Int j simul model 20 (2021) 1, 201-211

Original scientific paper

AN IMPROVED GENETIC ALGORITHM FOR RESOURCE-

CONSTRAINED FLEXIBLE JOB-SHOP SCHEDULING
Wei, F. F.#; Cao, C. Y. & Zhang, H. P.
School of Management and Economics, North China University of Water Resources and
Electric Power, Zhengzhou 450046, China
E-Mail: [email protected] (# Corresponding author)
Abstract
Flexible job-shop scheduling could effectively lower the costs of manpower and materials. However,
there is little report on the scheduling algorithm or optimization model for the optimization of production
resources. This paper proposes a resource-constrained flexible job-shop scheduling algorithm based on
an improved genetic algorithm. Firstly, an optimization model was established for resource-constrained
FJSP, together with the objective functions about resources and time, as well as constraints. Next, the
multi-objective genetic algorithm (MOGA) was combined with the whale optimization algorithm
(WOA) into a combinatory method to solve the proposed model for resource-constrained FJSP.
Experimental results show that the combination enhances the adaptivity of crossover and mutation
probabilities, and improves the local search ability, presenting an effective solution to the FJSP.
(Received in December 2020, accepted in February 2021. This paper was with the authors 1 month for 1 revision.)

Key Words: Multi-Objective Genetic Algorithm (MOGA), Resource Constraints, Flexible Job-Shop

1. INTRODUCTION
For modern manufacturing enterprises, the most important means to enhance competitiveness
and gain an advantage in commercial competition include reducing the makespan and lowering
the production cost [1-4]. Relying on personal experience, the traditional job-shop scheduling
principles like first in, first out (FIFO), and shortest job first have not achieved the ideal effect
[5, 6]. Thus, the solution to the flexible job-shop scheduling problem (FJSP) has attracted wide
attention at home and abroad [7-10].
The current research on FJSP mainly deals with the algorithm design and optimization of
the scheduling problem, and the analysis and modelling of manufacturing energy consumption
[11-15]. Wang et al. [16] set the minimal maximum makespan as the objective of job-shop
scheduling, proposed a 5/3 approximate sorting algorithm and the lower limit of the optimal
solution, and discussed the properties of the optimal solution of the order-constrained graph.
Largescale customized production involves complex operations and machines [17, 18].
Morinaga et al. [19] improved the particle swarm optimization (PSO) based on Bayesian
network, which satisfies the needs of solving largescale FJSPs. Kurniawati and Nugroho [20]
took the minimization of maximum makespan, standby energy consumption, and power cost as
the optimization goals, constructed a fuzzy opportunity-constrained programming model for
remanufacturing job-shops, and solved dynamic events like machine failure.
At this stage, there are relatively few studies on the comprehensive scheduling algorithm of
the FJSP [21, 22]. For the multi-machine job-shop scheduling problem, Hammami and Stützle
[23] presented an operation priority sorting strategy and a machine-driven strategy with delay
constraints, which effectively increases the parallelism of processing supply and demand. Viana
et al. [24] put forward a waiting strategy that suspends eligible jobs based on batch optimization.
Kundakcı and Kulak [25] constructed the objective function of the batch scheduling, established
the corresponding fuzzy mathematical model, and solved the batch scheduling problem based
on the hybrid meta-heuristic algorithm.
Many have attempted to optimize the energy consumption of single strategy optimization
and integrated optimization. For instance, Zhu et al. [26] proposed a linear optimization

https://doi.org/10.2507/IJSIMM20-1-CO5 201
 Wei, Cao, Zhang: An Improved Genetic Algorithm for Resource-Constrained Flexible …

algorithm for operation planning, while Zhou et al. [27] developed a genetic algorithm-based
method for optimizing the idle energy consumption of machines. The existing evolutionary
solution algorithms for FJSP mainly optimize the running speed and optimization efficiency
via genetic operations, encoding and decoding, and fitness setting. But there is little report on
the scheduling algorithm or optimization model for production resources.
This paper proposes a resource-constrained flexible job-shop scheduling algorithm based
on an improved genetic algorithm, which combines the multi-objective genetic algorithm
(MOGA) with the whale optimization algorithm (WOA). Experimental results show that our
algorithm enhances the adaptivity of crossover and mutation probabilities, and improves the
local search ability.

