Page 107 - ITUJournal Future and evolving technologies Volume 2 (2021), Issue 1

ITU Journal on Future and Evolving Technologies, Volume 2 (2021), Issue 1

Table 1 – MADM decision matrix. ing, Dempster‐Shafer theory, to name a few. Particularly,
Multi‐Attribute Decision Making (MADM) methods [15]
1 2 ... are commonly used for NIS. MADM methods are interest‐
1 2
1 11 12 ... 1 ing as they rank several alternatives, based on their at‐
tributes as well as the relative importance associated to
2 21 22 ... 2 those attributes.
... ... ... ... ... The problem can be modelled with a decision matrix as
1 2 ... shown in Table 1. It is composed of = { | =

1, 2, ..., } the set of the alternatives, = { | =

1, 2, ..., } the set of the attributes and = { | =

1, 2, ..., } the set of the weights associated to each at‐
tribute. Applied to NIS, is the set of technologies, the
set of attributes associated to those and the data re‐
quirements. The MADM methods take as input a decision
matrix and output a ranking of the alternatives. Several
MADM methods exist, the most known being: Simple Ad‐
ditive Weighting (SAW), Weighting Product (WP), Analyti‐
cal Hierarchy Process (AHP) and Gray Relational Analysis
(GRA).
One of the most used and studied methods is Technique
for Order Preference by Similarity to Ideal Solution (TOP‐
SIS) [16]. TOPSIS ranks alternatives depending on their
Fig. 1 – Representation of TOPSIS with three alternatives and two at‐ relative mathematical distance to the ideal solution. The
tributes. TOPSIS method runs the following steps:
sion (SCHC). This increases network lexibility, however
it requires a speci ic virtual network operator. While the 1. The values of each attribute from the decision ma‐

precedent work focuses on the device side, in [12] the au‐ trix (cf. Table 1) are normalized according to Equa‐
thors propose a cloud‐based virtual network operator for tion (1).
multi‐modal LPWA networks. This operator takes care of = (1)
the con iguration and management of heterogeneous LP‐ √∑ 2
WAN equipment. Again this requires speci ic infrastruc‐ =1
ture on the operator side.
2. The normalized values are weighted according to

In [13], a green path selection inter‐MAC selection pro‐ Equation (2).
tocol is detailed. This protocol allows path selection at
the MAC layer while focusing on energy consumption
and radio frequency minimization. However, it does not = , ∑ = 1 (2)

give any information about the routing layer. The arti‐ =1
cle [14] presents the ORCHESTRA framework which man‐
ages real‐time inter‐technology handovers. It is based on 3. The positive and negative ideal alternatives and
+
a virtual MAC layer which coordinates the different lay‐ are constructed according to Equation (3).
−
ers from different technology with a unique MAC address.
This work also focus on the link layer and not on the rout‐ = [ ... ]
+
+
+

ing layer. − 1 − − (3)
= [ ... ]

1
The aforementioned works increase WSN’s lexibility.
However several limitations are still present, such as the
need of a dedicated infrastructure. In this article we pro‐ 4. The attribute values of the ideal alternatives are de‐
pose a routing protocol adapted to MTN, which greatly termined according to Equation (4) for upward at‐
increase WSN’s capabilities while requesting only multi‐ tributes (e.q. range) or Equation (5) for downward
RAT nodes. attributes (e.q. latency).
+
= { , = 1, ..., }

3. TECHNOLOGY SELECTION BACKGROUND (4)
−
= { , = 1, ..., }

Multi‐technology devices have to autonomously select the
best communication technology based on many factors.
In the literature, several tools are available to perform this +
Network Interface Selection (NIS): utility and cost func‐ = { , = 1, ..., } (5)

−
tions, Markov chains, fuzzy logic, game theory, data min‐ = { , = 1, ..., }

© International Telecommunication Union, 2021 91