Application of Swarm Intelligence in Internet of

Things
Tanmay Chakraborty* and Soumya Kanti Datta**
* Student, Adamas University and Intern, Future Tech Lab, India
** Future Tech Lab, India
Emails: [email protected] and [email protected]

Abstract—This paper looks into application of swarm in- to the Internet by 2020, which would generate around 1.2
telligence (SI) in the Internet of Things (IoT). SI algorithms zettabytes of data. Therefore, optimization of data intensive
are well established and applied in well known problems. The IoT processes using the distributed and decentralized control
IoT processes being data intensive, the quality of IoT services
depend a lot on IoT data processing capabilities. With diverse and self organization is highly desirable.
variety and volume of IoT data, SI algorithms like Ant Colony Utilization of swarm intelligence algorithms in IoT can
Optimization (ACO) can play an important role in optimizing make the IoT data processing much faster and efficient.
the IoT processes. We envision the application of ACO in three Potential applications of the algorithms include Connected
IoT domains and outline how they can be improved with the Cars, IoT network design for routing high volume of IoT
same.
data as well as an Edge Computing system that optimize IoT
Keywords-Internet of Things; Optimization; Swarm Intelli- data processing. This would give a boost to the emerging
gence. IoT ecosystems, improve the consumer experience of IoT
applications and services.
I. I NTRODUCTION
II. S WARM I NTELLIGENCE AND ITS A PPLICATIONS
Large number of simple organisms performing some simple
task in order to get a complex or greater task accomplished Swarm Intelligence (SI) is the collective behavior of a self-
with ease is what we understand by swarm intelligence. All organized and decentralized system that may be natural or ar-
the individuals coordinate using decentralized control and self- tificial intelligence. We briefly describe two most widely used
organization. The focus in mainly on local interactions of swarm intelligence algorithms that are biologically inspired.
the individuals with each other and their environment. Swarm A. Ant Colony Optimization
intelligence can be found widely in Colonies of ants, school
Ants are one of the best example of natural swarm. They
of fishes, honey bees etc. Using of a distributed form of
live in colonies. They lay pheromone (i.e volatile chemical
control makes a system more efficient, effective and scalable.
substance) on the way from their nest to the food source.
The main aim of swarm intelligence [1] is to increase the
Single ants do not have enough intellect to find shortest path
performance and robustness. Swarm algorithms are faster and
to their food. But when in colonies they can perform a lot
more robust solutions to solve complex set of problems.There
complex tasks with ease. Ant Colony Optimization (ACO) is
are different algorithms like artificial fish swarm algorithm,
based on the path optimization done by ant colonies while
dynamic optimization with which different SI systems were
finding the shortest route to the food source. ACO algorithms
built which could solve various problems. Till now, swarm
find its usage in optimization problems, traveling salesman
intelligence algorithms have been widely applied to robotics,
problem, scheduling problems, vehicle routing problems [3]
DNA computing, vehicle routing problem to name a few.
etc. Its application in SI is as a class of algorithm which
Recently, these algorithms are being applied to data intensive
inspires the forging behavior of the ant. The main iteration in
Internet of Things (IoT) operations and services.
the ant system ACO is the updating of the pheromone level.
The IoT is an umbrella of technologies comprising of
sensing, actuation, data processing, device management and B. Bee Colony Optimization
more that open new vistas for application and service develop- Bees are another good example of natural and dynamic
ment [2]. IoT services like environmental monitoring, fitness swarms. They display their intelligence by dividing their work
tracking, home appliance control, smart waste management among other bees. They perform tasks like foraging, storing,
are already being deployed in Smart Cities. The smart device honey distribution, collecting pollen, retrieving, communica-
or Things generate data in high volume, velocity, variety and tion and adapting themselves to changes. Bees are very good in
veracity leading to Big Data. They are usually transported to a organizing their colonies [4]. Following algorithm highlights
Cloud Computing Platform for analysis and long term storage. a bee colony functionalities -
It is estimated that 20-25 Billion Things will be connected 1.Bees by nature are organized and divided into groups of
978-1-5386-2189-9/17/$31.00 2017
c IEEE employed bees and unemployed bees.

