RU2781391C2 - Method for decentralized control of distributed generalized network including movable robotic means and docking stations - Google Patents

Abstract

FIELD: robotics.

SUBSTANCE: invention relates to a method for decentralized control of a distributed generalized network of movable robotic means and docking stations. For decentralized control, decentralized interaction is performed between each pair formed by a movable robotic means performing the movement and a docking station. Maintenance of robotic means or their disabling are determined by an operator or an automated command link in a certain way by the specified criteria of network functioning, taking into account accumulated information about previous interactions.

EFFECT: increase in the efficiency of autonomous functioning of a group of movable robotic means.

1 cl, 2 dwg

Description

The invention relates to control methods in technical robotic distributed systems, including objects of two different types, between pairs of heterogeneous objects of each such system, interaction must be established and performed, with one or another periodicity, in order to achieve the goal of functioning of this system.

In [1 - A Distributed Framework for Energy Trading Between UAVs and Charging Stations for Critical Applications)), Vikas Hassija, Vinay Chamola, Dara Nanda Gopala Krishna and Mohsen Guizani, Fellow IEEE, 2020], a dialog model of interaction between a charging station and an UAV is considered group, which allows balancing the supply and demand for electricity, as well as the safe and reliable purchase of energy by UAVs from the station.

In [2 - "Energy-controlled Optimization Algorithm for Rechargeable Unmanned Aerial Vehicle Network)", Li Li, Jie Wu, Yixiang Xu, Jun Che, Jin Liang, 2017], the problem of maximizing the operation time of each UAV group is solved based on the solution of the corresponding traveling salesman problem.

The models of interaction between UAVs of a group and recharging stations considered in articles [1-2] do not take into account local factors that affect the efficiency of UAVs performing their tasks and are caused by preparation for interactions with individual stations and these interactions themselves; do not take into account the evolution of these factors over time. In addition, these models assume an unlimited supply of energy at each station, which can only be the case when such stations are connected to a centralized power grid. This is a significant limitation of such models, which exclude the consideration of recharging stations based on renewable energy sources.

Works [1-2] will be considered analogues.

Of the known technical solutions, the closest in technical essence to the claimed method is "MULTI-USE UAV DOCKING STATION SYSTEMIS AND METHODS)) (patent US 9387928B1, publ. 12.07.2016). This patent considers a generalized centralized method for managing the operation of a group of mobile autonomous objects of various nature for the implementation of such tasks as the transportation of goods, recharging and repair at service stations, etc., taking into account feedback from sensors that measure the current values of the spatial field of quantities, characterizing the environment of functioning of a group of mobile robots.

This patent discloses systems and methods for ensuring the interaction of multi-purpose docking stations with mobile robotic vehicles (MRS). Docking stations can be networked with a central control panel and multiple PRSs. Docking stations may include a number of services to facilitate the pointing and maintenance of the MRS. Docking stations may include equipment for the handling of delivered goods/packages and may serve as both the final destination and the delivery hub. Docking stations can expand the coverage of the PRS by providing the latter with the ability to recharge / refuel. In the case of PRS with electric propulsion, recharging can be contact or non-contact.

The docking stations may include navigation aids for guiding the drones to the docking stations and providing route information from a central control unit. These stations can be equipped with special measuring equipment for estimating the current characteristics of the states of the MRS, in particular, unmanned aerial vehicles (UAVs), the station itself, the environment in the vicinity of the station (for example, meteorological equipment).

Docking stations can be built into existing structures such as cell towers, lighting and electrical poles, and buildings. Docking stations may also include stand-alone structures to provide value-added services in underserved areas.

The basis of the proposed in the prototype method for planning the trajectories of the movement of each RRS of the group and managing the interaction of the RRS with docking stations is the centralized decision-making by the operator based on: a) the goals and objectives of the functioning of the RRS group; b) current data coming from the measuring equipment of the stations and, possibly, from each ORS of the group; c) other data coming from other sources and related to the current or future processes of the functioning of the PRS group and docking stations.

The disadvantages of the method claimed in the prototype are: 1) the lack of decentralized control in the system, which significantly limits the scope of the method of this patent. The degree of this limitation grows with the increase in the number of elements of the system, the number of potential modes of their operation and the area of the zone of its action. In addition, the reliability of the functioning of the entire system and the probability of successful completion of the mission are reduced due to the significant risk of failure of the control center in the absence of redundancy of its functions. 2) the inability to use the accumulated information about the previous states of the system to optimize the process of subsequent interactions of moving and stationary objects of the system according to a given criterion of its functioning, i.e. the lack of self-learning of the system to better solve the tasks assigned to it based on the experience it has accumulated in the past.

