CC BY-NC-ND
0
DOI: 10.17323/2587-814X.2024.1.36.51
Agent-based modeling and optimization of the characteristics for research-and-production
clusters
Gayane L. Beklaryan
E-mail: glbeklaryan@gmail.com
Central Economics and Mathematics Institute, Russian Academy of Sciences Address: 47, Nakhimovsky Ave., Moscow 117418, Russia
Abstract
This paper presents a developed agent-based simulation model for the development of research-and-production clusters in Russia implemented with the use of high-tech enterprises located in four science cities (Troitsk, Obninsk, Pushchino and Protvino) as the case study. A new approach to modeling and optimization of gross metropolitan product (GMP) is proposed, taking into account the influence of the "gravity effect" on the redistribution of labor resources between developing science cities and appropriate enterprises united in single research and research-and-production clusters An important element of this approach is the formation of various scenarios for the strategic development of the research-and-production clusters being assessed and support for the possibility of choosing the most preferable scenario using an evolutionary optimization algorithm. An enlarged simulation model has been developed and implemented in AnyLogic describing the possible development trajectories of science cities with a corresponding change in the values of the most important characteristics: the number of economically active population, the number of research-and-production enterprises, the volume of products produced in high-tech sectors of the economy, GMP, etc. The designed framework is intended primarily for the
management of research-and-production clusters implementing the strategy of innovative development. Such a framework uses methods of system dynamics and agent-based simulation modeling supported in the AnyLogic system, genetic optimization algorithms and GIS mapping for science cities, etc. to implement the required functionality. The approbation of the framework was completed with the use of real data published in the approved strategies of the relevant science cities development. As a result of the numerical experiments carried out, some recommendations were proposed for the development of the research-and-production clusters under study considering their mutual influence and the existing base of resources.
Keywords: research-and-production cluster, high-tech enterprise, science city, production characteristics, simulation modeling for enterprises, gross metropolitan product, agent-based modeling, system dynamics, gravity effect, production function, AnyLogic
Citation: Beklaryan G.L. (2024) Agent-based modeling and optimization of the characteristics for research-and-production clusters. Business Informatics, vol. 18, no. 1, pp. 36-51. DOI: 10.17323/2587-814X.2024.1.36.51
Introduction
Currently, a very pressing problem of increasing the efficiency of the public administration system is the development of research and production clusters using the existing human resource potential, including those formed on the basis of research institutes and research-and-production clusters that have experience in using own developments in the real sector of the economy.
Determining the best development scenarios for high-tech enterprises located in science cities, developing and implementing decision support systems for managing research and production clusters is the most important area of business informatics aimed at increasing the efficiency of managing innovative development processes.
As an example, high-tech enterprises located in science cities united within a single conceptual model of innovative development will be considered. Within such a model, it is possible to obtain a significant syn-ergistic effect, due, firstly, to the possibilities of staff
cooperation, as well as the use of a common high-tech resource base (e.g., supercomputers, data centers, industrial systems for 3D printing of integrated circuits, etc.). In particular, high-tech enterprises located in cities such as Troitsk, Obninsk, Pushchino and Protvino near the Moscow agglomeration and being essentially neighboring enterprises can be united into single research-and-production clusters, in particular, for the production of a microelectronic element base, precision instrumentation products, etc. At the same time, it is important to ensure the balanced development of high-tech enterprises in all small cities united in the respective research-and-production clusters in order to avoid a one-way flow of highly qualified staff and ensure the harmonious strategic development of all agent enterprises (research institutes, research-and-production clusters, etc.) as existing ones, and those formed in the corresponding science cities.
Therefore, it is necessary to develop a decision support system (software framework) for managing hightech enterprises located in science cities and implementing an innovative development strategy. Such a
complex uses methods of system dynamics and agent-based simulation modeling supported in the AnyLogic system, genetic optimization algorithms and GIS maps of research-and-production clusters, etc. to implement the required functionality.
The first fundamental work on the study of the dynamics of the development of a city and its enterprises is the book [1], in which Forrester formulated the basic principles for constructing simulation models that describe various scenarios for the evolution of the urban environment, taking into account the complex system of existing feedbacks, lag dependencies and internal urban interactions, etc.
Among the recent works in this area, it is worth highlighting [2—4], in which system dynamics methods are used to study the possibilities of achieving sustainable economic development of large cities. In particular, in [2] four scenarios of possible development for the city of Beijing were modeled using a system-dynamic approach. It is shown that the dynamics of sustainable development of the city are implemented differently during three periods (from the growth of the city's economy at the beginning to its slowdown and subsequent growth).
