Vol.:(0123456789) 1 3
Evolutionary Intelligence
https://doi.org/10.1007/s12065-018-0188-7
SPECIAL ISSUE
Particle swarm optimization with adaptive inertia weight based
on cumulative binomial probability
Ankit Agrawal
1
· Sarsij Tripathi
1
Received: 22 May 2018 / Revised: 20 October 2018 / Accepted: 31 October 2018
© Springer-Verlag GmbH Germany, part of Springer Nature 2018
Abstract
Particle swarm optimization (PSO) is a population oriented heuristic numerical optimization algorithm, infuenced by the
combined behavior of some birds. Since its introduction in 1995, a large number of variants of PSO algorithm have been
introduced that improves its performance. The performance of the algorithm mostly rely upon inertia weight and optimal
parameter setting. Inertia weight brings equivalence among exploitation and exploration while searching optimal solution
within the search region. This paper presents a new improved version of PSO that uses adaptive inertia weight technique
which is based on cumulative binomial probability (CBPPSO). The proposed approach along with four other PSO variants
are tested over a set of ten well-known optimization test problems. The result confrms that the performance of proposed
algorithm (CBPPSO) is better than other PSO variants in most of the cases. Also, the proposed algorithm has been evaluated
on three real-world engineering problems and the results obtained are promising.
Keywords Inertia weight · Particle swarm optimization (PSO) · Exploration and exploitation · Convergence
1 Introduction
The particle swarm optimization is a population oriented
meta-heuristic optimization technique which was frst pro-
posed by Russell Eberhart and James Kennedy [1]. This
algorithm is motivated from the social behavior of some ani-
mal groups like fsh schools or bird focks. Similar to other
meta-heuristic optimization algorithms, in PSO, a swarm of
possible solutions is derived in succeeding iterations. PSO
is relatively easy to understand and implement since there
are very few parameter settings that are required to tune in
comparison to other optimization strategies.
In PSO, each particle represents a prospective solution to
an optimization problem. The group of particles (or swarm)
fy within search space by trailing the current optimum solu-
tions. Every particle remembers its best coordinate position
in the search region which is related to the best position it
has attained so far. This position is called pbest. Similarly,
the best current position of the particle within whole swarm,
called as gbest is also remembered. In PSO, each particle
changes its velocity towards its gbest and pbest location after
each time step. The basic PSO [1] is slow in most cases and
prematurely converges to local optima. The solution to their
problem is use of inertia weight. The inertia weight [2] is
the primary parameter of the PSO, and has an important
part in balancing the exploration and exploitation. After the
introduction of inertia weight, many versions of the PSO
algorithm have been presented aiming balanced explora-
tion–exploitation trade-of.
Shi and Eberhart [2] examined diferent constant values
of inertia weight and reached to the conclusion that within
specifc range, PSO search the optimum global solution
within a acceptable number of iterations. Mostly with lower
value of inertia weight, PSO converges in local optima
region and with higher value of inertia weight; it diverges
and hence performs little exploration in the search region.
For fnding an optimum solution in a dynamic environment,
a random inertia weight strategy is used by Russell Eberhart
and Shi [3] and varies in the range [0.5, 1]. This approach is
used to alter exploitation and exploration randomly.
A large number of inertia weight variants use time-vary-
ing inertia weight method which utilizes iteration number to
determine the value of inertia weight. Most famous among
them is linear decreasing inertia weight technique [4], in
* Ankit Agrawal
aagrawal.phd2017.cse@nitrr.ac.in
1
Department of Computer Science and Engineering,
National Institute of Technology Raipur, G.E Road, Raipur,
Chhattisgarh 492001, India