Fully centralized control of such a system is justified in those cases or those modes when the control of the operator is required, coordinating the actions of all elements of the system at the top level of its control. However, with an increase in the number of elements of the group and an increase in the complexity of taking into account the entire complex of processes that determine the functioning of this group, a combination of centralized and decentralized management methods becomes a more priority solution. The complex of problems of decentralized management of such a group includes the task of organizing the optimal interaction of its elements in order to achieve a certain goal of its functioning by the group.

In the case of organizing energy exchange between mobile mobile objects of the system and its stationary elements - recharging stations, the solution of this problem plays a key role in developing the concept of optimal energy consumption of this group.

The technical result of the proposed method. The technical result of the invention is to increase the efficiency of the autonomous operation of a group of mobile robotic means, in particular, to reduce the probability of failure of its elements, to reduce the mission time for each element of the group and the group as a whole, to reduce the participation of the operator or the centralized control link in coordinating the actions of the group, to increase the efficiency of solving typical problems with an increase in the time of self-learning of the system, including due to adaptive adjustment of the configuration of the static elements of the system or / modes of operation of the PRS, the possibility of functioning of the group of PRS in conditions of limited energy resources of recharging stations.

This result is based on the concept of decentralized interaction between mobile and stationary objects of some technical system, which implies the possibility for each agent to choose the potential interaction of the corresponding companion; the basis of such a choice is the consideration of the prehistory of the process, when in order to make a decision - the choice of an object for interaction, both current information about these objects at a given point in time and the evaluated results of previously made choices that have shown their respective effectiveness in relation to some target criterion are used.

The integration of the proposed method with methods for planning and controlling the movement of each mobile robotic vehicle, as well as methods for controlling the operation of each station, ensures the autonomous execution of the mission by the RRS group with the achievement of a given target quality of this performance under given restrictions.

The technical result is achieved by the fact that the claimed method performs decentralized control of a distributed generalized network, including mobile robotic tools and docking stations, in which each network element is equipped with a module for organizing the reception and transmission of information signals, a computing module for autonomous analysis and decision-making on interaction with others. network elements, sensors for measuring characteristic intrinsic values or values of the environment, mobile robotic vehicles are additionally equipped with navigation modules and special autopilots, payload modules, these vehicles carry out autonomous or remotely controlled movement between destinations, including the stages of docking and landing on specially equipped docking station sites, in which the initial routes of each mobile robotic vehicle, as well as the moments when network elements require significant correction adjustment of their operating modes, repair maintenance or their failure are determined by the operator or an automated command link based on the purpose and objectives of the functioning of the entire system. The proposed method implements decentralized regulation of interactions between each pair, including a mobile robotic vehicle and a docking station, the decision on a potential interaction between each such pair is based on the choices made by both of its elements or one of them, with the possibility of using the accumulated information about previous interactions for optimization the process of subsequent interactions of mobile robotic means and docking stations according to the specified criteria for the functioning of the network, in which forced connection of an operator or an automated command link to the control loop is allowed. Based on the results of functioning of a distributed generalized network, it becomes possible to adaptively adjust the configurations of docking stations and the modes of operation of mobile robotic means to optimize the process of network functioning according to specified quality criteria.

The analysis of technical solutions made it possible to establish that analogues, characterized by a set of features identical to all the features of the claimed technical solution, are absent in known media, which indicates that the claimed method complies with the patentability condition "novelty".

The search for known solutions in this and related fields of technology in order to identify the claimed set of features that are different from the prototype showed the absence of both an explicit formulation of such features, and any technical solutions where the specified set of features would be fully implemented.

The analysis of the present level of technology did not allow us to conclude that the impact of the transformations provided for by the essential features of the claimed invention on the achievement of the specified technical result is known. Therefore, the claimed invention meets the condition of patentability "inventive step".

The claimed method is applicable to the adaptive control of the process of functioning of a generalized two-connected technical system - a distributed generalized network (GRN), including a static distributed subnet (SRP) and a dynamic distributed subnet (DRP). The SRP includes stationary objects (hereinafter referred to as docking stations or simply stations) that perform certain functions of traffic support and / or maintenance of mobile robotic vehicles included in the DRP.

This method makes it possible to optimize the control of the ROS operation processes according to the specified criteria: a) maximum autonomy of operation / minimum participation of the operator in the control loop and maintenance of the ROS elements; b) the criterion for the speed of the mission; c) the minimum percentage of failures of system elements; d) minimization of imbalances in energy and information interactions between ROS elements.

In addition, the proposed method allows for a posteriori adjustment of the configuration of the static elements of the system and the modes of operation of its dynamic elements - PRS, in order to achieve a given quality criterion.