Among the disadvantages of using system dynamics methods, one should note the difficulties of modeling internal interactions, processes of redistribution of labor resources, difficulties with the analysis of spatial development of the city, etc. To overcome such difficulties, it is advisable to use agent-based simulation modeling methods that allow modeling interactions at the microscopic level, taking into account individual adoption systems decisions of each economic agent [5, 6].
Among the important works in this area, [7-10] should be highlighted. In particular, [7] presents an aggregated agent-based model of migration flows of the European Union countries and describes the influence of the "gravity effect" on the inter-country redistribution of human flows. This effect arises due to factors influencing the migration behavior of people, in particular, when there is a significant difference in the ratio of wages, labor market size and gross domestic product (GDP) per capita between a given coun-
try (region) and neighboring countries (regions). For instance, relatively low wages combined with a small number of available jobs tend to lead to an outflow of the population to more favorable regions. The "gravity effect" also influences internal migration between cities, when the population (and corresponding labor resources) moves from the agglomerations that are less developed to economically more developed ones. In [8] an agent-based model of population dynamics of two interacting communities consisting of migrants and natives is presented. In such a model, natives search for high-tech workplaces, and migrants search for low-tech workplaces, which is also relevant for enterprises in science cities that actively attract external labor resources. Agent-based modeling can be combined with other methods, e.g., system dynamics, discrete event modeling, etc. to create digital twins of complex socio-economic systems [9, 10]. The developed simulation models can be aggregated according to objective functions and constraints with evolutionary optimization algorithms, in particular, genetic algorithms [11, 12] to optimize the characteristics of such large-scale systems. Agent-based models of territorial development of regional and urban agglomerations can aim at solving important environmental problems, such as reducing the concentration of harmful emissions in socially significant urban areas (e.g., near kindergartens and schools) through landscaping [13], reducing emissions from industrial enterprises due to their environmental modernization [14], transformation of the urban environment [15]. To model the movement of human flows and labor resources in an urban environment, it is advisable to use the phenomenological approach previously proposed in [16], which allows us to take into account various scenarios of interaction of various agents with each other, simulate the behavior of the crowd, etc. Also noteworthy are the works [1721] devoted to both methodological issues [17] and specific case studies carried out to study the possibilities of developing urban architecture [19], improving traffic [20] and increasing returns in certain sectors of the urban economy [21].
Thus, the combined use of various simulation modeling methods, heuristic optimization algorithms, etc., in particular, supported in the AnyLogic
system, makes it possible to design a decision support system for the sustainable economic development of research-and-production clusters and corresponding science cities.
The purpose of this paper is to develop economic, mathematical and computer tools for studying the dynamics of development of individual research-and-production clusters of the Russian Federation under various scenario conditions within the framework of which the balanced development of all interacting economic agents-enterprises (research institutes, research-and-production clusters, etc.) is ensured.
1. Simulation model for the development of research-and-production clusters
This section presents the developed simulation model for the development of research-and-produc-tion clusters with implementation using the example of high-tech enterprises located in the science cities of Troitsk, Obninsk, Pushchino and Protvino. The proposed agent-based model consists of several interconnected levels (Fig. 1).
Top level model - the environment in which agent enterprises operate, belonging to different cities, between which flow interaction is realized, in particular, redistribution (internal migration) of the population with a corresponding change in labor resources, which can be involved in the high-tech sector of the economy (in research institutes, SPAs and etc.) in the case of the formation of new jobs (Fig. 1). Such flow interaction is carried out due to the influence of the "gravitational" effect [7] caused by differences in the level of wages, the size of the labor market and GDP per capita in the corresponding science cities.
The Low level model ensures the implementation of the logic of behavior of agent enterprises, the characteristics of which are described using system dynamics methods (Fig. 1). Such internal models are systems of finite-difference equations with their own resource and target characteristics [11, 22]. At the same time, to determine the values of individual parameters, inte-
gration with the upper level of the simulation model is required. In particular, the main control parameters of the model, the values of which must be the same for all agents, are set at the environment level. At the same time, individual characteristics calculated at the agent level are used in the top-level model, for example, to calculate the gross metropolitan product (GMP) generated by all enterprises.
For each enterprise agent, three possible states are specified:
♦ the first state, corresponding to low economic growth rates and not exceeding a given lower threshold value, is highlighted in "red";
♦ the second state, corresponding to the average economic growth rates, the valuesof which are in the interval between the lower and upper threshold values, is highlighted in "yellow";
♦ the third state, corresponding to high rates of economic growth, the values of which exceed the upper threshold value, is highlighted as "green".
The economic growth rate of agent enterprises (the dynamics of GMP) is influenced by various factors, among which the most important are investments in fixed capital, the number of jobs created, average monthly wages, etc. The production resources that enterprises already have (fixed assets, staff) are also important: the level of scientific and technological progress (STP), the ability to attract highly qualified staff, etc. Within the framework of the developed model, a centralized allocation of resources (investments, new workplaces, etc.) is proposed with their subsequent distribution among enterprises of each science city, taking into account greater support for problem organizations characterized by low rates of economic growth. This approach is relevant, first of all, for state enterprises, e.g., institutes of the Russian Academy of Sciences, research institutes, as well as commercial organizations with significant government participation.
