211 Imitation Modeling of UAV’s Multi-Purpose System of Optimal Structure Based on Generalized Method of Dynamic Condensation* Quang Thuong Nguyen 1[0000-0001-7694-4401] Аnh Hien Vu 2[ 0000-0002-8308-7222] and Thanh Long Nguyen 3[0000-0002-1950-2866] 1 State university of management, 99 Ryazan Ave., Moscow, 109542, Russia tikrus20.21@gmail.com 2.3 Institut of Military Science and Technology, 17, Hoang Sam Str., Hanoi, Vietnam nguyenthanhlong_676@yahoo.com, vahien@mail.ru Abstract. The article presents the task of substantiating the selection of the de- sign characteristics of a multi-purpose UAV system to simulate the optimal struc- ture of a multi-purpose UAV system in the study of tropical cyclones, the task of justifying the choice of characteristics for building a multi-purpose UAV system, the algorithm of the generalized dynamic concentration method (GDCM) to solve a class of optimization problems using the criterion of the minimum cost of the UAV system's target tasks at a given efficiency of their implementation. The so- lution to the problem of substantiating the choice of characteristics for building a multi-purpose UAV system, the simulation results of the optimal multi-purpose UAV system for monitoring tropical cyclones in Vietnam are proposed. The re- sults of the solution of the formulated scientific problem should be used in the future when forming the tactical and technical requirements (TTZ, TZ) of the Customer for the creation (modernization) of promising multi-purpose UAV sys- tems when solving remote sensing problems in terms of evaluating the effective- ness of their functioning, and evaluating the resource provision, in particular, costs. Keywords: Unmanned Aerial Vehicle (UAV), a Multi-Purpose UAV System, Imitation Modeling, Generalized Method of Dynamic Condensation (GMDC), Algorithm, Statistical Sample, Optimization, Criterion, Optimal multi-purpose, Remote sensing. 1 Introduction Currently, many countries of the world are actively developing and using systems of unmanned aerial vehicles (UAV) that would deliver special platforms for data collec- * Copyright 2021 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0). 212 tion (PDC) to a certain area of the world ocean, to solve the problems of analysis, as- sessment, monitoring, and forecasting the formation of tropical cyclones in specific regions [1, 2]. At the same time, such special platforms should be so distributed over the area of the ocean and its height that measurements of the characteristics of the at- mosphere at various points of the probed object are synchronized. This synchronization is necessary to build a General and maximally accurate model of the development of atmospheric vortices with the possibility of its practical application on a time scale close to the real one [3]. The multi-purpose feature of the UAV system is determined by the variety of tasks performed by the system, the conditions of its operation, and the set of elements that make up this system. In General, the construction of this system consists of the rational distribution of a given set of target tasks {X} between individual elements of the UAV system. Replacing the external target set of tasks {X} with some "design characteristic" in this case will be incorrect since the previously obtained "design characteristics" can lead to a significant systematic error in evaluating the effectiveness of a multi-purpose UAV system. In General, regardless of what type of UAV should be used (operated) in the optimal structure of a multi-purpose system, the justification for the choice of characteristics to build the structure (framework) of such a system inevitably leads to the need to solve a class of optimization problems using the criterion of the minimum cost of the UAV system's target tasks at a given efficiency of their implementation. The purpose of this work is to simulate the optimal structure of a multi-purpose UAV system using the generalized method of dynamic condensation in the study of tropical cyclones one of the most destructive and regularly recurring natural phenomena in the territories of the Earth and the world Ocean adjacent to the Socialist Republic of Vi- etnam. 