research and innovation programme (Grant Agreement No. 101044629). Themistoklis Charalambous 1 , Zheng Chen 2 , and Christoforos N. Hadjicostis 1 Abstract In this paper, we address the average consensus problem of multi-agent systems over wireless networks. We propose a distributed average consensus algorithm by invoking the concept of over-the-air aggregation, which exploits the signal superposition property of wireless multiple-access channels. The proposed algorithm deploys a modified version of the well-known Ratio Consensus algorithm with an additional normalization step for compensating for the arbitrary channel coefficients. We show that, when the noise level at the receivers is negligible, the algorithm converges asymptotically to the average for time-invariant and time-varying channels. Numerical simulations corroborate the validity of our results. I Introduction The unprecedented interconnection of multi-agent systems we experience today necessitates efficient network management. One problem of particular interest over multi-agent systems is reaching consensus in a distributed fashion, i.e., reaching agreement among distributed agents towards a common decision. Average consensus constitutes a special case of consensus in which agents are initially endowed with numerical states and aim to calculate the average among these initial states. In classical consensus approaches, each node acquires separately the states of all the neighboring agents, whereas only a function (e.g., a linear combination) of the states is needed [ 1 ] . Acquiring the individual states over wireless networks increases the communication overhead and cost considerably and, in addition, requires advanced coordination in order to avoid interference and packet collisions. A key characteristic of the wireless channel is signal superposition, meaning that a receiving node can receive signals from multiple transmitting nodes simultaneously. In traditional wireless communication systems, concurrent transmissions from multiple sources are usually undesired due to the negative effect of interference on the decodability of the intended signal. On the other hand, signal superposition in multiple access channels (MACs) can also be exploited to perform function computation efficiently [ 2 ] . Inspired by this concept, over-the-air aggregation has recently attracted wide attention in distributed settings where the goal is to aggregate information from a large number of source nodes. Specifically, all nodes operate in the same frequency-time blocks and transmit analog signals carrying the information of their states, thus forming a waveform at the receiver side that corresponds to the aggregated information. There are limited existing works in the literature considering distributed consensus with over-the-air aggregation as the communication approach. Some earlier works have explored this direction without considering the fading effect [ 3 , 4 ] . In [ 5 , 6 ] , over-the-air aggregation over fading channels for \max -consensus is considered. 1 1 1 The \max -consensus algorithm is a distributed algorithm for computing the maximum value among the initial states of agents in multi-agent systems. Furthermore, in [ 7 ] , a weighted average consensus protocol is proposed under the same conditions. Even though consensus is reached, the protocol does not necessarily reach the average. Recently, a stochastic approximation-based (diminishing step size) distributed consensus protocol has been proposed in [ 8 ] , which guarantees convergence to the average in the mean square sense, for the case with noisy channels and complex-valued channel coefficients without phase alignment. However, the channel and noise statistics are assumed to be known in these works. A similar method with non-coherent processing was proposed in [ 9 ] for the distributed optimization setting. In this paper, we propose the Over-the-Air Ratio Consensus algorithm with which we can achieve the exact average consensus in both time-invariant and time-varying channels, when the noise level at the receivers is negligible [ 5 , 6 , 7 ] , but without requiring any knowledge about the channels. Our proposed method relies on the inherent and fundamental property of channel