Thus, the most important task of the system under consideration is to minimize the total number of problem ("red") enterprises with the minimum required investment and operating costs.
Top level model
States of all agent enterprises
R Y G •
O
5
Population of Obninsk
O
Population of Pushchino
X>
5
Population of Troitsk
O
Population of Protvino
O
State of an agent enterprise
Output volume of an agent enterprise
City population States of all agent enterprises Investments in fixed assets of all agent enterprises Number of new workplaces for all agent enterprises Average monthly salary
t
Low level model
Fig.1. Aggregated architecture of a simulation model of research-and-production clusters.
1.1. Model of labor resource redistribution
The top-level model presented in Fig. 1 ensures the implementation of the logic of redistribution of population and labor resources, taking into account the influence of the "gravity effect." A similar effect was first described in works [23, 24] devoted to the analysis of human behavior under the influence of social forces. In particular, the decision to migrate to another city is influenced by multiple factors: differences in wages, the size of the labor market, GDP per capita, climatic characteristics, level of social security, etc. To a greater extent, the "gravitational" influence of such differentiation manifests itself at the level of countries that differ significantly in their level and quality of life. However, similar factors promoting migration also apply to cities, including science cities, the features of which allow their residents to move while remaining employed in high-tech sectors of the economy.
Here,
♦ T = {tp t2, ..., t]T|} is the set of moments of the model time (by year), |T is the strategic planning horizon (10 years);
♦ I = {i"p i2, ..., i|I|} is the set of indices of science cities, |I - total number of science cities;
♦ {v,, n}, i e I are the fertility and mortality rates in the i-th science city;
♦ {GMP.(tk), P,(tk)}, i e I are the gross metropolitan product and population of the i-th science city at moment tk, (tke T);
♦ {W( tk), L ( tk)}, i e I are the level of wage and amount of labor resources of the i-th science city employed in high-tech sectors of the economy at moment tk,
(t^ T);
♦ {Wj, w2, w3} e [0, 1] are the weighting coefficients that determine the level of influence of individual factors on migration flows, where Wj + w2 + #3 = 1;
♦ xt e [0, 1] is the coefficient that determines the level of influence of the "gravity effect" on emigration to the i-th science city.
An emigration rate from the i-th science city (ie I) to the neighboring Z-th science city (£e I, Z^ i) at moment tk, (tke T) is equal to
w=
where
x.
1
0, if giS(tk)> 1,
-1
P^JM gJtk)<\,
(1)
+#2ln
+ w3ln
GMP,(tk_,) GMPAt^)
V s
P^)
(2)
The equations (1)-(2) describe the influence of the "gravity effect" on migration flows. Expression (2) reflects the influence of such key factors as wages W(tk_i), number of employed labor resources Li(tk l) and GDP per capita
1)
on the migration behavior of population (i.e., migration rates) in the i-th science city (i e I). The choice of such factors is determined, first of all, by the observed difference in their values in relation to the science cities under consideration (i.e., Troitsk, Obninsk, etc.). The influence of other potentially important characteristics, e.g., differences in climate, transport accessibility, ecology, etc., is not significant from the geographic proximity and spatial homogeneity of the studied agglomerations. The use of relative and scaled (loga-rithmized) values of influencing factors allows them to be aggregated using (2) for subsequent calculation of migration rates to neighboring science cities.
From (1)-(2) it follows that if a given science city is less attractive than a neighboring one which has higher wages, more jobs and more GDP per capita, then a positive emigration flow to this science city is provided. The opposite scenario is also possible, when a given
science city is more attractive to residents of neighboring cities. In this case, the formation of positive immigration flows will be ensured.
The population of the i-th science city (i e I) considering the emigration to the neighboring Z-th science cities (Z e I, Z ^ i), the immigration from the neighboring Z-ths science cities (Z e I, Z ^ i), as well as the fertility and mortality at moment tk, (tke T) is equal to
Together with the redistribution of the population between neighboring science cities, the redistribution of highly qualified labor is being implemented accordingly.
The amount of labor resources that can be employed in high-tech enterprises of the z'-th science city (i e I) at moment tk, (tke T) is equal to
(4)
where
Ht e [0, 1] is the share of economically active population in the z'-th science city (i e I);
a.e [0, 1] is the share of highly qualified labor resources from the economically active population in the i-th science city (i e I).