2 The Task of Justifying the Choice of Characteristics for Building a Multi-Purpose UAV System Find   aD E  x     C W , D, E    min min C W , D, E   ,  (1) where C[W,D,E( )] - the cost of a multi-purpose UAV system; W - the external target set of tasks defined by vectors     m pl ,  ,  , H  W ; mpl – the payload mass, latitude - , longitude -  and altitude - H of a given cyclone point; E   - the target distribution function, which is defined by elementary distribution functions eij:eij=1, if the i-th task is performed by the j-th type of UAV and eij=0, oth- erwise. 213     E   e 1,1 , e 1, 2  ,..., e  i, j  ,..., e  n, m  ; i  1; n, j  1; m, Here n is the given number of target tasks, m is the given number of UAV types. D is a set of strategies for building a variant of the UAV system (Dirichlet region). e(i,j)=1 if the i-th task is completed by the j-th UAV type; e(i,j)=0, otherwise. Thus, the task of justifying the choice of characteristics of the structure of a multi- purpose UAV system is to search for such options (scenarios) for the survey of one of N goals using various types of UAVs to obtain operational remote sensing information, in which the probability of completing the target task Ptcwill be maximum. In this task, not only the optimal distribution of target tasks by UAV types is selected, but also the corresponding design solutions for each of the separate UAV types. Prob- lem (1) is solved under the following parametric constraints on design solutions [4, 5]:  P0 min  P0  P0 max ;   PS min   PS   PS max ;  mW min  mW  mW max ;  M 0 min  M 0  M 0 max . (2) where the vector of design parameters a defines the key characteristics of the a  P ,  0 PS W 0 , ,m ,M  UAV's appearance and has the following components: where P0 is the thrust of the propulsion system (PS); PS is the relative mass of the propulsion system; mW is the weight of the wing; M0 is the UAV’s initial mass. Functional restrictions have the form: m g1  Di  D j  ; g 2  Dj W j 1 , (3) here j   D    W E ( )  j , j  1, m , - the Dirichlet region. It is assumed that the solution to the problem of the functioning of a multi-purpose UAV system is related to the area of acceptable design solutions for completing UAVs with special (PDC) and service equipment, which is determined by parametric and functional restrictions of the form [6]:  D  a  A, u U , amin i  ai  amax i , i  1, n, u min j  u j  u max j , j  1, m, g r  d   0, r  1, g  , (4) where A, U are the permissible regions in the parameters and control; amin i  ai  amax i , i  1, n, umin j  u j  umax j , j  1, m are the restrictions on the design parameters of the UAV and its motion control func- g r  d   0, r  1, g d   a, u  t   tions, respectively; are the functional restrictions; is the UAV design solution vector;  is the vector of characteristics of the target task of a multi-purpose UAV system. 214 The first functional restriction in the task (1) means that the set of target tasks per- formed by the i-th type of UAV does not intersect with the set of target tasks performed by the j-th type of UAV. The physical meaning of the second functional constraint in problem (1) means that the structure of the UAV system must be constructed specially. More specifically, it should be possible to redistribute a given set of target tasks to each other on a time scale that is close to real-time. At the same time, the cost of the entire UAV system for ob- taining the necessary remote sensing information with the required frequency and up- dating it should be minimal. The generated problem of the form (1) belongs to the class of problems of multi- criteria multiparametric multi-factor identification of indicators and characteristics of complex organizational and technical systems and their structural and parametric opti- mization. This problem is solved by the generalized method of dynamic condensation [7, 8, 9]. 