The key process of the functioning of the ROS is the interaction between its elements to perform the tasks or achieve a specific goal. For example, the purpose of functioning can be: a) organizing the flow of goods between a number of points for receiving / issuing goods; b) organization of monitoring of the area; c) mapping; d) conducting a rescue operation; e) conducting a combat operation in a certain territory; f) carrying out a clearing / demining operation, etc.; g) other.

Interaction between ROS elements may include: information exchange; loading and unloading of components transported by the RRS at the station; energy exchange between the PRS element and the PRS, which, as a rule, comes down to recharging the PRS at the station; other.

Stationary facilities may be stations that perform one or more of the following functions: a) loading and unloading of goods/cargoes transported by the MRS; b) replenishment of energy losses of individual PRS - recharging due to a centralized supply of electricity or the generation of energy of one nature or another using renewable sources; c) PRS repair; d) other. These stations can be additionally equipped with measuring equipment, in the general case, fixing the current characteristics of the states of the MRS, the station itself, the environment in the vicinity of the station (for example, using a weather station). PSA objects, in addition to docking stations, may include various transceiver equipment with an appropriate control unit that generates, receives, transmits or retransmits signals of any nature (radio, optical, acoustic, etc. signals) for the exchange of information between all or part of the elements of a distributed generalized network.

Mobile robotic means can be: a) UAVs; b) autonomous robotic boats; c) autonomous robotic underwater vehicles; d) autonomous robotic aeronautical platforms.

A variant of implementation of ROS is a distributed generalized power grid (DSG), in addition, DS can be a DSW and perform additional functions of servicing the DC.

An important special case of ROES is a distributed network, including recharging stations of the PRS and a group of PRS; the task of such a network is to fulfill a given mission by a group of PRS with the possibility of recharging with energy at the stations of the network.

The claimed method, together with methods for planning and controlling the movement of each RRS, as well as methods for controlling the operation of each charging station, ensures the autonomous execution of the mission by the RRS group with minimization of the time for this execution under given restrictions on the percentage of failures of the ROS elements.

The proposed method is implemented in a distributed generalized network control system, including mobile robotic devices and docking stations, the functional diagram of which is shown in Fig. one.

Before the start of the functioning of the distributed generalized network, the goal is formed and the tasks of the corresponding mission are developed in block 1, then a rule is established according to which the operator or/automated centralized command link can interfere in the ROS control process at each or some of the stages of this control. In one limiting case, only the operator has the right to make decisions at all stages - a centralized control mode is obtained, in the other - the influence of the operator / centralized link is completely excluded, at least until the end of the mission. In the general case, the operator / centralized link can intervene in the process with one or another periodicity.

Then, in block 3, the interaction of pairs of elements of the SRP and DRP is planned, based on the algorithm below (Fig. 2), in block 4, the planning of the movements of the PRS and the operation of stations is performed, providing, without taking into account random and a priori uncertain disturbances within the system, the implementation of a map of planned interactions , outlined in block 3. Block 5 provides low-level regulatory control for each of the ROS elements, ensuring the minimum possible deviation of the real trajectories of the elements in the phase spaces of their state variables from the corresponding program trajectories laid down in block 4. Implementation of self-learning, acquisition of "experience" ROS elements in relation to the criteria for the best performance of the tasks imputed to them, is provided by block 6, where the weight coefficients are adjusted - the parameters of the algorithm for the decentralized selection of interacting heterogeneous ROS elements. In the case when it is possible or expedient to change the configuration, the mutual arrangement of the elements of the PDS or the setting values of each of the stations, as well as to adjust the characteristic modes of movement of the PRS to optimize the given quality criteria of the ROS, block 7 is used; the results of such changes made in block 7 according to some adaptive algorithm, correct the sequential operation of blocks 3-6; however, block 7 can be turned on no more than after an interval of time necessary and sufficient for the accumulation of relevant information about the functioning of the ROS in the previous configuration and with the previous values of the set values of its elements.

Let there be a set of fixed destinations, goods acceptance and delivery points and / or recharging stations, etc., in a certain territory, between which transport communication is carried out using mobile robotic objects.

, а объекты противоположной статической/динамической подсистемы - к классу

.Interaction with a part of the stations can be planned in advance for the PRS, and the possibility of interaction with the rest of the stations is generally determined by the proposed method. In this case, the objects of the dynamic/static distributed subsystem will be referred to the class

, and objects of the opposite static/dynamic subsystem - to the class

.

In the process of movement, the PRS can use the services of one of the stations located in the designated area. Likewise, stations may also select individual RSPs to interact based on the information available to them, based on experience gained.

In the general case, in order to make a decision on the interaction of a given pair, it is necessary and sufficient to fulfill a strict condition so that each of the elements of this pair chooses the opposite element of this pair, in the way that is incorporated in the claimed method. It is also possible that the decision on interaction can only be made by objects of one of the classes - SRP or DRP.