1.2. Model of behavior of enterprise agents
The lower-level model presented in Fig. 1 is used to calculate the dynamics of the average annual growth rate of output of agent enterprises with a corresponding assessment of the states of such agents. The most important characteristic of such enterprises is the production function, which describes the influence of fixed assets, labor resources, the level of scientific and technological development and other factors on the dynamics of output volume. A detailed description of the principles for constructing production functions, including the Cobb-Douglas type used in this work
is presented in [25, 26]. There are also examples of constructing multiplicative production functions that take into account the influence of wages on output, as a factor that significantly affects the productivity of labor resources (e.g., [27]). A feature of the production function proposed in this article is that it takes into account the relative (in relation to all neighboring science cities) level of wages. The higher this level, the more qualified labor resources can be employed at enterprises in a given city.
Let be,
♦ Jt = {jiVjn, —J^ }, ie I is the set of indices of agent enterprises located in the z'-th science city, where J | is the total number of enterprises;
♦ j (tk) e {1, 2, 3} are the states of the yz-th agent enterprise of the z'-th science city at momen tk, (tk<= T): Sj (tk) = 1 is the first state corresponding to low economic growth rates ("red"), Sj (tk) = 2 is the second state, corresponding to the average rate of economic growth ("yellow"), s. (t.) = 3 is the
Jt k
third state, corresponding to high rates of economic growth ("green");
♦ {Kj (tk), L. (tk)}, i e I, j. e Jt are the fixed assets and labor resources of the j-th agent enterprise of the i-th science city at moment tk, (tke T);
♦ {8K (tk), 8Lj (ft)}, i e I, jt e J are the rate of new fixed assets commissioning and the labor resources inflow of the j-th agent enterprise of the i-th science city at moment tk, (tke T);
♦ {Qi(tk), Yi(tk)}, i e I are investments in fixed assets and the number of new workplaces allocated by the i-th science city at moment tk, (tke T);
♦ {a;. , p.} (0, 1), i e I, ji e J. are the coefficients of elasticity of output volume in relation to fixed assets and labor resources of the j-th agent enterprise of the i-th science city, where aj + /? = 1;
♦ A., i e I,jt e J is the factor (coefficient) reflecting the influence of scientific and technological progress (STP) on the dynamics of output volume;
♦ {«, i} are the known retirement rates of fixed assets and labor resources, respectively.
Then, the output volume of the j-th agent enterprise (j e J) of the i-th science city ( i e I) can be specified with the use of the Cobb-Douglas type production function at moment^, (tke T):
where
S =
h
w.
* i=i
Ji=i
if
1 + y
j,= 1
j(=i
if ^(^> = 3,
l-V
10,0*-.)
Jl=l
8L.(tk) =
if =
1 + y
ii=l
TfjSh-.)
Ji=i
l-yi
(5)
(6)
(8)
(9)
(10)
rjSo=
0, if s.{tk)*\.
(11)
Figure 2 shows the state-flow chart of the yi-th agent enterprise (j e J) of the i-th science city ( i e I) implementing the transition rules to new states (tk) e {1, 2, 3}.
Transitions to new states (indicated by the sign (?) in Fig. 2) are carried out using the following rule:
1 X V (K) 2, if ^
■x -f 1 V № . 3, if —> —><p,.
Here, {<pp <p2| are the specified threshold values that determine the conditions for compliance of agent enterprises with states of low, medium and high rates of economic growth.
I
Fig.2. State-flow chart of an agent enterprise.
The proposed model takes into account the influence of the relative wage level (6) on output (5). The use of relations (9)—(11) allows us to provide priority support to agent enterprises that are in the first state of low economic growth rates. Here, y e [0, 1] is the coefficient determining the overall level of such support. Thus, investments in fixed assets and new jobs allocated for each science city are distributed among "problem" agent enterprises in proportion to their contribution to the total output.
1.3. Optimization
problem statement
A system is considered in which it is necessary to ensure the minimization of the total number of enterprises characterized by low rates of economic growth for all selected science cities. For the sustainable development of research-and-production clusters, it is desirable to ensure the transition of the majority of problem enterprises to states of medium and high rates of economic growth and the preservation of such favorable states for the longest possible time interval. However, in the current conditions of a shortage of investment capital, as well as limited opportunities to create new high-tech jobs, it is important to find such states and modes of operation of the system in which minimizing the number of problem organizations is achieved with minimal costs at least by some given point in model time. As a result, a qualitative improvement in the characteristics of the corresponding enterprises will be ensured, practically guaranteeing the preservation of the states they have achieved over a sufficiently long time interval due to the inertia of the values of production factors. In the absence of statistical information about the existing budget and investment restrictions, it is justified to reduce such an optimization problem to a single-criteria one with the following objective functional, estimated at the final moment of model time t,T,, (t,T, e T):
<=1 4=1
(13)
where
[l,if s.(tk) = \, |0,if Sj (tk)*\.
