Page 105 - ITU Journal Future and evolving technologies Volume 2 (2021), Issue 1

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

RODENT: A FLEXIBLE TOPSIS BASED ROUTING PROTOCOL FOR
MULTI‐TECHNOLOGY DEVICES IN WIRELESS SENSOR NETWORKS

1
Brandon Foubert and Nathalie Mitton 1
1 Inria, Lille, France

NOTE: Corresponding author: Brandon Foubert, brandon.foubert@inria.fr

Abstract – Wireless Sensor Networks (WSN) are icient tools for many use cases, such as environmental monitoring.
However WSN deployment is sometimes limited by the characteristics of the Radio Access Technologies (RATs) they use. To
overcome some of these limitations, we propose to leverage the use of a Multiple Technologies Network (MTN). What we refer
to as MTN is a network composed of nodes which are able to use several RAT and communicating wirelessly through multi‐hop
paths. The management of the RAT and routes must be handled by the nodes themselves, in a local and distributed way, with a
suitable communication protocol stack. Nodes may reach multiple neighbors over multiple RAT . Therefore, each stack’s layer
has to take the technologies’ heterogeneity of the devices into account. In this article, we introduce our custom Routing Over
Different Existing Network Technologies protocol (RODENT), designed for MTN. It enables dynamic (re)selection of the best
route and RAT based on the data type and requirements that may evolve over time, potentially mixing each technology over a
single path. RODENT relies on a multi‐criteria route selection performed with a custom lightweight TOPSIS method. To assess
RODENT’s performances, we implemented a functional prototype on real WSN hardware, Pycom FiPy devices. Unlike related
prototypes, ours has the advantage not to rely on speci ic infrastructure on the operator’s side. Results show that RODENT
enables energy savings, an increased coverage as well as multiple data requirements support.

Keywords – heterogeneous, Pycom FiPy, routing, TOPSIS, WSN

1. INTRODUCTION based on the routes’ availability and costs, in terms of en‐
ergy, money, etc. If the environment changes, and the se‐
Wireless Sensor Networks (WSN) enable a remote mon‐ lected route’s quality decreases, a node can dynamically
itoring of various metrics and many more use cases [1]. select a better route and RAT. Nodes that support several
Such networks usually rely either on a medium distance data requirements (e.g., temperature and video monitor‐
Radio Access Technology (RAT) (e.g., IEEE 802.15.4) and ing) can follow several paths accordingly. Network re‐
a multi‐hop path routing or on a long distance RAT (e.g., siliency is increased, as in case of a RAT failure, a node
LoRaWAN) and a star topology. The latter simpli ies the
can switch to an alternative technology.
network structure and enables a wider coverage. When
deployed, WSN usually use a single RAT shared by all Thus, nodes have to use speci ic methods to au‐
nodes. Deployments are thus constrained by the lim‐ tonomously and dynamically choose which technology
its of the chosen RAT, in terms of coverage and perfor‐ is the best suited depending on the data requirements
mance (throughput, energy consumption, costs, etc). For and current context. This issue is known as Network
instance, the network of Sigfox, an operator‐based RAT, Interface Selection (NIS). Several tools are available in
provides long range communication (up to km) but is not the literature to tackle the NIS problem. Among them
available worldwide. Some RATs are even so constrained are the Multiple Attribute Decision‐Making (MADM)
that they may not be able to comply with speci ic data re‐ methods. MADM methods provide a ranking of different
quirements such as delay‐intolerant data, high through‐ alternatives based on their attributes and their associated
put or irmware over‐the‐air upgrade. Additionally, out‐ weights. One of the most used and studied MADM meth‐
door nodes have to bear the weather changes (e.g., rain) ods is Technique for Order of Preference by Similarity to
which greatly impact the wireless links’ quality. Ideal Solution (TOPSIS). Said simply, TOPSIS compares
Traditional WSN lack lexibility to support multiple use‐ candidates based on their mathematical distances to two
cases. Many different RATs are available for WSN nowa‐ ideal positive and negative alternatives.
days [2]. Different RATs come with different perfor‐ However, TOPSIS suffers from an issue known as rank
mances and capabilities. Multiple Technologies Networks reversal. A rank reversal happens when the ranking is
(MTN) could overcome the aforementioned issues [3]. modi ied following the removing of one of the alterna‐
With several RATs built‐in, the nodes’ range of deploy‐ tives under study. This can alter the quality of the ranking
ment would be extended, as nodes could switch from and lead to a sub‐optimal NIS. In our case, this could out‐
one RAT to another at each hop and relay data through come in too many useless and costly technology switches.
multi‐hop. An MTN’s nodes would be able to select the Moreover, considering hardware constrained WSN nodes,
best technology and route available. The choice would be TOPSIS computation is resource‐intensive. This would

© International Telecommunication Union, 2021 89