Page 67 - ITU Journal Future and evolving technologies Volume 2 (2021), Issue 7 – Terahertz communications

ITU Journal on Future and Evolving Technologies, Volume 2 (2021), Issue 7

awake duration (or percentage of ) , but the beginning Density Estimator for Dense Networks (DEDeN) is an al‑

of the awake interval is different for each node, and is gorithm tailored to wireless nano‑networks. It allows a
randomly deter‑ mined. The mechanism ensures that if a node to estimate the number of its neighbours (also called
node is able to pick the irst bit of a packet, it will be able node degree, node density, neighbour density, neighbour‑
to pick all the following ones. More details on this hood density, local density, or simply density).
sleeping mechanism are described in [14].
Fig. 8 depicts an example, where receiver nodes Recv1,
DEDeN works by making nodes transmit an initialization
Recv2 and Recv3 wake up at different times, but for the
message, and all nodes that receive it start the discovery
same duration. Recv1 and Recv2 are able to pick the bits
process. With a very low (but known in advance) proba‑
from lows 1 and 2, as they arrive when they are awake.
bility, they may start sending small beacons. As the
Recv3 is able to pick bits only from low 3.
probability is known, nodes receiving those beacons can
easily compute an estimated local density. If the
con idence in the estimation in not high enough, nodes
start another round, with a higher beaconing
probability. The con idence increases with the number
of packets received and the probability to transmit.

The con idence and the error range of the estimation can
be adjusted to the requirements of the user with a pre‑
dictable overhead. Depending on how it is initiated, DE‑
DeN enables a unique node or all nodes to estimate the
neighbourhood density. The algorithm may be executed
each time this neighbourhood density is needed.
Fig. 8 – Sleep mechanism with three nodes and three lows.
For additional information about the density estimator al‑
The very high frequency of awake and sleep transitions
gorithm, refer to [15].
make it icantly different from a traditional duty‑
cycle. Fully turning on and off processing capabilities at
such a speed may not be technically achievable, but that
is not the main point of our protocol anyway. Instead, we 4. RETRANSMISSION ALGORITHM AT THE
just consider that nodes do not process bits (pulses in TS‑ DESTINATION ZONE
OOK) outside of their awaken duration.
As nodes do not turn off most of the hardware, the energy
saving is limited (but still de initely exists, given that As explained in the previous section, the proposed slee-
ignored frames are not processed and potentially ping mechanism should reduce the congestion and
retransmitted). preserve node resources consumption. When a packet
As we will see later, the bene it of this mechanism lies in arrives at the destination zone, if the destination node
an increase of the overall effective channel capacity. This is asleep, it will lose the data.
bene it becomes really signi icant as the local node den‑
sity increases and nodes can have individual lower and It is important to note that the proposed sleeping mecha‑
lower awaken durations.
nism was implemented and tested in a homogeneous net‑
work. In this paper we are now evaluating this protocol in
3.3 Density estimator algorithm an heterogeneous network (areas of different nodes den‑

Because nodes do not have much memory or processing sity). All nodes have the same awake‑sleep time , but
they are different in their awaken duration and awaken
power, elaborate routing protocols that try to ind opti‑
starting time.
mum forwarder(s) cannot work well if at all. Hopefully,
we can improve routing protocols by using the local den‑
sity of the network to limit their retransmission rate. We We propose an algorithm that increases the destination
need a way for the nodes to discover by themselves how node chance of receiving a packet even if it was asleep
many neighbours they have. when the packet reached the destination zone. The algo‑
A traditional solution to this problem is that each node rithm is to be used only by nodes at the destination zone.
sends ”HELLO” packets. All nodes would only need to No matter the network density, it never saturates the ra‑
maintain a list of the neighbours from which they have dio channel and does not require much memory or com‑
received an HELLO packet to know the local density. This putations. The only memory needed is the buffer to store
solution is simple and ef icient if the density is low, but the received packet to retransmit it at the end of the node
it becomes prohibitively costly in resources consumption awaken duration. A summary of the packet retransmis‑
as the density increases. sion algorithm is presented in Algorithm 1.

© International Telecommunication Union, 2021 55