(14)
Here,
♦ {£>,.(4), Yj(tj), W(tj)}, i e I are the investments in fixed assets, number of new workplaces and wages in the i-th science city at moment tk, (tk e T) which are control parameters of the model;
♦ {«j, n2, n3} are the normalizing factors that are used to reduce the values of the elements of the objective function values to the uniform scale.
The choice of values of the control parameters of the model is carried out at each moment tk, (tk e T) and leads to changes in the production characteristics of agent enterprises, i.e., fixed assets, labor resources and output volumes in accordance with (5)—(11). As a result, the impact on the average annual rate of economic growth and the state of the corresponding enterprises is ensured, see (12)-(14).
Then the optimization problem can be formulated as follows.
Problem A. The need to minimize the value of the total number of enterprises characterized by low rates of economic growth at the final moment of the model time ZjT , (zjT | e T), as well as the investment and operating costs associated with the implementation of the strategy:
min .NUJ (15)
{«W.W.'WC,
s.t.
Y<Y(tk)<Y, W<Wt(tk)<W, iel, k = 1, 2, ..., |r|.
Here, {Q, Y, W], {Q, Y, W] are known lower and upper limits of the simulation model control parameter values.
Problem A can be solved using a genetic optimization algorithm aggregated by objective function with the proposed simulation model. The implementation of such algorithms is based on methods of evolutionary search for the best potential solutions and is described in detail
in [12, 28]. To simplify the problem under consideration, the values of control parameters in such a system can be specified for each i-th science city (i e I) as constants at the initial moment of the model time.
2. Software implementation of the model
The software implementation of the developed simulation model (1)—(15) was carried out in the Any-Logic system. A feature of the proposed approach is the use of GIS maps to visualize the states of enterprise agents located in given spatial coordinates, as well as the integration of the model with a previously developed genetic optimization algorithm [28], which, in particular, provides a solution to Problem A.
The developed software package provides the ability to navigate through all the science cities being studied,
vary the values of the control parameters of the model and visualize the states of agent enterprises on the map.
Figure 3 shows the implementation of the model for the redistribution of population and labor resources between the science cities under consideration (toplevel models) in AnyLogic.
As seen from Fig. 3 the system dynamics methods are used to construct this model (e.g., [11, 22]). The population of Obninsk, Troitsk, Protvino and Pushchino is modeled using four system levels ("reservoirs") interacting through flows at rates whose values depend on the difference in wage levels, the size of the labor market, etc. Thus, the influence of the "gravity effect" on the rate of internal migration. Fertility and mortality rates with their known values are also taken into account. The reinforcing and balancing feedbacks present in the model are marked with the letters R and B, respectively (Fig. 3).
Flow1_P
Flow41
Population_of_Pusthchinp
Flov;4 P
G^—>
I 1 "fl Flow4_C /
0W3 0W4
^ Mortality
Fig. 3. Model of redistribution of labor resources in AnyLogic.
STP alpha beta
Investments_in_fixed_assets & (^—--^Ci-x Ratio_in_salary-to_other_cities
vOutput_volume
Of3 w C
Rate_of_inflow_fixed_assets X Rate of outfiow\ tc\ ^ event
Rate_of_inflow_labor_resources
Share_of_highly_qualified\staff Number_economically_active\population
Share_of_economically_active_population
Popu|ation_size
d Output_volume_1 T
state
O
0 Average_annual_growth_rare_of_output_volume statechart
1
( LoW rates of economic growth
TT
Average rates of economic growth
TT
High rates of economic growth
Fig.4. Model of behavior of enterprise agents in AnyLogic.
Figure 4 shows the implementation of the behavior model of enterprise agents (low-level model) in AnyLogic.
Figure 4 shows that to build this model, both the system dynamics and agent-based modeling methods are used (e.g., [5, 6, 13, 14, 16]). In particular, to model possible changes in the states of agents in accordance with the given rules, a state-flow chart is used (that is shown in Fig. 4). At the same time, to determine the values of the average annual growth rate of output volume, the value of which is compared with the given threshold values for the implementation of transitions between the states of each agent, the calculation uses the system dynamics model (the left part of Fig. 4). The model (5)—(11) uses the function of the Cobb-Doug-las type to calculate the output volume, taking into account the values of control parameters. The results of simulation modeling, in particular, the state of each
agent enterprise, the output volume, the amount of labor resources, etc. are automatically uploaded to the top-level model and used to simulate the "gravity effect" that determines the process of redistribution of population and labor resources (Fig. 3).