3 The Algorithm of the Generalized Method of Dynamic Condensation (GMDC) The generalized method of dynamic condensation consists of using a combination of the possibilities of the optimization method in the space of inverse functions and the possibilities of the dynamic condensation method. Based on the method of dynamic condensation, a certain verification function is formed that allows you to build the final distribution of target tasks between individual types of UAVs based on the distances between the design parameters of the UAV. To apply the inverse function method, you must set parametric restrictions on the variable parameters of the UAV type (2) and the corresponding functional restrictions. ' The measure of adequacy between a subset A  A and an object y is defined as follows [10]:  '  aA' D A , y    a d 1  a, y    '   a   2  a, y  , a A (5) where (a) is " the weight" coefficient in the context of task (1), i.e. the required number of modules required to justify the configuration of all types of UAV multi- purpose system. The function of representation (selection) of UAV parameters for a set of target tasks g(P)=L matches each cluster Pk with its representative lk by the condition:  lk  g  Pk   a j  Pk max d a j , y  ,  j   (6) p where y  R - cluster center of gravity Pk. 215  a j Pk      a j  0, a j y (7)   a j  , a Pj k Thus, the object y that is closest to the center of gravity of the cluster with Pk is selected as the cluster representative. The assignment function (distributing targets by UAV type) f(L)=P takes integer values 1,2,…,k and distributes objects to clusters:  Pi  a  A d  a, li   d a, l j  , (8) In the case, if d(a,li)=d(a,lj), then aP i when i, that characterizes the set of UAV design parameters and the target tasks to be solved is formed as a set of partitions by UAV types and its corre- sponding set of representatives so that the condition is met:  * *   mink i1 aP   a  d 1  a, Li , K W P ,L (10) i The algorithm for finding the optimal pair formation and partition of sets {Pk*, L*} consists of the following iterative steps [8, 11]: Step1. Search for the optimal partition of the set Pk at k=const. ' Step2. Selecting a candidate object a  A for the (k+1)-th cluster representatives Pk. Step3. Checking for convergence to the desired solution. The rule for selecting object y as a representative of the new (k+1) -th cluster Pk+1 has the form:   lk 1   a j  A min min d ( a j , lk )  , (11)  k a j Pk  216 that is, the object y with the minimum similarity measure to its cluster Pk+1 is se- lected as a candidate for representatives, or, in other words, the most remote object in metrics ρ(a,y). Indicator of optimal splitting of a set of modules A into clusters Pk , A  Pk  ; Pi  Pj   has the form: K K Arg min W  P, L    D  Pi , li       a  d  a, li  , 1 (12) i 1 i 1 aPi where li- the representative of the i-th cluster, selected based on the results of an assessment of proximity to the center of gravity of this cluster. To estimate the end of the iterative process, use the condition: max  ( a j , lk )   , Pk  A (13) a j Pk where ε - the given accuracy of the target function. The algorithm of the generalized method of dynamic condensation (GMDC) is shown in figure 1 (see fig. 1). [5, 6]. Fig. 1. Algorithm of the generalized method of dynamic condensation (GMDC). 217 Figure 2 shows the algorithm of the generalized method of dynamic condensation (GMDC). In its final form the GMDC algorithm consists of the following sequence of iterative steps [9, 11]: Step 1. Put k=1, Pk =A (li can be selected at random). Step 2. Execute the GMDC algorithm for the current value k=const. Step 3. For the given accuracy  check whether the condition (9) is met. a. If this condition is met, the algorithm is complete, and the division of the set of design parameters by UAV Pk type and representation Lk by UAV target tasks are con- sidered final. b. If this condition is not met, then you need to go to step 4. Step 4. Define the object y  A that meets the condition (11). Step 5. Put k=k+1, Pk =y, lk=y you can go to step 2 of the algorithm by excluding the object from the cluster that it belonged to in the previous step. Thus, the generalized method of dynamic condensation consists of using a combina- tion of the possibilities of the optimization method in the space of inverse functions and the possibilities of the dynamic condensation method. Based on the method of dynamic condensation, a certain verification function is formed that allows you to build the final distribution of target tasks between individual types of UAVs based on the distances between the design parameters of the UAV. 4 The solution to the Problem of Substantiating the Choice of Characteristics for Building a Multi-Purpose UAV System The solution of a set of tasks for modeling the optimal structure of a multi-purpose UAV system for sensing the Earth's atmosphere is carried out in the following se- quence. Step 1. For each payload using a UAV, a statistical sample of the