2. CONSTRUCTION OF OPTIMIZATION MODEL

2.1 Resource and time constraints
The production resources of FJSP can be divided into renewable resources, partially renewable
resources, non-renewable resources, and restricted resources. Despite having an upper limit on
supply, the renewable resources are supplemented as the operations are implemented. Let RRsU
be the upper limit on the supply of type s renewable resource, and DRisU be the demand for type
s renewable resource in the ith operation. Then, the constraint of the renewable resource is:
 DR
iPt
U
is  RRsU , s,t (1)
where, Pt is the set of operations being implemented in period t. The total amount of non-
renewable resources is determined at the start of the production in the flexible job-shop. But
the non-renewable resources will not be supplemented as the operations are implemented
continuously. Let NRRvT be the total amount of type v non-renewable resource, and DNRivT be
the demand for type v non-renewable resource in the ith operation. Then, the constraint of the
non-renewable resource is:
V

 NDR
v =1
T
iv  NRRvT , v (2)

For resource-constrained FJSP, the optimization is often constrained by the time window,
that is, the start time of an operation is limited in a fixed time range. Let Δtijmin and Δtijmax be
the minimum and maximum differences between the start times of the ith and the jth operations,
respectively. Then, the time window constraint is:
si + tijmin  s j  si + tijmax (3)
where, si and sj are the start times of the ith and the jth operations, respectively.
2.2 Resource and time objective functions
Let COSTs(RRs) be the cost function of type s production resource, with RRs be the
replenishment of type s renewable resource. Then, the cost minimization function is:
S
min  COSTs (RRs ) (4)
s =1

The total demand for type s resource in period t is:

RDst =  DRis (5)
iPt

The mean demand for the type of resource in the production cycle is:

202
 Wei, Cao, Zhang: An Improved Genetic Algorithm for Resource-Constrained Flexible …

TT

 RD st
(6)
RDs = t =1
TT
where, TT is the total time of the production cycle. The minimal resource cost can be defined
as:
( )
TT
min  RDst − RDs (7)
t =1

If multiple kinds of production resources need to be scheduled throughout the production

cycle, different weights should be assigned to each type of resource. Thus, the objective
function can be written as:
( )
S TT
min  ωs  RDst − RDs
2
(8)
s =1 t =1

As for the objective function of time, this paper chooses four objectives for optimizing the
resource-constrained FJSP, namely, makespan, makespan expectation, operation delay, and
cycle delay. Let ei be the completion time of the ith operation. Then, the minimization of the
makespan is:
TT
min  ei (9)
t =1

The minimization of makespan expectation, that is, minimizing the difference between the
actual completion time ei and the latest completion time e'i of an operation, can be defined as:
min ei − ei (10)
After assigning weights to different operations by importance, the minimization of the
weighted operation delay is:
M
min  ωi max0,ei − ei (11)
i =1

The minimization of the cycle delay is:

 TT

min max 0,  e − ei (12)
 t =1 
where, e' is the latest completion time of the entire production cycle.
2.3 Mathematical modelling of scheduling
As shown in Fig. 1, the optimization of resource-constrained FJSP is a directed graph
DG = (P, R). Let M be the number of operations, and P be the set of operations. The start time,
completion time, and makespan of the ith operation satisfies the equation ei = si + Δei. Fig. 2
shows the priority relations of four kinds of operations: start-start, end-start, end-end, and start-
end. Then, the set of priority relations R in the directed graph DG can be established.

Figure 1: Structure of a node-type directed graph network.

203
 Wei, Cao, Zhang: An Improved Genetic Algorithm for Resource-Constrained Flexible …

Figure 2: Priority relations of four kinds of operations.

For the optimization model for resource-constrained FJSP, the minimization of production
cycle can be defined as:
TT
OF1: min  ei (13)
t =1

Eq. (13) needs to satisfy the following constraints:

 si  max er , i
 rRi
 DRU  RRU , s,t

(14)
is s
iPt
Eq. (14) specifies that all the operations that have priority relations with the jth operation
must be completed earlier than the start time of the jth operation, and the demand of this
operation for a renewable resource must be lower than the upper limit on the supply of that
resource. Hence, a decision variable can be constructed as:
1, if the 𝑖 th operation is completed in period 𝑡
𝛿𝑖𝑡 = { (15)
0, otherwise
The optimization model for resource-constrained FJSP can be updated as:
TT
OF2 : min  ei (16)
t =1

Eq. (16) needs to satisfy the following constraints:

(17)

Eq. (17) limits the start time of each operation, the demand for renewable resources, and
the feasible range of the decision variable.

