Efficient and Scalable Distributed Autonomous Spatial Aloha Networks via Local Leader Election

TL;DR

This paper introduces a distributed algorithm for spatial Aloha networks where local leaders self-tune access probabilities to optimize throughput and stability using only local information, enabling scalable and efficient network operation.

Contribution

The paper proposes a novel local leader election algorithm that ensures network stability and near-optimal throughput with minimal information sharing.

Findings

The algorithm achieves close-to-Pareto-front throughput in simulations.

Network stability is maintained with R ≤ 2 at local leaders.

The method requires only local information, reducing complexity.

Abstract

This paper uses a spatial Aloha model to describe a distributed autonomous wireless network in which a group of transmit-receive pairs (users) shares a common collision channel via slotted-Aloha-like random access. The objective of this study is to develop an intelligent algorithm to be embedded into the transceivers so that all users know how to self-tune their medium access probability (MAP) to achieve overall Pareto optimality in terms of network throughput under spatial reuse while maintaining network stability. While the optimal solution requires each user to have complete information about the network, our proposed algorithm only requires users to have local information. The fundamental of our algorithm is that the users will first self-organize into a number of non-overlapping neighborhoods, and the user with the maximum node degree in each neighborhood is elected as the local leader (LL). Each LL then adjusts its MAP according to a parameter R which indicates the radio intensity level in its neighboring region, whereas the remaining users in the neighborhood simply follow the same MAP value. We show that by ensuring R less than or equal to 2 at the LLs, the stability of the entire network can be assured even when each user only has partial network information. For practical implementation, we propose each LL to use R=2 as the constant reference signal to its built-in proportional and integral controller. The settings of the control parameters are discussed and we validate through simulations that the proposed method is able to achieve close-to-Pareto-front throughput.