form is constructed (see Table 1) [7]: Table 1. The statistical sample of UAV design parameters. a c N i where N - the volume of the statistical sample, a - the vector of UAV design parame- ters, cΣi - the total cost of i-th acceptable variant of building a multi-purpose UAV sys- tem. Step 2. Based on the obtained statistical samples, in the class of power polynomials, for each payload using UAVs, dependencies  Ci  Ci a , i  1,8 are constructed, and  optimization problems Ci  min Ci a , i  1,8 are solved by the generalized method of aD 218 dynamic condensation, which determines the optimal vectors of UAV design parame- ters a opt , i  1,8 and the minimum cost of a multi-purpose UAV system. Step 3. To implement strategies for structural selection of optimal variants of a multi- purpose UAV system, all design parameters a are converted to a dimensionless form, and new variants of the structure of UAV systems are formed for different accuracy εopt(i,j) of building a multi-purpose UAV system. Step 4. The total cost of a multi-purpose UAV system was calculated using the for- mula [10, 12]: s p C    n j c j , (14) i 1 j 1 where s is the number of UAV types in a multi-purpose UAV system, p is the number of payloads (PDC) delivered by the i-th type of UAV, nj is the number of UAVs re- quired to service the j-th payload (PDC). The matrix of configuration options for a multi-purpose UAV system has the form (Table 2). Table 2. Matrix of options for completing a multi-purpose UAV system. x1 x2 … xnц J(X,A,E(x)) (a1)1,j1 (a1)2,j2 … (a1)nц,jnц J1(X,A,E(x)) (a2)1,j1 (a2)2,j2 … (a2)nц,jnц J2(X,A,E(x)) … … … … … (an)1,j1 (an)2,j2 … (an)nц,jnц Jnц(X,A,E(x)) is the optimality criterion of the i-th target task, i  e(i, j )  is a vector of opt Here, J i characteristics of i-th target task, а j is the vector of design parameters characterizing the j-th type of UAV [10, 13].  min а j , i  e(i, j )  , opt Ji а j A   (15) The vector of characteristics of the i-th target problem for the j-th type of UAV i(e(i,j)) characterizes the dependence of the optimality criterion J iopt on which type of UAV the i-th target problem is performed in the interests of the UAV system. Be- sides, for each target, you must generate the corresponding optimality criterion, in par- ticular, these criteria may coincide. Based on the values of partial optimality criteria, a generalized integral criterion for structural and parametric selection of a multi-purpose UAV system is formed [14]:  J opt  F J1, J 2 ,..., J i ,..., J nц . (16) 219 Here F(…) is a convolution of partial criteria of optimality J1,J2,…,Ji,…,Jnij. In par- ticular, when solving special problems, the most common convolution is the additive convolution of particular criteria, which is typical for evaluating technical, economic, and cost criteria [17,22]. Optimal accuracy of building a multi-purpose UAV system is achieved by formaliz- ing algorithmic procedures that exclude subjectivism when justifying the structure of a multi-purpose UAV system. Previously, the corresponding cost of the structure of a multi-purpose UAV system is determined for various specified values and a statistical sample of the type is formed (Table 3) Table 3. The statistical sample of the cost of a multi-purpose UAV system structure. ε J(A,W,E()) ε1 J1(A,W,E()) ε2 J2(A,W,E()) … J(A,W,E()) εN JN(A,W,E()) Based on the data from a statistical sample, a dependency of the following type is formed [15]:  2 n J  A,W , E ( )   c1 1  c2  ...  cn , (17) Linear and nonlinear parameters 1 2c , c ,..., c ,  n 1 2,  ,...,  n are determined from the minimum condition of the regularity criterion, and the optimal accuracy of the struc- ture of a multi-purpose UAV system is determined from the condition: J  A, W , E ( )  opt 0 ,  ( ) (18) A special case of building a multi-purpose UAV system is a scenario where the j-th target task is characterized by the delivered payload mass mplj with the launch fre- quency nj, where j  1, n . The cost criterion CΣ for a multi-purpose UAV system is used as an optimality criterion. It is assumed that if at a given value P(mplj) is the thrust of the UAV propulsion system, the design solution allows you to deliver a payload of mass mplj to a given height Нgv, then the same type of UAV can deliver payloads of a smaller mass mplj-1, mplj-2,…, mpl1 to the same height. Based on statistical samples from table 4, for each target problem, polynomials in the power function class are restored from a given external set of targets [15]:  (1) 1 (1) (1)  2 (1) (1)  n(1) C1  C1 d1  C2 d 2  ...  