Energy and bandwidth-efficient channel access for local area machine-to-machine communication

Abstract

Ticket Election Multiple Access (TEMA) is introduced for local machine-to-machine communication that is energy and bandwidth-efficient. TEMA is based on distributed elections held among nodes to gain interference-free access to the channel in either unicast or broadcast mode. Non-transmitting nodes can infer whether or not they are the intended receiver of a transmission and act accordingly to save energy, without the need for particular traffic patterns or explicit future transmission schedules. TEMA is shown to be correct in the sense that the channel access schedules are collision-free at the intended receivers, and intended receivers are always in receiving state. An analytical model of the performance of the protocol is used to show that TEMA attains energy-efficiency and high channel utilization even under heavy traffic and high node density conditions. A simulation-based performance analysis validates the analytical results and shows that TEMA outperforms representatives of contention-based and interference-free protocols in terms of energy efficiency, network goodput, and channel access delay. More specifically, it reduces energy consumption to half of that of state-of-the-art distributed election-based protocols while providing up to 25% increase in goodput and up to 50% decrease in channel access delay.

Appendix

We derive an expression for the expected value of the size of the closed two-hop neighborhood ($n_{2}$) of a wireless node when nodes are randomly placed over a bounded bi-dimensional region R following a homogeneous Poisson Point Process (PPP) with intensity $\lambda (x)=\lambda \ge 0$.

We further assume that the radio range of every node equals r, and hence, that the closed one-hop neighborhood of a node u located at position x is composed of the nodes in region $D(x,r) \subset R$, where D(x, r) denotes a disk of radio r with center at x. It is important to point out that this result holds for nodes located at a position x such that $D(x,2r) \subset R$.

From the previous assumptions, the expected value of the closed one-hop neighborhood of u equals $\lambda \pi r^2$.

On the other hand, a node v is a two-hop neighbor of u if it is in region $D(x,2r) - D(x,r)$ and it has a neighbor in D(x, r). By applying the thinning property of the PPPs, the number of two hop-neighbors of u can be described by a new PPP over region $D(x,2r) - D(x,r) \subset R$, with intensity $\lambda _\alpha (x') =\alpha (x') \lambda$, and where $1-\alpha (x')$ is the probability that a node located at $x' \in D(x,2r) - D(x,r)$ does not have a neighbor in D(x, r).

Let l be the distance between x and $x'$ with $r < l\le 2r$, this probability equals

$$\begin{aligned} p_N(0)=\frac{(\lambda \int _D {\mathrm {d}}l)^0}{0!} e^{-\lambda \int _D {\mathrm {d}}l}=e^{-\lambda \int _D {\mathrm {d}}l} \end{aligned}$$

where $D = D(x,r) \cap D(x',r)$.

The expected value of the new process can be computed using Eq. 23, where A(r, l) is the area of region D that can be computed using Eq. 24.

$$\begin{aligned}&\int _r^{2r} \lambda (1-e^{-\lambda A(r,l)})2 \pi l\, {\mathrm {d}} l \end{aligned}$$

(23)

$$\begin{aligned}&A(r,l) = r^2 \left[ 2\arctan \left( \frac{\sqrt{4r^2-l^2}}{l}\right) -\frac{l\sqrt{4r^2-l^2}}{2r^2} \right] \end{aligned}$$

(24)

Lastly,

$$\begin{aligned} n_{2}=\lambda \pi r^2 +\int _r^{2r} \lambda (1-e^{-\lambda A(r,l)})2 \pi l \, {\mathrm {d}}l \end{aligned}$$

(25)