Journal of Systems Engineering and Electronics
Vol. 28, No. 4, August 2017, pp.717 – 724
Uncertain bilevel knapsack problem and its solution
Junjie Xue
1,*
, Ying Wang
1
, and Jiyang Xiao
2
1. College of Equipment Management and Safety Engineering, Air Force Engineering University, Xi’an 710051, China;
2. Science Research Center, Air Force Engineering University, Xi’an 710051, China
Abstract:
This paper aims at providing an uncertain bilevel knap-
sack problem (UBKP) model, which is a type of BKPs involving
uncertain variables. And then an uncertain solution for the UBKP
is proposed by defining P
E
Nash equilibrium and P
E
Stackelberg
Nash equilibrium. In order to improve the computational efficiency
of the uncertain solution, several operators (binary coding dis-
tance, inversion operator, explosion operator and binary back
learning operator) are applied to the basic fireworks algorithm to
design the binary backward fireworks algorithm (BBFWA), which
has a good performance in solving the BKP. As an illustration, a
case study of the UBKP model and the P
E
uncertain solution is
applied to an armaments transportation problem.
Keywords: uncertainty, bilevel programming, knapsack problem,
binary backward fireworks algorithm.
DOI: 10.21629/JSEE.2017.04.11
1. Introduction
The bilevel knapsack problem (BKP) is a typical bilevel
programming which models the hierarchical relationship
between two autonomous, and possibly conflicting, deci-
sion makers: the leader and the follower [1]. BKP is first
studied by Dempe and Richter [2], and it is formulated as
a mixed-integer bilevel programming. Recently, more and
more researchers focused on the BKP and developed sev-
eral kinds of algorithm to solve the BKP, such as branch-
and-cut [1,2], approximation algorithm [3], dynamic pro-
gramming [4,5], heuristic optimization algorithm [6,7].
However, almost all of them are studied under the condi-
tion of deterministic assumptions.
However, parameters in the BKP are often indetermi-
nism. Traditionally, the probability theory and fuzzy the-
ory are applied to solve indeterminism [8]. As we know,
a premise of applying probability theory is that the ob-
tained probability distribution is close enough to the true
frequency. However, the plenty of practical sample data to
Manuscript received May 27, 2016.
*Corresponding author.
This work was supported by the National Natural Science Foundation
of China (71601183; 61502522).
generate the fitting probability distribution are always im-
possible. And then, the fuzzy theory may be applied by
inviting some domain experts to evaluate the belief degree
that each event happens. However, a lot of surveys show
that human beings usually estimate a much wider range of
values than the object actually takes. Considering all these
limitations, Liu declared that it was inappropriate to ap-
ply both the probability theory and fuzzy theory to handle
uncertainty, for both theories may lead to counterintuitive
results [9].
Mostly, the scale of sample data is always too small to
generate the probability distribution, but sufficient to ob-
tain the belief degree of these quantities. These kinds of
indeterminacy are neither like randomness nor like fuzzi-
ness, it is uncertain which is ubiquitous in reality [10], es-
pecially issues related to combat, such as the armaments
transportation problem. In order to deal with problems un-
der the uncertain condition, Liu found the uncertainty the-
ory, which provides the commonness of the probability
theory, credibility theory and chance theory to handle un-
certainty [11]. Parameters of the BKP are always uncertain
variables, and these kinds of BKPs are defined as the un-
certain BKP (UBKP).
In order to deal with the UBKP, an uncertain solution
is proposed by presenting P
E
Nash equilibrium and P
E
Stackelberg Nash equilibrium to transform the UBKP to
the corresponding equivalent deterministic model. Consid-
ering the BKP is NP-hard [12], a heuristic optimization
algorithm should be applied to improvecomputational effi-
ciency. As a relatively new member of swarm intelligence,
the fireworks algorithm (FWA) is proposed and enhanced
[13,14]. Due to its simplicity, effectiveness and efficiency,
FWA has been adopted by researchers in a variety of fields,
including variable-rate fertilization in oil crop production
[15], and digital filters design [16]. Considering the com-
binatorial character of the UBKP, a novel evolutionary al-
gorithm named binary backward FWA (BBFWA) is de-
signed, which uses three binary definitions (binary cod-
ing distance, inversion operator and explosion operator) to