3. PROBLEM SOLVING
The current MOGA can effectively shorten the time of problem-solving, while outputting
diverse solutions. However, the algorithm is, to a certain extent, limited in the adaptation to
crossover and mutation probabilities, and the local search ability. This paper combines MOGA
with WOA to solve the resource-constrained FJSP.
Fig. 3 explains the coding method of the MOGA improved for solving the resource-
constrained FJSP. Let CP and MP be the crossover and mutation probabilities of MOGA,
respectively. Then, the CP and MP values were adjusted based on fitness. Through dynamic
control of crossover and mutation operators, the optimal individuals with high fitness were
preserved to obtain the set of pareto optimal solutions.

204
 Wei, Cao, Zhang: An Improved Genetic Algorithm for Resource-Constrained Flexible …

Figure 3: Coding method of improved MOGA.

This paper improves the fixed CP and MP into adaptive CP and MP. Specifically, the
maximums (CPmax and MPmax) and minimums (CPmin and MPmin) of CP and MP were selected,
and the adaptive CP and MP would change adaptively to fitness. Let Fitnessi be the fitness of
the ith individual, and Fitnessmax and Fitnessmin be the maximum and minimum fitness,
respectively. Then, the crossover probability of the ith individual can be computed by:
CPi = CPmin − (Fitnessi − Fitnessmin )/ ( + (Fitnessmax − Fitnessmin ))  (CPmax − CPmin ) (18)
where, ε is a small real number that prevents the denominator from being zero. The mutation
probability of the ith individual can be calculated by:
MPi = MPmin − (Fitnessi − Fitnessmin )/ ( + (Fitnessmax − Fitnessmin ))  (MPmax − MPmin ) (19)

Figure 4: Generation of the new population based on elite retention strategy.

As shown in Eqs. (18) and (19), if an individual has a higher fitness than other individuals
in the population, then it will have small CPi and MPi, that is, it is difficult to undergo crossover
and mutation. To avoid damages to the set of optimal solutions, such an individual was directly
preserved to the next generation. Fig. 4 explains the generation of the new population.
The WOA consists of three phases, namely, encircling prey, bubble-net attacking, and
search for prey. Let W = (W1, W2, …, WN) be the position of each whale in an N-dimensional
solution space. For each whale, the probability of encircling prey is equal to that of bubble-net
attacking: EP = AP = 0.5. Based on the principle of best search agency, it was assumed that the
position from which the ith whale chooses to swim towards the target prey as the position of the
whale closest to the prey Pbest. Then, the position of the ith whale can be updated by:
Pi t +1 = Pbest
t
− A  C  Pbest
t
− Pi t (20)
where, t is the number of current iteration; A and C are coefficient vectors, in which each
dimension consists of random numbers from a certain range. Suppose γ decreases linearly from
2 to 0 with the growth in t. Then, A and C can be calculated by:

205
 Wei, Cao, Zhang: An Improved Genetic Algorithm for Resource-Constrained Flexible …

 A = 2  1 − 
 (21)
 C = 2  2
where, μ1 and μ2 are both random vectors in [0, 1]. The value of A determines when a whale
chooses to swim towards the optimal target, and when it selects a target. If the modulus of A is
smaller than 1, then the whale will swim towards the optimal individual; if the modulus of A is
equal to or greater than 1, then the whale will swim towards a random individual. Fig. 5 shows
all possible positions of the whale moving from the current position (CLx, CLy) to the target
position (TLx, TLy) in the two-dimensional space, when the modulus of A is smaller than 1.

Figure 5: Positions of a whale encircling the prey.

During bubble-net attacking, the trajectory of a whale is a shrinking circle about the prey
(Fig. 6). In this phase, the position of the whale can be updated by:
Pi t +1 = Pbest
t
− Pi t  el  cos(2 l ) + Pi t +1 (22)
where, υ is a constant defining the shape of a logarithmic spiral; l is a random number uniformly
distributed in [-1, 1]. Fig. 7 shows how our algorithm solves the resource-constrained FJSP.

Figure 6: Trajectory of a whale in bubble-net attacking.

206
 Wei, Cao, Zhang: An Improved Genetic Algorithm for Resource-Constrained Flexible …

Figure 7: Flow chart of our algorithm.