Cn d n ,  C2  C1(2) d11  C2(2) d 2 2  ...  Cn(2) d n n , (2) (2) (2)  ... ... ... ... ... ... ... ... ... ... ... ...  ( n) Cn  C1( n ) d11  C2( n) d 2 2  ...  Cn( n) d n n . (n) (n) (19) 220 A generalized criterion for the cost of creating a multi-purpose UAV system of op- timal structure can be presented in the following form [10]: C  J  X , A, E ( x)  , (20) which characterizes, in General, the total cost of creating a multi-purpose UAV sys- tem of the optimal structure taking into account the features of its design technical char- acteristics and the diversity of the target tasks to be solved. The cost of a multi-purpose UAV system is defined as the sum of the costs of the i- th UAV structure, which are necessary for the formation of this variant of a multi-pur- pose system, and the cost of R & d in the development of appropriate types of UAVs. R & d costs for the development of i-th type UAVs are defined as: Ci R &d  Cyd М 0i , (21) where Cyd is specific R & d costs, M0i is the starting mass of the i-th type of UAV. Table 4. The statistical sample of options for completing a multi-purpose UAV system. The total cost of a multi-purpose UAV system is as follows: n n C   C j   C j R & d , j 1 j 1 (22) (1) The obtained value of the optimality criterion Arg min C  C   is compared 1opt with the best extreme value of the target function: (1) C  C   , (23) where δ - the given accuracy of cost estimates of variants of the structure of a multi- purpose UAV system. (k ) ( k 1) For the k-th step we have: С  C  C , and local optimization is per- formed according to the criterion [10]:  2 С( k )   2 С( k )  (k ) ( k )   x1 cos    x2 sin   (k )  x3  , C  C   , (24) J  min С  C (k ) ,  x3  221 The condition for the end of the process of optimizing the accuracy of the structure of a multi-purpose UAV system is the output of the variable parameter ε to the given constraints. The optimal multi-purpose UAV system was built from the condition of delivery of eight target remote sensing data collection platforms (PDC) remote sensing of the Earth (RSE) to various points (targets) in the South China sea. The origin of the coordinates was located at the start point of the UAV in the area of da Nang, Vietnam. Target remote sensing data collection platforms (PDC RSE) for monitoring tropical cyclones were located at points with coordinates (figure 2) (in meters): 1. p. А (350 000; 50 000) 5. p. E (250 000; 40 000) 2. p. В (300 000; 80 000) 6. p. F (310 000; 10 000) 3. p. С (35 000; 20 000) 7. p. G (360 000; 90 000) 4. p. D (280 000; 70 000) 8. p. H (250 000; 10 000) Fig. 2. Point of delivery platforms to collect remote sensing data RSE using a UAV (in km). 5 Simulation Result The probabilities of completing the target tasks P1,P2,…,P8 and the number of UAV launches n1,n2,…n8 are shown in Table 5. The required number of UAVs for a multi-purpose system with the required target payloads (PDC) m pl i , i  1,8 is determined based on the given efficiency Pi of the de- livery of the i-th PDC to the specified area. Table 5. Structural-parametric synthesis of a multi-purpose UAV system. P1 P2 P3 P4 P5 P6 P7 P8 0,6 0,7 0,65 0,5 0,8 0,75 0,8 0,8 n1 n2 n3 n4 n5 n6 n7 n8 4 3 3 5 2 3 2 2 222 Probability of covering the i-th point of the goal [4, 5]: ny0 Pcovi  .e  i, j  , (25) n where ny0 is the number of successful realizations (when there is a convergence with the i-th goal); nΣ is a total number of realizations. The required number of UAVs of the i-th type is defined as [6]: lg 1  Р  ni  , (26) lg 1  Рi  where PΣ is the total efficiency of a multi-purpose UAV system, which in this task is assumed to be equal to PΣ =0.7. The following numerical results are obtained: Parameters The distribution of the targets according to the types of UAV X(1) = 1,367 e(1,1) =1- the first target task is solved by 1-th type of UAV; X(2) = 2,892 e(2,2) =1- the second target task is solved by 2-th type of UAV; X(3) =1,245 e(3,1) =1- the 3-th target task is solved by 1-th type of UAV; X(4) =2,080 e(4,2) =1- the 4-th target task is solved by 2-th type of UAV; X(5) =2,591 e(5,2) =1- the 5-th target task is solved by 2-th type of UAV; X(6) =3,969 e(6,3) =1- the 