3. Results of optimization experiments
To improve the state of the entire ensemble of agent enterprises, in particular, to minimize the number of organizations with low rates of economic growth (less than 2% per year), the optimization experiments were carried out aimed at finding recommended values of control parameters of the model. As initial data, aggregated statistics on science cities were used, that is shared in [29], as well as the strategies for the socioeconomic development of science cities published on
the websites of city administrations. An example of initial data on agent enterprises for the Obninsk city is presented in Table 1.
Figure 5 presents a visualization of the states of agent enterprises obtained as a result of optimizing the values of control parameters of the simulation model ("Scenario based on recommended parameters") in comparison with previously approved investment and staff strategies ("Initial scenario (before optimization)".
The states of agent enterprises presented in Fig. 5 correspond to a ten-year strategic planning horizon. A scenario for changes in such states based on recommended parameters (Fig. 5) was generated using a genetic optimization algorithm, aggregated by target functionality with a developed simulation model implemented in AnyLogic. Table 2 presents the values of the control parameters of the model that correspond to the scenarios under consideration.
Table 1.
An example of initial dataset for agent enterprises of the Obninsk city (for 2016 year)
No Company Longitude Latitude Labor resources, people Investments in fixed assets, thousand rubles Fixed assets, thousand rubles Issue volume, thousand rubles
1 JSC "SSC RF-IPPE" 55.089851 36.591257 2612 14782 594124 3199431
2 JSC ONPP "Technology" 55.105885 36.636379 2641 488089 6527163 5856985
3 JSC "SSC RF NIFHI" 55.063435 36.623261 528 103226 1300044 803209
4 MRNC named after A.F. Tsyba -the branch of the FSBI "NMRRC" of the Ministry of Health of the Russian Federation 55.111844 36.617534 1559 136132 2230550 2052357
5 ASP "Typhoon" 55.104454 36.609027 580 1551 80820 293792
6 VNIIRAE 55.106651 36.638687 270 3058 110706 242231
7 Public JST "Instrument Plant "Signal" 55.11013 36.59102 308 48217 996640 1061128
8 LLC "Hemofarm" 55.13496 36.6424 209 169412 2247154 1374894
9 JSC "Progress-ecology" 55.12122 36.58286 236 14758 696989 1709746
10 LLC "Nearmedic Pharma" 55.11442 36.61703 125 2810003 29994145 3506284
Obnnsk city Initial scenario (before optimization) Scenario based on recommended parameters From Table 2 it follows that in order to improve the conditions of agent enterprises in the science cities under study, it is necessary to significantly increase
LLC "Hemofarm" « (Y investments in fixed assets (mainly in Troitsk, Pro-
JSC "Progress-ecology" <G (G tvino and Pushchino), as well as to multiply the
LLC "Nearmedic Pharma" ( H (Y number of workplaces that should be created. At
MRNC named after A.F. Tsyba -the branch of the FSBI "NMRRC" Y Y the same time, it is recommended to increase the level of wages in Obninsk and Pushchino in order
Public JST "Instrument Plant "Signal" Y (G to prevent the outflow of staff to neighboring cities
JSC ONPP "Technology" (Y (Y (in particular, Troitsk). As a result, a fundamental
ASP "Typhoon" G G improvement in the conditions of agent enterprises
JSC "SSC RF-IPPE" G G will be ensured (see Fig. 5), most of which will move
VNIIRAE G G to the third state of high economic growth rates.
JSC "SSC RF NIFHI" 4JI (y
B y First stage Second stage Third stage (low growth rate) (average growth rate) (high growth rate) Fig.5. States of agent enterprise before and after optimization. Conclusion This paper presents a new agent-based simulation model for the development of individual research-and-production clusters in Russia with implementation in the AnyLogic system. The model
Table 2.
Values of model control parameters
Control parameters Science city Basic scenario (before optimization) Scenario based on recommended settings
Investments in fixed assets (million rubles/year) Obninsk 3320 4341
Troitsk 453 3108
Protvino 246 2252
Pushchino 3 -•J 2100
Number of new jobs (units/year) Obninsk 1000 3020
Troitsk 500 2550
Protvino 300 1551
Pushchino 200 1121
Average monthly salary (rubles/month) Obninsk 62 000 72 000
Troitsk 66 000 66 000
Protvino 60 000 60 000
Pushchino 45 000 55 000
developed consists of two levels (Fig. 1). The top-level model implements the environment in which agent enterprises belonging to various science cities operate, between which flow interaction is realized with a corresponding redistribution of the population and labor resources. The lower-level model provides the implementation of the logic of behavior of enterprise agents, the characteristics of which are described using system dynamics methods. An important optimization problem has been formulated to minimize the total number of enterprises characterized by low rates of economic growth, taking into account investment and operating expenses. The proposed two-level simulation model is implemented in the form of a software package (using the example of research-and-production clusters of Obninsk, Troitsk, Protvino and Pushchino), intended primarily for municipal and regional authorities imple-
menting the strategy of innovative development. The use of such a system made it possible to formulate recommended values of control parameters (such as investments in fixed assets, the number of new jobs, etc.), providing a significant improvement in the conditions of agent enterprises located in the science cities under study over a ten-year strategic planning horizon (Fig. 5 and Table 2). The economic-mathematical and computer toolkit developed is intended primarily for managing research-and-production clusters implementing the strategy of innovative development.
