A multi-agent system distributed optimal resource allocation method and system

Abstract

This application discloses a distributed optimal resource allocation method and system for multi-agent systems, relating to the field of distributed resource allocation control technology for multi-agent systems. The method includes real-time acquisition of actual system requirements, communication network topology data, status data, and resource data to be allocated from each physical intelligent device in the cluster control system; construction of a multi-agent resource allocation mathematical model; solving the multi-agent resource allocation mathematical model using a distributed extreme value search optimization method based on an anti-saturation compensation mechanism and a dynamic event triggering mechanism to determine the optimal resource allocation scheme and generate control commands; and sending the control commands to the cluster control system to control the cluster control system to allocate resources to each physical intelligent device, thus completing the distributed optimal resource allocation of the cluster control system. This application can achieve distributed optimal resource allocation, improving the performance and communication resource utilization of multi-agent systems.

Description

Technical Field

[0001] This application relates to the field of distributed resource allocation and control technology for multi-agent systems, and in particular to a method and system for distributed optimal resource allocation in multi-agent systems. Background Technology

[0002] Distributed optimal resource allocation is one of the core research directions of multi-agent systems. It achieves the global optimal allocation of limited resources through local information interaction among multiple physical intelligent devices and is widely used in engineering fields such as distributed energy scheduling, multi-robot collaboration, and communication network resource management.

[0003] Existing distributed optimization methods face several technical bottlenecks in practical engineering applications: First, most existing distributed optimization methods assume that the analytical expression of the local cost function and gradient information of the physical intelligent device are known. However, in real-world scenarios, the cost function is often affected by environmental and operating conditions, exhibiting unknown characteristics. The physical intelligent device can only obtain real-time measurement values, making it impossible to directly apply distributed optimization methods. Second, the actuators of the physical intelligent device have physical saturation constraints, which can lead to integral saturation, disrupting the convergence of the algorithm and causing response delays and decreased steady-state accuracy in multi-agent systems. Most existing methods ignore this nonlinear constraint or only perform simple truncation, failing to effectively eliminate the impact of integral saturation and thus reducing the performance of multi-agent systems. Third, existing methods are based on the assumption of ideal continuous communication. However, communication resources in real-world multi-agent systems are limited, and continuous information interaction can lead to a huge communication burden and even network congestion. The triggering conditions of some event-triggered algorithms are fixed, making it impossible to adaptively adjust the information interaction frequency according to the system state, resulting in low utilization of communication resources.

[0004] To address the aforementioned issues, there is an urgent need to design a distributed optimal resource allocation method for multi-agent systems, in order to achieve distributed optimal resource allocation and improve the performance and communication resource utilization of multi-agent systems. Summary of the Invention

[0005] The purpose of this application is to provide a distributed optimal resource allocation method and system for multi-agent systems, which can realize distributed optimal resource allocation and improve the performance and communication resource utilization of multi-agent systems.

[0006] To achieve the above objectives, this application provides the following solution.

[0007] In a first aspect, this application provides a distributed optimal resource allocation method for a multi-agent system, applied to a distributed resource allocation control scenario of a cluster control system including multiple physical intelligent devices. The distributed optimal resource allocation method for a multi-agent system includes the following steps.

[0008] The actual system requirements, communication network topology data, status data, and resource data to be allocated for each physical intelligent device in the cluster control system are collected in real time.

[0009] Based on the acquired basic attribute data and the actual system requirements, a multi-agent resource allocation mathematical model is constructed. The multi-agent resource allocation mathematical model is used to represent a distributed resource allocation problem with an unknown cost function and subject to actuator saturation constraints. In the distributed resource allocation problem, each physical intelligent device can obtain a real-time measurement value of the local cost function, and the control input of the physical intelligent device is subject to nonlinear truncation constraints due to actuator saturation.

[0010] Based on the communication network topology data, the resource data to be allocated, and the state data, a distributed extreme value search optimization method based on anti-saturation compensation mechanism and dynamic event triggering mechanism is used to solve the multi-agent resource allocation mathematical model, determine the optimal resource allocation scheme, and generate control commands.

[0011] The control command is sent to the cluster control system to control the cluster control system to allocate resources to each of the physical intelligent devices, thereby completing the distributed optimal resource allocation of the cluster control system.

[0012] Optionally, the basic attribute data includes the number of physical intelligent devices, the unique identifier of each physical intelligent device, and physical characteristic parameters; the communication network topology data includes the communication link relationship between each physical intelligent device, communication bandwidth, and communication latency; and the resource data to be allocated includes the total amount of resources to be allocated.

[0013] Optionally, the mathematical model for multi-agent resource allocation is expressed as follows.

[0014] .

[0015] .

[0016] in, This represents the system's global optimization goal. For the global cost function, For the first Local cost function of a physical intelligent device The number of physical smart devices. This represents a global equality constraint. For the first The output status of each physical smart device, and , Represents an n-dimensional real vector space. For the first The actual system requirements of a physical intelligent device.