4. EXPERIMENTS AND RESULTS ANALYSIS

To demonstrate its effectiveness, our algorithm for resource-constrained FJSP was tested with
the actual production data from the processing shop of a carmaker as the benchmark example.
The example contains 15 jobs and 6 machines. The improved MOGA was compared with the
original MOGA to reflect the reasonability of the proposed hybrid algorithm. As shown in Fig. 8
improved algorithm had a certain advantage in both optimization speed and the quality of the
initial population.
Next, our algorithm was subject to a comparison experiment under different pareto optimal
fronts. The initial population size, number of genetic iterations, initial number of whales, and
number of generations were all set to 100; the maximum and minimum crossover probabilities
were set to 0.6 and 0.9, respectively; the maximum and minimum mutation probabilities were
set to 0.06 and 0.25, respectively.

207
 Wei, Cao, Zhang: An Improved Genetic Algorithm for Resource-Constrained Flexible …

a)Before improvement b)After improvement

Figure 8: Maximum makespan curves obtained by the original and improved MOGAs.

a) b)

c) d)

Figure 9: Performance comparison between original and improved MOGAs.

As shown in Figs. 9 a and 9 b, after 100 iterations, the set of pareto optimal solutions
obtained by the original algorithm was far away from the actual pareto front, while that obtained
by the improved algorithm approximated that front. In Table I, the reversal distance of the
improved algorithm was smaller than that of the original algorithm by 0.00087893, while the
hypervolume index of the former was larger than that of the original algorithm by 0.02193. This
means the improved algorithm outputs better solution set, converges to more diverse solutions,
and achieves better convergence.
Without changing other parameters, the original and improved MOGAs were tested on
different test functions. The relevant results are displayed in Figs. 9 c and 9 d and Table I. It can
be seen that the set of pareto optimal solutions obtained by the improved algorithm was closer
to the actual pareto front than that obtained by the original algorithm. In Table I, the reversal

208
 Wei, Cao, Zhang: An Improved Genetic Algorithm for Resource-Constrained Flexible …

distance of the improved algorithm was smaller than that of the original algorithm by
0.0013196, while the hypervolume index of the former was larger than that of the original
algorithm by 0.04013. The results confirm the effectiveness of the improved algorithm.
Table I: Performance of original and improved algorithms.
Test number Algorithm Reversal distance Hypervolume index
Before improvement 0.00156227 0.65276
1
After improvement 0.00068334 0.67469
Before improvement 0.00184162 0.28140
2
After improvement 0.00052196 0.32153

Table II: Optional machines and makespans.

Optional machines and makespans
Job Operation
E1 E2 E 3 E4 E5 E6 E7
1 3 - 4 - 2 1 2
1 2 - 4 - 2 5 3 1
3 3 - - 2 4 2 -
1 - 2 - 2 3 2 -
2 2 - 4 5 - 2 1 4
3 3 - 4 1 - 5 7
1 4 - 3 - 2 3 3
3 2 3 - 5 2 - 4 -
3 3 - 6 4 - 8 6
1 3 4 5 - 4 - 7
4 2 - 3 6 3 5 - -
3 2 - 2 2 - 6 1
1 - 3 - 2 5 6 -
5
3 3 - 4 5 - 5 3

Based on the said benchmark example (Table II), the production data on 5 jobs, 7 machines,
and 14 operations were sorted out. Without changing the parameter setting or test parameters,
the proposed improved algorithm was implemented 10 times consecutively. The optimal Gantt
chart is shown in Fig. 10. It can be seen that the improved algorithm output the minimum
makespan at 11.75 h, shorter than that obtained by the original algorithm (13.42 h), i.e., it is
superior in solving resource-constrained FJSP.

Figure 10: Gantt chart for resource-constrained FJSP.

209
 Wei, Cao, Zhang: An Improved Genetic Algorithm for Resource-Constrained Flexible …

5. CONCLUSIONS
This paper designs a resource-constrained flexible job-shop scheduling algorithm based on an
improved genetic algorithm. The ability of our algorithm to solve multiple objectives was tested
under different pareto optimal fronts. The algorithm was found to output better solution set,
converge to more diverse solutions, and achieve better convergence than the original algorithm.
Besides, the optimal Gantt chart was plotted with the actual production data, which verifies the
superiority of our algorithm in solving resource-constrained FJSP.