Further research will be aimed at studying the problems associated with overcoming the shortage of highly qualified labor, as well as the creation of detailed agent-based models for the development of research-and-production clusters. ■
References
1. Forrester J.W. (1969) Urban dynamics. M.I.T. Press, Cambridge.
2. Li G., Kou C., Wang Y., Yang H. (2020) System dynamics modelling for improving urban resilience in Beijing, China. Resources, Conservation and Recycling, vol. 161, 104954. https://doi.org/10.1016/j.resconrec.2020.104954
3. Diemer A., Nedelciu C.E. (2020) System dynamics for sustainable urban planning. Sustainable Cities and Communities. Encyclopedia of the UN Sustainable Development Goals (eds. W. Leal Filho,
A. Marisa Azul, L. Brandli, P. Gokgin Ozuyar, T. Wall). Springer, Cham, pp. 760-773. https://doi.org/10.1007/978-3-319-95717-3_115
4. Armenia S., Barnabe F., Pompei A., Scolozzi R. (2021) System dynamics modelling for urban sustainability. Urban Sustainability. Springer Texts in Business and Economics (eds. J. Papathanasiou, G. Tsaples, A. Blouchoutzi). Springer, Cham, pp. 131-173. https://doi.org/10.1007/978-3-030-67016-0_4
5. Makarov V., Bakhtizin A., Epstein J. (2022) Agent-based modeling for a complex world. Part 1. Economics and the Mathematical Methods, vol. 58, no. 1, pp. 5-26 (in Russian). https://doi.org/10.31857/S042473880018970-6
6. Makarov V., Bakhtizin A., Epstein J., (2022) Agent-based modeling for a complex world. Part 2. Economics and the Mathematical Methods, vol. 58, no. 2, pp. 7-21 (in Russian). https://doi.org/10.31857/S042473880020009-8
7. Makarov V., Bakhtizin A., Beklaryan G., Akopov A., Rovenskaya E., Strelkovskiy N. (2019) Aggregated agent-based simulation model of migration flows of the European Union countries. Economics and the Mathematical Methods, vol. 55, no. 1, pp. 3-15 (in Russian). https://doi.org/10.31857/S042473880004044-7
8. Makarov V., Bakhtizin A., Beklaryan G., Akopov A. (2020) Agent-based modelling of population dynamics of two interacting social communities: migrants and natives. Economics and the Mathematical Methods,
vol. 56, no. 2, pp. 5-19 (in Russian). https://doi.org/10.31857/S042473880009217-7
9. Makarov V.L., Bakhtizin A.R., Beklaryan G.L., Akopov A.S. (2021) Digital plant: methods of discrete-event modeling and optimization of production characteristics. Business Informatics, vol. 15, no. 2, pp. 7-20. https://doi.org/10.17323/2587-814X.2021.2J.20
10. Makarov V.L., Bakhtizin A.R., Beklaryan G.L., Akopov A.S. (2019) Development of software framework for large-scale agent-based modeling of complex social systems. Programmnaya Ingeneria (Software Engineering), vol. 10, no. 4, pp. 167-177 (in Russian). https://doi.org/10.17587/prin.10.167-177
11. Akopov A.S. (2012) Designing of integrated system-dynamics models for an oil company. International Journal of Computer Applications in Technology, vol. 45, no. 4, pp. 220-230. https://doi.org/10.1504/IJCAT.2012.051122
12. Akopov A.S. (2014) Parallel genetic algorithm with fading selection. International Journal of Computer Applications in Technology, vol. 49, no. 3-4, pp. 325-331. https://doi.org/10.1504/IJCAT.2014.062368
13. Akopov A.S., Beklaryan L.A., Saghatelyan A.K. (2019) Agent-based modelling of interactions between air pollutants and greenery using a case study of Yerevan, Armenia. Environmental Modelling & Software, vol. 116, pp. 7-25. https://doi.org/10.1016/j.envsoft.2019.02.003
14. Akopov A.S., Beklaryan L.A., Saghatelyan A.K. (2017) Agent-based modelling for ecological economics: A case study of the republic of Armenia. Ecological Modelling, vol. 346, pp. 99-118. https://doi.org/10.1016/j.ecolmodel.2016.11.012
15. Kislitsyn E.V., Gogulin V.V. (2021) Simulation of the environmental situation in a megalopolis. Modeli, sistemy, seti v ekonomike, tekhnike, prirode i obshchestve (Models, systems, networks in economics, technology, nature and society), vol. 1, no. 37, pp. 92-106 (in Russian). https://doi.org/10.21685/2227-8486-2021-1-8