класса

для объекта

класса

в текущей ситуации с номером s, когда системе управления

требуется принять решение о взаимодействии с некоторым объектом противоположного класса в отношении реализации некоторой задачи или достижения цели функционирования отдельных ПРС или распределенной динамической подсети в целом. Такие задачи мы будем называть базовыми; ими могут быть: 1. подзарядка отдельных ПРО в процессе движения по произвольным или регулярным траекториям; 2. организация обмена товарами/грузами между ПРО и пунктами отправки назначения; 3. прочие задачи. Каждая базовая задача индуцирует соответствующую относительную эффективность данного объекта одного класса по отношению к некоторому другому объекту противоположного класса при их возможных взаимодействиях в данный момент времени.Let us introduce the concept of the relative efficiency of an object

class

for object

class

in the current situation with number s, when the control system

it is required to make a decision on interaction with some object of the opposite class in relation to the implementation of a certain task or the achievement of the goal of the functioning of individual DRSs or a distributed dynamic subnetwork as a whole. We will call such tasks basic; they can be: 1. recharging individual missile defense systems in the process of moving along arbitrary or regular trajectories; 2. organization of the exchange of goods/cargo between missile defense and points of departure and destination; 3. other tasks. Each basic task induces a corresponding relative efficiency of a given object of one class in relation to some other object of the opposite class in their possible interactions at a given time.

функционал, количественно выражающий относительную эффективность потенциально возможного взаимодействия

р-го объекта класса

с q-м объектом класса

на шаге s_p в отношении данной базовой задачи, решаемой р-м объектом класса

пусть, далее,

- функционал, количественно выражающий относительную эффективность того же взаимодействия, но в отношении базовой задачи, решаемой q-м объекта класса

на шаге s_q, причем шаги s_p и s_q относятся к одному и тому же времени.Let

functional expressing quantitatively the relative efficiency of a potential interaction

p-th class object

with q-th class object

at step s _p with respect to the given basic problem solved by the p-th object of the class

let, further,

- a functional that quantitatively expresses the relative efficiency of the same interaction, but in relation to the basic task solved by the q-th object of the class

at step s _q , and steps s _p and s _q refer to the same time.

и

; во-вторых, накопленный к настоящему времени опыт взаимодействия

со объектами класса

и

с объектами класса

Возможен вариант, когда, например, каждому объекту

становится доступен аналогичный предыдущий опыт всех других объектов класса

These functionals should take into account: firstly, the current situation, which imposes spatio-temporal, informational, energy, and so on. restrictions on the ability to interact

and

; secondly, the experience of interaction accumulated to date

with class objects

and

with class objects

It is possible that, for example, each object

the same previous experience of all other objects of the class becomes available

Эти факторы будем называть частичными. Ими могут быть: 1. степень энергоемкости станции для данного ПРС в данной ситуации, выражаемая как разность между заявленной станцией энергией, потенциально доступной для данного ПРС, и энергетическим запросом самого ПРС с учетом движения его до этой станции; 2. затраты дополнительного времени на достижение станции, последующую зарядку и процесс возврата на целевую траекторию, выражаемые как реальное время подзарядки плюс время достижения самой станции; 3. точность оценивания времени движения до ПРС, определяющаяся в том числе особенностями расположения инфраструктуры и влияния природных факторов в малой окрестности станции, может быть выражена как разность между оцениваемым и реальным временем по п. 2; 4. возможность технического обслуживания данной станцией данного ПРС; 5. возможность принятия груза, перевозимого данным ПРС на данной станции; 6. прочие факторы и соответствующие величины.Consider the factors affecting the value of the functional

These factors will be called partial. They can be: 1. the degree of energy intensity of the station for a given PRS in a given situation, expressed as the difference between the energy declared by the station, potentially available for this PRS, and the energy demand of the PRS itself, taking into account its movement to this station; 2. the cost of additional time to reach the station, subsequent charging and the process of returning to the target trajectory, expressed as the real time of recharging plus the time to reach the station itself; 3. the accuracy of estimating the time of movement to the MRS, which is determined, among other things, by the peculiarities of the location of the infrastructure and the influence of natural factors in a small vicinity of the station, can be expressed as the difference between the estimated and real time according to paragraph 2; 4. the possibility of maintenance by this station of this PRS; 5. the possibility of accepting the cargo transported by this PRS at this station; 6. other factors and relevant quantities.

