Transport Capacity of Distributed Wireless CSMA Networks

TL;DR

This paper demonstrates that distributed CSMA protocols can achieve the optimal transport capacity scaling law in large multi-hop wireless networks without centralized control, using percolation theory and parameter tuning.

Contribution

It shows that optimal capacity scaling can be achieved with fully distributed algorithms, providing bounds and validation through simulations.

Findings

Achieves throughput of Θ(1/√n) in distributed CSMA networks.

Provides upper and lower bounds for transport capacity.

Validates bounds with simulation results.

Abstract

In this paper, we study the transport capacity of large multi-hop wireless CSMA networks. Different from previous studies which rely on the use of centralized scheduling algorithm and/or centralized routing algorithm to achieve the optimal capacity scaling law, we show that the optimal capacity scaling law can be achieved using entirely distributed routing and scheduling algorithms. Specifically, we consider a network with nodes Poissonly distributed with unit intensity on a $\sqrt{n}\times\sqrt{n}$ square $B_{n}\subset\Re^{2}$. Furthermore, each node chooses its destination randomly and independently and transmits following a CSMA protocol. By resorting to the percolation theory and by carefully tuning the three controllable parameters in CSMA protocols, i.e. transmission power, carrier-sensing threshold and count-down timer, we show that a throughput of $\Theta\left(\frac{1}{\sqrt{n}}\right)$ is achievable in distributed CSMA networks. Furthermore, we derive the pre-constant preceding the order of the transport capacity by giving an upper and a lower bound of the transport capacity. The tightness of the bounds is validated using simulations.