ACKNOWLEDGEMENTS
This paper is supported by Postgraduate Education Reform Project of Henan Province, China
(2019SJGLX045Y), and Postgraduate Education Reform and Quality Improvement Project of Henan
Province, China (HNYJS2018KC02).

REFERENCES
[1] He, P. (2018). Optimization and simulation of remanufacturing production scheduling under
uncertainties, International Journal of Simulation Modelling, Vol. 17, No. 4, 734-743,
doi:10.2507/IJSIMM17(4)CO20
[2] Nazifa, T. H.; Ramachandran, K. K. (2019). Information sharing in supply chain management: a
case study between the cooperative partners in manufacturing industry, Journal of System and
Management Sciences, Vol. 9, No. 1, 19-47, doi:10.33168/JSMS.2019.0102
[3] Liao, J.; Lin, C. (2019). Optimization and simulation of job-shop supply chain scheduling in
manufacturing enterprises based on particle swarm optimization, International Journal of
Simulation Modelling, Vol. 18, No. 1, 187-196, doi:10.2507/IJSIMM18(1)CO5
[4] Ikechukwu, A. J. (2019). Assessment of organizational performance of private manufacturing
companies: the impact of supply chain management responsiveness, Journal of System and
Management Sciences, Vol. 9, No. 3, 26-44, doi:10.33168/JSMS.2019.0302
[5] Bustos-Tellez, C. A.; Tenjo-García, J. S.; Figueroa-García, J. C. (2018). Solving job-shop
scheduling problems with fuzzy processing times and fuzzy due dates, Medina, J. (Ed.),
Information Processing and Management of Uncertainty in Knowledge-Based Systems, Springer,
Cham, 790-799, doi:10.1007/978-3-319-91473-2_66
[6] Li, H.; Cao, B.; Zhu, H. (2017). A variable neighborhood migrating birds optimization algorithm
for flexible job shop scheduling, International Journal of Performability Engineering, Vol. 13, No.
7, 1020-1029, doi:10.23940/ijpe.17.07.p3.10201029
[7] Mhasawade, S. N.; Bewoor, L. A. (2017). A survey of hybrid metaheuristics to minimize makespan
of job shop scheduling problem, Proceedings of the 2017 International Conference on Energy,
Communication, Data Analytics and Soft Computing, 1957-1960, doi:10.1109/
ICECDS.2017.8389792
[8] Mishra, S. K.; Bose, P. S. C.; Rao, C. S. P. (2017). An invasive weed optimization approach for
job shop scheduling problems, The International Journal of Advanced Manufacturing Technology,
Vol. 91, No. 9, 4233-4241, doi:10.1007/s00170-017-0091-x
[9] Nouri, H. E.; Driss, O. B.; Ghédira, K. (2016). Simultaneous scheduling of machines and transport
robots in flexible job shop environment using hybrid metaheuristics based on clustered holonic
multiagent model, Computers & Industrial Engineering, Vol. 102, 488-501, doi:10.1016/
j.cie.2016.02.024
[10] Teekeng, W.; Thammano, A.; Unkaw, P.; Kiatwuthiamorn, J. (2016). A new algorithm for flexible
job-shop scheduling problem based on particle swarm optimization, Artificial Life and Robotics,
Vol. 21, No. 1, 18-23, doi:10.1007/s10015-015-0259-0
[11] Selmair, M.; Claus, T.; Trost, M.; Bley, A.; Herrmann, F. (2016). Job shop scheduling with flexible
energy prices, Proceedings of the 2016 European Council for Modeling and Simulation, 488-494,
doi:10.7148/2016-0488