16. Akopov A.S., Beklaryan L.A. (2015) An agent model of crowd behavior in emergencies. Automation and Remote Control, vol. 76, no. 10, pp. 1817-1827. https://doi.org/10.1134/S0005117915100094
17. Crooks A., Heppenstall A., Malleson N., Manley E. (2021) Agent-based modeling and the city: A gallery of applications. Urban Informatics. The Urban Book Series (eds. W. Shi, M.F. Goodchild, M. Batty,
M.P. Kwan, A. Zhang). Springer, Singapore, pp. 885-910. https://doi.org/10.1007/978-981-15-8983-6_46
18. Chen L. (2012) Agent-based modeling in urban and architectural research: A brief literature review. Frontiers of Architectural Research, vol. 1, no. 2, pp. 166-177. https://doi.org/10.1016/j.foar.2012.03.003
19. Tian G., Qiao Z. (2014) Modeling urban expansion policy scenarios using an agent-based approach for Guangzhou Metropolitan Region of China. Ecology and Society, vol. 19, no. 3, art. 52. https://doi.org/10.5751/ES-06909-190352
20. Yun T.-S., Kim D., Moon I.-C., Bae J.W. (2022) Agent-based model for urban administration: A case study of bridge construction and its traffic dispersion effect. Journal of Artificial Societies and Social Simulation, vol. 25, no. 4, art. 5. https://doi.org/10.18564/jasss.4923
21. Ghandar A., Theodoropoulos G., Zhong M., Zhen B., Chen S., Gong Y., Ahmed A. (2019) An agent-based modelling framework for urban agriculture. 2019 Winter Simulation Conference (WSC), National Harbor, MD, USA, pp. 1767-1778. https://doi.org/10.1109/WSC40007.2019.9004854
22. Akopov A.S. (2012) System dynamics modeling of banking group strategy. Business Informatics, vol. 2, no. 20, pp. 10-19 (in Russian).
23. Stewart Q.J. (1950) The development of social physics. American Journal of Physics, vol. 18, pp. 239-253. https://doi.org/10.1119/L1932559
24. Yap Y.L. (1977) The attraction of cities: A review of the migration literature. Journal of Development Economics, vol. 4, no. 3, pp. 239-264. https://doi.org/10.1016/0304-3878(77)90030-x
25. Kleiner G.B. (1986) Production functions. Theory, methods, application. Moscow: Finance and Statistics (in Russian).
26. Suvorov N.V., Akhunov R.R., Gubarev R.V., Dzyuba E.I., Fayzullin F.S. (2020) Applying the Cobb-Douglas production function for analysing the region's industry. Economy of Regions, vol. 16, no. 1. pp. 187-200
(in Russian). https://doi.org/10.17059/2020-1-14
27. Hellerstein J.K., Neumark D. (2007) Production function and wage equation estimation with heterogeneous labor: Evidence from a new matched employer-employee data set. Hard-to-Measure Goods and Services: Essays in Honor of Zvi Griliches (eds. Ernst R. Berndt and Charles R. Hulten). Chicago: University of Chicago Press, pp. 31-72. Available at: https://www.nber.org/system/files/chapters/c0873/c0873.pdf
(accessed 21 February 2024).
28. Beklaryan G.L., Akopov A.S., Khachatryan N.K. (2019) Optimisation of system dynamics models using a real-coded genetic algorithm with fuzzy control. Cybernetics and Information Technologies, vol. 19, no. 2, pp. 87-103. https://doi.org/10.2478/cait-2019-0017
29. Guide on the results of the analysis of the compliance of the indicators of scientific and industrial complexes of science cities of the Russian Federation with the requirements established by paragraph 8 of Article 2.1
of Federal Law No. 70-FZ "On the status of the science city of the Russian Federation", and the achievement of the results provided for by the action plans for the implementation of the strategies of socio-economic development of science cities of the Russian Federation in 2022 (in Russian). Available at: https://minobrnauki.gov.ru/upload/iblock/77c/cemgzf9g61hhktvme7dfmm9feddbfzvv.pdf (accessed 21 February 2024).
About the author
Gayane L. Beklaryan
Cand. Sci. (Econ.);
Senior Researcher, Department of Theoretical Economics and Mathematics Research, Laboratory of Computer Modeling of Social and Economic Processes, Central Economics and Mathematics Institute, Russian Academy of Sciences, 47, Nakhimovsky Ave., Moscow 117418, Russia;
E-mail: glbeklaryan@gmail.com
ORCID: 0000-0002-1286-0345