поставим в соответствие частичный функционал

, зависящий от а) вектора X_pq(s) всех учитываемых величин, влияющих на эффективность взаимодействия

и могущих быть измеренными р-м объектом класса

б) вектора

всех учитываемых величин, влияющих на эффективность взаимодействия

и могущих быть измеренными q-м объектом класса

Для оценки такого функционала, произведенной р-м объектом класса

априорно, до взаимодействия с q-м объектом класса

можно записать:To each k-th partial factor

we associate the partial functional

, depending on a) the vector X _pq (s) of all quantities taken into account that affect the efficiency of interaction

and can be measured by the p-th object of the class

b) vector

all taken into account quantities that affect the effectiveness of interaction

and can be measured by the qth object of the class

To evaluate such a functional, produced by the p-th object of the class

a priori, before interacting with the q-th class object

can be written:

в следующем общем виде:The basis of the proposed method is the representation of the functional

in the following general form:

Here

, определенных по (1) на основании измерений

и

- K × 1 -vector of estimates of all partial functionals

determined by (1) based on measurements

and

на всех предыдущих этапах взаимодействий

с объектами класса

определенных на основании измеренных векторов

, и интегрирующий в себе информацию об эффективности влияния каждого k-го фактора на результат этих взаимодействий

в прошлом. Значениями векторов

и

элементы противоположных классов должны обмениваться через соответствующий информационный канал для корректного выбора подходящего объекта для взаимодействия.- 1×K - vector of weight coefficients depending on the array of dimension s _p ×K values of functionals

at all previous stages of interactions

with class objects

determined based on the measured vectors

, and integrating information about the effectiveness of the influence of each k-th factor on the result of these interactions

in the past. Vector values

and

elements of opposite classes must be exchanged through the appropriate information channel to correctly select the appropriate object for interaction.

получаются из соответствующих равенств (2)-(4) заменами

в том числе в нижних индексах.Expressions for the functional

are obtained from the corresponding equalities (2)-(4) by the replacements

including subscripts.

The functions of many variables F and f are determined by the features of the ROS and the tasks it performs.

A distinctive part of the proposed method corresponds to an algorithm for decentralized regulation of interactions between each pair of mobile and stationary network objects, a generalized block diagram of which is shown in Fig. 2; this algorithm is implemented in on-board computers of all or some elements of a distributed generalized network. Let us put for definiteness

On FIG. 2, the following designations are introduced: 1 - incrementing the number of the element of the pair at the first position; 2 - incrementing the number of the element of the pair at the second position; 3 - checking the condition: "Are there any pairs at the moment of time, the interaction between which is required to be implemented?"; 4 - implementation of previously planned interactions; 5 - adjustment of the weights of all elements of the system, determined by interacting elements; 6 - checking the conditions for forcing the interaction of this pair; 7 - forced assignment of a pair for interaction; 8 - the implementation of information exchange between the elements of the pair; 9 - Checking the condition: "the current number of the element of the pair in the second position is less than the total number of elements of the second class (B)"; 10 - calculation of the value of the target functional, determined by the element in the first position; 11-calculation of the value of the target functional, determined by the element in the second position; 12- checking the condition: "the number of the element in the second position is equal to the total number of elements of the second class (B)"; 13 - optimization of the functional for the element in the first position with obtaining the number of the corresponding optimal element in the second position (second class); 14 - checking the condition: "the number of the element in the first position is equal to the total number of elements of the first class (A)"; 15 - optimization of functionals for each element in the second position with obtaining the numbers of the corresponding optimal elements in the first positions (class A); 16 - determination of all pairs of elements that satisfy the cross criterion of optimal selection; 17 - adding pairs determined by the forced choice condition; 18 - planning interactions of selected pairs.

и

реализация принудительного, производимого оператором назначения пар, подлежащих взаимодействию, производится в блоках 6 и 7 (если таковое назначение имеет место). Если после этого не достигается последний элемент первого класса А, т.е. не выполнено условие p=N_A блока 14, то происходит возврат к пункту инкрементирования номера р, иначе - производятся действия, описываемые далее. Если условие блока 6 не выполнено, тогда в блоке 8 проверяется условие возможности осуществления информационного обмена парой (р, q); если такой обмен не возможен, то производится переход к следующему элементу через блоки 10 и 14 проверок условий q<N_B и p=N_A, или же - переход к этапу обработки данных после циклов (блок 15 и следующие). Если указанный обмен возможен, то он производится, и по его результатам рассчитываются функционалы Ф_АВ(р,q,s_p) и Ф_BA(q,p,s_q) в блоках 9 и 11. Если далее выполнено условие q=N_B блока 12, то в блоке 13 производится оптимизация функционала Ф_АВ(р,q,s_p) q с получением оптимального элемента q_opt для данного р; затем, после проверки условия блока 14, происходит переход к следующему номеру р, либо к блоку 15 оптимизации функционала Ф_BA(q,p,s_q) по р с получением оптимального элемента p_opt для данного q, исключая принудительно отобранные пары. На основании результатов работы блоков 13 и 15 в блоке 16 производится отбор всех пар