6-th target task is solved by 3-th type of UAV; X(7) =3,003 e(7,3) =1- the 7-th target task is solved by 3-th type of UAV; X(8) =3,479 e(8,3) =1- the 8-th target task is solved by 3-th type of UAV. Thus, the obtained multi-purpose UAV system of an optimal structure consists of three types: the first type of UAV serves target platforms with masses mpl=150kg and mpl=250kg; the second type of UAV serves platforms with masses mpl=200kg, mpl=300kg, and mpl=350kg; the third type of UAV serves platforms with masses mpl=400kg, mpl=450kg and mpl=500kg. The total cost of the optimal structure of a multi-purpose optimal UAV system is: opt 6 С  0,86  10 y.e. , which gives a cost-benefit of about 4-5% less than using a UAV system with the same type of specialized UAVs. 6 Conclusions The developed generalized method of dynamic condensation (GMDC) consists in com- prehensive use of the possibilities of solving a class of optimization problems in terms of justification and selection of structures of multi-purpose UAV systems of a given accuracy both in the space of its inverse characteristics and taking into account the possibilities of final prioritization of the distribution of target tasks by types of UAVs and restrictions on their design parameters. As a result of solving the problem of finding the optimal design solution, the struc- ture of a multi-purpose UAV system costs about 4-5% less than the use of a system with the same type of specialized UAVs is obtained. 223 The results of the solution of the formulated scientific problem should be used in the future when forming the tactical and technical requirements (TTZ, TZ) of the Customer for the creation (modernization) of promising multi-purpose UAV systems when solv- ing remote sensing problems in terms of evaluating the effectiveness of their function- ing, and evaluating the resource provision, in particular, costs. References 1. Garbuk, S. V., Gershenson, V. E.: Space Systems of Remote Sensing of the Earth. Moscow: Publishing house A and B, (1997). 2. Hain, A. P., Sutyrin, G. G.: Tropical Cyclones and their Interaction with the Ocean. L.: Hy- drometeoizdat, (1983). 3. Hain, A. P.: Mathematical Modeling of Tropical Cyclones. L.: Hydrometeoizdat, (1984). 4. Zbigniew, Koruba, Edyta, Ładyżyńska-Kozdraś: The Dynamic Model of a Combat Target Homing System of an Unmanned Aerial Vehicle. Journal of Theoretical and Applied Me- chanics. Warsaw 2010, 48, 3. P. 551−566. (2010) 5. Cao, J. et al.: Research of Integrated Design about Guidance and Control of Homing UAV. Applied Mechanics and Materials. V. 373−375. P. 1428−1431. (2013). 6. Gundlach, J.: Designing Unmanned Aircraft Systems. A Comprehensive Approach. AIAA, Reston. Virginia. (2012). 7. Balyk, V. M., Malenkov, A. A.: Method of Dynamic Condensations in the Problem of Con- structing a System of Aircraft of the Optimal Type. 14th international conference "aviation and cosmonautics-2015", Moscow aviation Institute (national research University), Mos- cow, Russia, November 16-20, (2015). 8. Nguyen, Q.Th., Nguyen, Th.L.: The Problem of the Optimal Type of Unmanned Aerial Ve- hicles by a Generalized Method of Dynamic Condensations. Journal Science-Intensive tech- nologies, V. 2, pp. 15-20. (2016) 9. Nguyen, Q.Th., Nguyen, Th.L.: Modification of the Method of Dynamic Condensations in the Problems of Building a System of Aircraft of the Optimal Type. Journal Success of modern radio electronics, V. 6, p. 81-85. (2016) 10. Balyk, V.M.: Statistical Synthesis of Design Solutions in the Development of Complex Sys- tems. Moscow: MAI Publishing house, 280 p. (2011) 11. Thuong, N.Q., Dong, P.T., Long, N.T.: A Structural and Parametric Synthesis Method for Type-Optimal Multipurpose System of Unmanned Aerial Vehicles (UAVs) Sensing Tropi- cal Cyclones, International Journal of Applied Management Science. Vol.58; V. 3; (2019). 12. Box, G. E. P., Jenkins, G. M., Reinsel, G. C.: Time Series Analysis: Forecasting and Control. 3rd ed. Englewood Cliffs, NJ: Prentice Hall, (1994). 13. George, E. P., Box, Gwilym, M., Jenkins, Gregory, C. Reinsel: Time Series Analysis: Fore- casting and Control. Wiley Series in Probability and Statistics. PP. 92-161. (2013). 14. Nocedal, J., Wright, S.J.: Numerical Optimization, Springer, (1999). Woodward, W.A., Gray, H. L. & Elliott, A. C. Applied Time Series Analysis. CRC Press. PP. 197-230. (2012).