210
 Wei, Cao, Zhang: An Improved Genetic Algorithm for Resource-Constrained Flexible …

[12] Rochow, P.; Böning, C.; Prinzhorn, H. (2015). Qualitätsorientierte belegungsplanung, Zeitschrift
fuer Wirtschaftlichen Fabrikbetrieb, Vol. 110, No. 10, 599-602, doi:10.3139/104.111406
[13] Shen, Z.; Burnham, K. J.; Smalov, L. (2015). Optimised job-shop scheduling via genetic algorithm
for a manufacturing production system, Selvaraj, H.; Zydek, D.; Chmaj, G. (Eds.), Progress in
Systems Engineering, Advances in Intelligent Systems and Computing, Vol. 366, Springer, Cham,
89-92, doi:10.1007/978-3-319-08422-0_13
[14] Phanden, R. K.; Jain, A. (2015). Assessing the impact of changing available multiple process plans
of a job type on mean tardiness in job shop scheduling, The International Journal of Advanced
Manufacturing Technology, Vol. 80, No. 9, 1521-1545, doi:10.1007/s00170-015-7123-1
[15] Driss, I.; Mouss, K. N.; Laggoun, A. (2015). A new genetic algorithm for flexible job-shop
scheduling problems, Journal of Mechanical Science and Technology, Vol. 29, No. 3, 1273-1281,
doi:10.1007/s12206-015-0242-7
[16] Wang, C. L.; Rong, G.; Weng, W.; Feng, Y. P. (2015). Mining scheduling knowledge for job
shop scheduling problem, IFAC-PapersOnLine, Vol. 48, No. 3, 800-805, doi:10.1016/
j.ifacol.2015.06.181
[17] Mokhtari, H.; Dadgar, M. (2015). Scheduling optimization of a stochastic flexible job-shop system
with time-varying machine failure rate, Computers & Operations Research, Vol. 61, 31-45,
doi:10.1016/j.cor.2015.02.014
[18] Yazdani, M.; Zandieh, M.; Tavakkoli-Moghaddam, R.; Jolai, F. (2015). Two meta-heuristic
algorithms for the dual-resource constrained flexible job-shop scheduling problem, Scientia
Iranica, Vol. 22, No. 3, 1242-1257
[19] Morinaga, E.; Nakamura, T.; Wakamatsu, H.; Arai, E. (2017). A method of flexible job-shop
scheduling for highly-distributed manufacturing systems, Proceedings of the 24th International
Conference on Production Research, 592-597, doi:10.12783/dtetr/icpr2017/17676
[20] Kurniawati, D. A.; Nugroho, Y. I. (2017). Computational study of N-job M-machine flow shop
scheduling problems: SPT, EDD, NEH, NEH-EDD, and modified-NEH algorithms, Journal of
Advanced Manufacturing Systems, Vol. 16, No. 4, 375-384, doi:10.1142/S0219686717500226
[21] Rohaninejad, M.; Sahraeian, R.; Nouri, B. V. (2017). Minimising the total cost of tardiness and
overtime in a resumable capacitated job shop scheduling problem by using an efficient hybrid
algorithm, International Journal of Industrial and Systems Engineering, Vol. 26, No. 3, 318-343,
doi:10.1504/IJISE.2017.084421
[22] Yazdani, M.; Aleti, A.; Khalili, S. M.; Jolai, F. (2017). Optimizing the sum of maximum earliness
and tardiness of the job shop scheduling problem, Computers & Industrial Engineering, Vol. 107,
12-24, doi:10.1016/j.cie.2017.02.019
[23] Hammami, H.; Stützle, T. (2017). A computational study of neighborhood operators for job-shop
scheduling problems with regular objectives, Hu, B.; López-Ibáñez, M. (Eds.), Evolutionary
Computation in Combinatorial Optimization, Springer, Cham, 1-17, doi:10.1007/978-3-319-
55453-2_1
[24] Viana, M. S.; Junior, O. M.; Contreras, R. C. (2020). An improved local search genetic algorithm
with multi-crossover for job shop scheduling problem, Rutkowski, L.; Scherer, R.; Korytkowski,
M.; Pedrycz, W.; Tadeusiewicz, R.; Zurada, J. M. (Eds.), Artificial Intelligence and Soft
Computing, Springer, Cham, 464-479, doi:10.1007/978-3-030-61401-0_43
[25] Kundakci, N.; Kulak, O. (2016). Hybrid genetic algorithms for minimizing makespan in dynamic
job shop scheduling problem, Computers & Industrial Engineering, Vol. 96, 31-51,
doi:10.1016/j.cie.2016.03.011
[26] Zhu, H.; Deng, Q.; Zhang, L.; Hu, X.; Lin, W. (2020). Low carbon flexible job shop scheduling
problem considering worker learning using a memetic algorithm, Optimization and Engineering,
Vol. 21, No. 4, 1691-1716, doi:10.1007/s11081-020-09494-y
[27] Zhou, G.; Zhou, Y.; Zhao, R. (2021). Hybrid social spider optimization algorithm with differential
mutation operator for the job-shop scheduling problem, Journal of Industrial & Management
Optimization, Vol. 17, No. 2, 533-548, doi:10.3934/jimo.2019122

211