удовлетворяющих условию: p_opt(q,s_q)=p^q_opt(p,s_p)=q (в количестве М' штук, если М'=0, такие пары отсутствуют). Наконец, в блоке 17 происходит добавление принудительно отобранных пар с формированием полного множества пар (p_m,q_m),

, взаимодействие элементов которых планируется в блоке 18. На данном временном шаге может оказаться М=0.The logic of the algorithm corresponding to the flowchart in FIG. 2 next. After incrementing the numbers of both elements of the pair (blocks 1 and 2), in blocks 3 and 4, the existence of pairs is checked, the interaction between the elements of which is to be implemented at a given time step of the processor time of on-board computers (the times of on-board computers of various ROS elements are synchronous) and the implementation of this interaction ; based on the results of these interactions in block 5, the values of all partial functionals (1) are measured and then the updated values of the weights δ for the objects p and q are calculated using (4):

and

the implementation of the forced assignment of the pairs to be interacted by the operator is carried out in blocks 6 and 7 (if such an assignment takes place). If after that the last element of the first class A is not reached, i.e. the condition p=N _A of block 14 is not met, then the return to the point of incrementing the p number occurs, otherwise, the actions described below are performed. If the condition of block 6 is not met, then in block 8 the condition of the possibility of carrying out information exchange of the pair (p, q) is checked; if such an exchange is not possible, then the transition to the next element is made through blocks 10 and 14 of checking the conditions q<N _B and p=N _A , or - the transition to the stage of data processing after the cycles (block 15 and the next). If the specified exchange is possible, then it is performed, and according to its results, the functionals Ф _АВ (р, q, s _p ) and Ф _BA (q, p, s _q ) are calculated in blocks 9 and 11. If further the condition q=N _B block 12, then in block 13 the optimization of the functional Ф _АВ (р,q,s _p ) q is performed to obtain the optimal element q _opt for the given р; then, after checking the condition of block 14, there is a transition to the next number p, or to block 15 of optimizing the functional Ф _BA (q,p,s _q ) for p with obtaining the optimal element p _opt for this q, excluding forced selection of pairs. Based on the results of blocks 13 and 15, in block 16, all pairs are selected

satisfying the condition: p _opt (q,s _q )=p^q _opt (p,s _p )=q (in the amount of M' pieces, if M'=0, there are no such pairs). Finally, in block 17, the forcibly selected pairs are added to form the complete set of pairs (p _m ,q _m ),

, the interaction of elements of which is planned in block 18. At this time step, M=0 may turn out.

Algorithms implying the possibility of choosing only for objects of one class A or B can be obtained from the considered algorithm. If elements of class B are prohibited from selecting the most suitable elements of class A for interaction, then from the block diagram in FIG. 2, the blocks associated with the calculation and optimization of functionals of the type Ф _BA (q,p,s _q ) should be excluded, as well as the corresponding weight coefficients and partial functionals are not calculated; similarly, if such a ban is imputed to elements of class A, then, on the contrary, from the block diagram in FIG. 2, the blocks associated with the calculation and optimization of functionals of the type Ф _АВ (р, q, s _p ) should be excluded, the corresponding weight coefficients and partial functionals are also not calculated.

Consider the specification of the proposed method in relation to the ROS containing stations for recharging the UAV. Let the UAV group for the ROS be a DSP, and the SRP be a set of stations with which each UAV can dock in one way or another to perform the tasks of the ROS, and each station can be equipped with an automatic UAV recharging module. As a basic task to be implemented by the ROS, we define the organization of an optimal speed transport flow between destinations - the stations themselves. We will assume that the UAV group is homogeneous; some of the stations are equipped with recharging modules (MP) of the UAV, and the other part is not.

Each UAV, as soon as its energy reserve drops below the threshold value, can land at one of the stations equipped with MP, the UAV can land at other stations to deliver or pick up cargo for transportation; it is possible that the destination station is also used by the UAV as a recharging station.

Below we consider a special case when the decision on the interaction between the station and the UAV is made by the latter (this corresponds to the second mode of operation of the proposed method), thereby adjusting their routes between the given destinations in order to minimize their missions. Such an adjustment becomes possible after some time is required to determine an adequate set of weights (4) in relation to each station by each UAV based on the tasks it performs.

The decentralized management discussed above is of particular relevance for energy charging stations based on renewable energy sources. It is advisable to install such stations, firstly, in hard-to-reach places where centralized power supply is difficult or economically inefficient; secondly, in places where the stay of a human operator for servicing stations / mobile objects, for one reason or another, should be reduced or excluded. In both of these cases, the lack of centralized power generation and the requirement to reduce operator participation dictates the need to use the decentralized control method declared above, which implies, in particular, optimization of the power generation process by stations equipped with MP and UAV operating modes.