978-1-5386-2189-9/17/$31.00 2017
c IEEE 67
 2.Unemployed bees scouts the employed bees when a food
backward interactions of agents that explore the network. The
source has been detected by the employed bees.
system is quite complex with a lot of network nodes (e.g. IoT
3.Employed bees move from one flower to another flower
devices) being involved. Each node will launch ants in forward
randomly and find the food source.
direction at regular intervals, the ant will find a path to the
4.Once food source is found the unemployed bees follow
Cloud randomly based on routing table. These ants creates a
find the shortest path to the food source and once they find
stack which is inherited by the backward ant after the forward
they inform the other bees by waggle dance.
ant reaches its destination (e.g. Cloud). Backward ant pops the
5.Worker bees collect the nectar from the source and return
stack entry and gets back via the same path. Node tables get
back to their comb and store the nectar.
updated based on trip time of the forward and backward ant.
III. A PPLICATION OF S WARM I NTELLIGENCE IN I OT
P ROCESSES C. SI in Cloud computing for Data Optimization
Swarm Intelligence with it’s optimization capabilities can The value of IoT lies in processing, storage and visualiza-
ameliorate IoT processes that are data intensive. Many novel tion of data. 1.2 zettabytes of IoT data are estimated to be
approaches for incorporating SI with different IoT vertical generated by IoT devices which would undoubtedly require
domains are presented in this paper. We have chosen three Cloud for storage. With such a large volume of data, data
domains where SI could have a better solution. They include - processing algorithms might potentially have a performance
(i) vehicle routing problem for connected cars with Ant Colony loss thereby increase in processing time. Utilizing ACO for
Optimization, (ii) data routing from a wide spread sensor data optimization to reduce the overall data processing time
network with ACO and (iii) Cloud Computing techniques for is what we suggest [6]. ACO offers large scale optimization
data optimization. for fast data mining on large scale datasets, it can handle high
dimensional data so that the performance of the algorithm does
A. SI in Connected Cars not decay with large datasets, capacity of handling dynamical
Vehicle Routing Problem (VRP) is a well studied logistic data which provides almost real-time data processing, and
problem in SI. The problem deals with the running cost min- multi-objective optimization in ACO can manage data coming
imization of running ”n” vehicles that has to serve ”m” cus- from different sources. This is an efficient and robust solution
tomers. Vehicle routing is an important service for connected for data optimization in IoT.
cars offered by IoT platforms. Such cars have capabilities of
exposing in-built sensor data to a secure Cloud platform over IV. C ONCLUSION
the Internet. Using the data the current platforms are offering Swarm Intelligence algorithms have been widely studied
services like shortest path to a fuel station, predictive analysis and applied to many well known problems. In this paper we
of car health and more. Such services are more important in examine the intersection of SI with the IoT. The later is an
the context of emerging autonomous vehicles. They would umbrella of several verticals which required data intensive
generate Gigabytes of data per second to (i) create Locally IoT processes to provide consumer services. We advocate
Dynamic Maps (LDM), (ii) logistics delivery, (iii) transporting for utilizing ACO and other algorithms to optimize the IoT
people, (iv) vehicle health analytics, (v) usage based insurance processes and make them more efficient. As a future work, we
and more. Without a human driver to take decisions, the plan to implement the three scenarios described in the paper
autonomous vehicles will be totally dependent on external and evaluate their performances.
computing platforms (e.g. Cloud) to process the data to create
meaningful information. Using ACO, novel IoT services can R EFERENCES
be developed to provide route construction, trail update and [1] T. Gui, C. Ma, F. Wang, and D. E. Wilkins, “Survey on swarm intelligence
route improvement strategies [3]. based routing protocols for wireless sensor networks: An extensive study,”
in 2016 IEEE International Conference on Industrial Technology (ICIT),
pp. 1944–1949, March 2016.
B. SI in Data Routing [2] S. K. Datta, C. Bonnet, and N. Nikaein, “An iot gateway centric
SI finds wide application in communication for its applica- architecture to provide novel m2m services,” in 2014 IEEE World Forum
on Internet of Things (WF-IoT), pp. 514–519, March 2014.
tion of routing in telecommunication networks. In the IoT, we [3] J. E. Bell and P. R. McMullen, “Ant colony optimization techniques for
propose to use this concept for routing data into a centralized the vehicle routing problem,” Advanced Engineering Informatics, vol. 18,
cloud from a large distributed sensor network. It is estimated no. 1, pp. 41 – 48, 2004.
[4] L. Dou, M. Li, Y. Li, Q. Y. Zhao, J. Li, and Z. Wang, “A novel artificial
that 20-30 Billion sensors will be connected to the Internet bee colony optimization algorithm for global path planning of multi-
by 2020 [2] and intelligent routing, storage of the sensor data robot systems,” in 2014 IEEE International Conference on Robotics and
will be required for the communication backbone. Use of SI Biomimetics (ROBIO 2014), pp. 1186–1191, Dec 2014.
[5] I. Kassabalidis, M. A. El-Sharkawi, R. J. Marks, P. Arabshahi, and A. A.
would ensure systematic data routing and storage which in turn Gray, “Swarm intelligence for routing in communication networks,” in
prevent data loss. We propose utilization of AntNet algorithm Global Telecommunications Conference, 2001. GLOBECOM ’01. IEEE,
[5] to achieve this. AntNet is a strong routing algorithm that vol. 6, pp. 3613–3617 vol.6, 2001.
has become popular after outperforming the other routing [6] S. Cheng, Y. Shi, Q. Qin, and R. Bai, Swarm Intelligence in Big Data
Analytics, pp. 417–426. Berlin, Heidelberg: Springer Berlin Heidelberg,
algorithms. In the AntNet, routing is done by forward and 2013.

68 2017 IEEE International Symposium on Consumer Electronics (ISCE).