reciprocity [ 10 ] and employs a modified version of the well-celebrated Ratio Consensus algorithm [ 11 ] . II Network Model and Preliminaries II-A Notation The set of real (integer) numbers is denoted by \mathbb{R} ( \mathbb{Z} ) and the set of nonnegative numbers (integers) is denoted by \mathbb{R}_{+} ( \mathbb{Z}_{+} ). \mathbb{R}^{n}_{+} denotes the nonnegative orthant of the n -dimensional real space \mathbb{R}^{n} . Vectors are denoted by small letters whereas matrices are denoted by capital letters. The transpose of a matrix A is denoted by A^{\mbox{\tiny T}} . For A\in\mathbb{R}^{n\times n} , A_{ij} denotes the entry at row i and column j . By \mathbf{1} , we denote the all-ones vector, and by I we denote the identity matrix (of appropriate dimensions). II-B Network model We consider a wireless network consisting of n\in\mathbb{N} spatially distributed nodes, captured by a directed graph \mathcal{G}=(\mathcal{N},\mathcal{E}) , where \mathcal{N}=\{v_{1},\cdots,v_{n}\} is the set of nodes (representing the n agents) and \mathcal{E}\subseteq\mathcal{N}\times\mathcal{N} is the set of edges (representing the communication links between agents). The total number of edges in the network is denoted by m=|\mathcal{E}| . A directed edge \varepsilon_{ji}\triangleq(v_{j},v_{i})\in\mathcal{E} , where v_{j},v_{i}\in\mathcal{N} , indicates that node v_{j} can receive information from node v_{i} , i.e., v_{i}\rightarrow v_{j} . The nodes that transmit information to node v_{j} directly are called in-neighbors of node v_{j} , and belong to the set \mathcal{N}_{j}^{-}=\{v_{i}\in\mathcal{N}|\varepsilon_{ji}\in\mathcal{E}\} . The number of nodes in the in-neighborhood is called in-degree and it is represented by the cardinality of the set of in-neighbors, d_{j}^{-}=|\mathcal{N}_{j}^{-}| . The nodes that receive information from node v_{j} directly are called out-neighbors of node v_{j} , and belong to the set \mathcal{N}_{j}^{+}=\{v_{l}\in\mathcal{N}|\varepsilon_{lj}\in\mathcal{E}\} . The number of nodes in the out-neighborhood is called out-degree and it is represented by the cardinality of the set of out-neighbors, d_{j}^{+}=|\mathcal{N}_{j}^{+}| . Note that when self-loops are included in digraph \mathcal{G} , the number of in-coming links of node v_{j} is ( d_{j}^{-}+1 ) and similarly the number of its out-going links is ( d_{j}^{+}+1 ). II-C Ratio Consensus In [ 11 ] , an algorithm is suggested that solves the average consensus problem in a directed graph in which each node v_{j} distributively sets the weights on its self-link and outgoing-links to be p_{lj}=\frac{1}{1+d_{j}^{+}}\ \forall(v_{l},v_{j})\in\mathcal{E} , so that the resulting weight matrix P=[p_{lj}] (with P_{lj}=p_{lj} ) is column stochastic, but not necessarily row stochastic. Asymptotic average consensus is reached by using this weight matrix to run two iterations with appropriately chosen initial conditions. Proposition 1 ( [ 11 ] ) . Consider a strongly connected digraph \mathcal{G}(\mathcal{N},\mathcal{E}) . Let y_{j}[k] and x_{j}[k] (for all v_{j}\in\mathcal{N} and k=0,1,2,\ldots ) be the result of the iterations \displaystyle y_{j}[k+1]=p_{jj}y_{j}[k]+\sum_{v_{i}\in\mathcal{N}^{-}_{j}}p_{ji}y_{i}[k]\;, (1a) \displaystyle x_{j}[k+1]=p_{jj}x_{j}[k]+\sum_{v_{i}\in\mathcal{N}^{-}_{j}}p_{ji}x_{i}[k]\;, (1b) where p_{lj}=\frac{1}{1+d_{j}^{+}} for v_{l}\in\mathcal{N}_{j}^{+}\cup\{v_{j}\} (zero otherwise), and the initial conditions are {y_{j}[0]=S_{j}}\in\mathbb{R} and x_{j}[0]=1 , for v_{j}\in\mathcal{N} . Then, the solution to the average consensus problem can be asymptotically obtained as \displaystyle\lim_{k\rightarrow\infty}\mu_{j}[k]=\frac{\sum_{v_{j}\in\mathcal{N}}y_{j}[0]}{\sum_{v_{j}\in\mathcal{N}}x_{j}[0]}{=\frac{\sum_{v_{j}\in\mathcal{N}}S_{j}}{|\mathcal{N}|}}\;,\forall v_{j}\in\mathcal{N}\;, where \displaystyle\mu_{j}[k]=\frac{y_{j}[k]}{x_{j}[k]}\;. \lrcorner Remark 1 . Proposition 1 states a distributed algorithm with which the exact average is asymptotically reached, even if the directed graph is not balanced. \lrcorner Remark 2 . The algorithm described in Proposition 1 is for a specific choice of weights on each link that assumes that each node knows its out-degree. Note, however, that the

Read the complete text