Each charging module contains: 1) a carrier platform, which can be stationary or mobile, and on which the following equipment is installed; 2) an integrated power plant, consisting, in the general case, of a wind power plant and photoconverting elements, the joint operation of which is regulated by a single control unit; 3) UAV battery charge/change module or wireless charging module; 4) a module for analyzing the energy state and predicting the operation of this MP; 5) a transceiver module for the exchange of information signals with individual PRS of the group.

Each station receives recharging request signals from UAVs in its vicinity and analyzes the possibility of implementing each such recharging, based on its current energy reserve, and also, possibly based on a forecast of its power generation in the near future. In response, the station sends to each UAV that sent it a request, a response signal, which transmits information about the energy that can be available for this UAV at this station at the current time. Next, the on-board computer of this UAV selects the most suitable station with the number q=q _opt (s) and sends a signal to this station about the intention to land on it. The station with the number q _opt {s) after receiving this signal, reserves the corresponding part of its energy for charging the nominal given UAV. And the exchange of information of this station with all other UAVs that can turn later will be made taking into account this reserve, as well as taking into account information about all previously recorded reserves.

здесь состоит из одного элемента:

- реальной энергии, доступной для данного БПЛА на q-й станции в момент ее достижения, с учетом обращений к этой станции других БПЛА. Величины X_pq(s) и

актуальны только при выборе пары в качестве подлежащей взаимодействию.The components of the vector X _pq (s) are: _En,i (s) - the need for the p-th UAV in charge energy, taking into account the achievement of the q-th station; T _pq (s) is the real time that the UAV will spend on reaching the nearest target route point, taking into account the preliminary reaching the q-th station and recharging at it. Vector

here consists of one element:

- real energy available for the given UAV at the q-th station at the moment of its reaching, taking into account calls to this station by other UAVs. Values X _pq (s) and

relevant only when a pair is selected as the subject of interaction.

After the decision is made to select a certain charging station, an appropriate signal is generated, which is transmitted to the selected station using the transceiver module of this RRS.

Functional Ф _АВ (р,q,s _p ) depends on A, δ according to (2) as follows:

заданы выражениями:where the partial functionals

given by expressions:

;

- максимально возможная энергия, которая может быть выделена одному БПЛА подзарядки. Вектор весовых коэффициентов δ=[δ_Е, δ_Т] обновляется по следующему правилу:

s is the serial number of the landing made by this r-th UAV; Δ _T - upper bound estimate

;

- the maximum possible energy that can be allocated to one recharging UAV. The weight vector δ=[δ _E , δ _T ] is updated according to the following rule:

The weight coefficients δ _{E, pq} characterize the average degree of demand for the free energy accumulated at the q-th station for the p-th UAV, taking into account all previous (up to s-step) experiences of receiving / not receiving / not receiving energy by this UAV at this station; The weight coefficients δ _{T, pq} characterize the comparative average time costs that accompanied the recharging of the p-UAV at the q-th station in the past up to the s-step.

The coefficients δ _E,pq (s), δ _T,pq (s) for all p,q,s satisfy the normalization conditions:

As a result of maximizing the functional (5), we obtain a station with the number q _opt (s), which is chosen as the object for the nearest interaction of the UAV with the number p.

The efficiency of the adaptive selection of a suitable recharging station according to the proposed algorithm, specified by expressions (5)-(12), is increased each time by analyzing the information accumulated by the current time about the recharging experiences of a given UAV. This information is accumulated in the quantities η _pq (s) and ξ _pq (s).

; 3) непредвиденными изменениями на пути следования БПЛА до этой станции, которые могут происходить с самим БПЛА, в результате чего может резко повышаться величина E_n,pq(s), например, в результате спонтанных ветровых порывов, ухудшений условий видимости, появления случайных препятствий; 4) неточностью расчета значения энергии, которую станция потенциально может предоставить данному БПЛА; 5) помехой в канале связи, искажающей принимаемую на борту БПЛА информацию о величине

; 6) прочими причинами.The decrease in the weight δ _E,pq (s) for the q-station compared to its initial value in (8) can be caused by the following factors: 1) the rare use of the station by the p-th UAV due to the peculiarities of the regular trajectories it uses; 2) periodic and rather abrupt changes in the operating conditions of the station, leading to a significant decrease in real energy E _pq (s) in relation to the corresponding declared

; 3) unforeseen changes on the route of the UAV to this station, which can occur with the UAV itself, as a result of which the value of E _n,pq (s) can increase sharply, for example, as a result of spontaneous wind gusts, deterioration in visibility conditions, the appearance of random obstacles; 4) inaccuracy in calculating the value of energy that the station can potentially provide to this UAV; 5) interference in the communication channel, distorting the information received on board the UAV about the value

; 6) other reasons.

The decrease in weight δ _T,pq (s) is due to factors 1) and 3) from the list presented above, and is also determined by the peculiarity of the relative position of the area where UAV routes pass in relation to the location of the station.

The computing unit of each UAV learns to choose the most appropriate, in relation to some criterion, station given the territory of operation, the location of individual stations and the regular trajectories of the UAV. If these conditions are forced to change, training must be done anew. At the same time, the experience of the training carried out makes it possible to correct these conditions, if it is possible and / or appropriate, of course, with the involvement of forces external to the ROS, including the operator.

The claimed method enables individual UAVs to exchange accumulated experience, decentralized or with the help of periodic operator connection, which is advisable if they perform the same tasks or move on average along the same regular trajectories.

The decentralized adaptive selection of pairs of interacting objects considered here is especially effective for energy distributed systems, including charging stations for autonomous mobile objects, including UAVs, based on renewable wind and solar energy sources.

The effectiveness of the method should grow with an increase in the number of elements of the system and an increase in the number of interactions carried out by each given element of the system by the current moment, i.e. training time.

Technical and economic advantages of the proposed method over the known ones. The technical and economic advantage of the proposed method in relation to the existing centralized methods for coordinating the actions of the elements of the network under consideration, as well as to possible decentralized methods without self-learning, is to minimize the time for solving the tasks assigned to the RRS group, as well as the failure of the RRS during the mission, which is achieved by optimizing the choice of a pair of interacting stations and the RRS in each specific case based on both the current characteristics of all elements of the system and the history of all previous interactions for each of the elements.

One of the most important cases of this interaction is automatic recharging / refueling, including UAVs, both in stationary and non-stationary conditions, while a group of ORS of a given size can operate completely autonomously for a long time in a certain geographical area. At the same time, the location of individual autonomous charging stations in this territory depends on the shape, size, area of the latter, and the given/possible modes of operation of the PRS group on it.

This solution, being applied to an energy distributed network, including a group of mobile robotic objects and their charging stations, unlike the existing ones, allows to significantly extend the battery life of the entire group of RRS in a certain area. In the case when, for example, one-time operation of a group of UAVs is expected, for example, for a single survey of a non-extended infrastructure object, then not the entire network of recharging stations can be used, but only one or, at most, two such stations; at the same time, the time spent on the recharging process is reduced by at least 5 times compared to manually recharging one UAV by one operator, and at least 10 times compared to manually recharging two or more UAVs by one operator. With the work of a group of PRSs, numbering dozens of elements, the time spent on recharging can be reduced by at least two orders of magnitude; at the same time, there is a significant saving in the use of routine human labor.

The property of decentralized management of the interaction between the PRS and the docking stations, which is essential for the proposed method, makes it, in contrast to the existing ones, especially effective in those conditions when the operator's work is impossible, unsafe or ineffective: mountainous terrain, water surface, difficult areas of forests, etc. P.; places of radiation contamination, fires, northern regions, etc.; rural areas; in conditions of dense urban development, which makes the use of conventional transport ineffective.

The positive effect of the application of the proposed method may be associated with the functioning in urban or rural conditions, where the manual type of recharging, although it can be used, is much less effective than the automatic recharging system. Moreover, the gap in the efficiency of these two types of recharging will grow strongly, firstly, with an increase in the number of PRSs to be serviced, secondly, with an increase in the area of the territory that must be covered by this group of PRSs in the course of its operation, and, thirdly , - with an increase in the time of operation of the specified group in the given territory due to the effect of self-learning in decision-making.

Claims (2)

A method for decentralized control of a distributed generalized network of mobile robotic vehicles and docking stations, which consists in the fact that decentralized interaction between each pair is carried out, formed by a mobile robotic vehicle that performs autonomous or remotely controlled movement between destinations, including the stages of docking and landing on specially equipped stations sites , the initial routes of each mobile robotic vehicle, as well as the moments when the network elements require a significant adjustment of their operating modes, and the docking station, implemented using a module for receiving and transmitting information signals, a computing module, sensors for measuring their own values and the values of the external environment, with which each element of the network is equipped, as well as navigation modules, autopilots, payload modules, which are additionally equipped with mobile robotic vehicles, characterized in that the repair service or their failure is determined by the operator or automated command link based on the purpose and objectives of the functioning of the entire system, and the decision on the potential interaction between each such pair is based on the choices made by both of its elements or one of them, with the possibility of using the accumulated information about previous interactions to optimize the process of subsequent interactions of mobile robotic vehicles and docking stations according to the specified criteria for the functioning of the network, in which forced connection of an operator or an automated command link to the control loop is allowed.

Images