[0017] Optionally, based on the communication network topology data, the resource data to be allocated, and the state data, a distributed extreme value search optimization method based on an anti-saturation compensation mechanism and a dynamic event triggering mechanism is used to solve the multi-agent resource allocation mathematical model, determine the optimal resource allocation scheme, and generate control instructions, specifically including the following steps.

[0018] Based on the communication network topology data, the state data, and the resource data to be allocated, and combined with the local cost function characteristics of the multi-agent resource allocation mathematical model and the nonlinear truncation constraint boundary of actuator saturation, the range of values ​​for the parameters to be tuned is determined; the parameters to be tuned include the small positive parameters of the extreme value search control strategy, the threshold parameters of the dynamic event triggering mechanism, and the coefficients of the anti-saturation compensation term.

[0019] The tuned parameters are determined based on the range of values ​​for the parameters to be tuned.

[0020] The tuned parameters are substituted into the expression of the distributed extreme value search optimization method, and the output state of each physical intelligent device is used as the decision variable. The decision variables of each physical intelligent device are iteratively updated until they converge to the optimal solution of the multi-agent resource allocation mathematical model, thus obtaining the converged optimal decision variables of each physical intelligent device. The distributed extreme value search optimization method estimates the gradient information of the unknown local cost function based on the real-time measurement value of the local cost function of each physical intelligent device using an extreme value search control strategy, and uses an anti-saturation compensation term to eliminate the integral saturation phenomenon caused by actuator saturation. The dynamic event triggering mechanism determines the information exchange triggering condition based on the state estimation error of the physical intelligent device, and adaptively adjusts the information exchange frequency between adjacent physical intelligent devices based on the information exchange triggering condition to control the on-demand information interaction between the physical intelligent devices.

[0021] The optimal resource allocation scheme is determined based on the convergent optimal decision variables of each of the aforementioned physical intelligent devices.

[0022] Based on the optimal resource allocation scheme, corresponding control commands are generated.

[0023] Optionally, the communication network topology data is undirected strongly connected topology data, which includes a set of nodes, a set of edges, and topology matrix information characterizing the network topology; wherein, each node in the set of nodes represents a physical intelligent device, and each edge in the set of edges represents a communication connection between two physical intelligent devices.

[0024] Optionally, the expression for the distributed extreme value search optimization method is as follows.

[0025] .

[0026] in, Indicates the first Real-time rate of change of decision variables for each physical intelligent device For the saturation function of the actuator, For the first Input from a physical smart device for The first derivative with respect to time, All are adjustable gain. For the first Anti-saturation compensation term for each physical intelligent device For time, For the adjacent first The first physical smart device and the first Communication weights between physical intelligent devices Indicates the first Lagrange multiplier signals transmitted by a physical intelligent device For the small positive parameter of the extreme value search control strategy, For the first Real-time measurement of the local cost function of a physical intelligent device For the first Lagrange multipliers corresponding to the coupling constraints of each physical intelligent device for The first derivative with respect to time, For the first The accompanying variables of a physical intelligent device for The first derivative with respect to time.

[0027] The expression for the dynamic event triggering mechanism is as follows.

[0028] .

[0029] in, All are adjustable gain. In the dynamic event triggering mechanism, the first A physical smart device in The internal state variables at time t, for The first derivative with respect to time, For the first A physical smart device in The state estimation error at time t, and , This indicates the time of the kth trigger.

[0030] Optionally, the distributed extreme value search optimization method adopts a dual-time-scale dynamic adjustment structure.

[0031] The dual-timescale dynamic adjustment structure specifically includes the following:

[0032] The dynamic consistency estimator is adjusted based on the first time scale to achieve rapid estimation of global information.

[0033] The output state of the physical intelligent device is adjusted based on a second time scale to achieve convergence of the decision variables of the physical intelligent device to the optimal solution, and the convergence speed is affected by the cost function and constraints.

[0034] The adjustment rate of the first time scale is greater than the adjustment rate of the second time scale.

[0035] Optionally, after substituting the tuned parameters into the expression of the distributed extreme value search optimization method, using the output state of each physical intelligent device as a decision variable, iteratively updating the decision variables of each physical intelligent device until the decision variables of each physical intelligent device converge to the optimal solution of the multi-agent resource allocation mathematical model, and obtaining the converged optimal decision variables of each physical intelligent device, the multi-agent system distributed optimal resource allocation method further includes the following steps.

[0036] Based on singular perturbation theory, a convergence analysis is performed on the dual-timescale dynamic adjustment structure to verify the effectiveness of the distributed extreme value search optimization method in converging to the optimal solution of the distributed resource allocation problem.

[0037] Optionally, the distributed extreme value search optimization method satisfies the following assumptions: the local cost function of each physical intelligent device is quadratically continuous and differentiable and uniformly strong convex with respect to the decision variables, and the gradient of the local cost function satisfies the Lipschitz continuity property; the distributed resource allocation problem has an optimal solution.

[0038] Secondly, this application provides a distributed optimal resource allocation system for a multi-agent system, applied to a distributed resource allocation control scenario of a cluster control system including multiple physical intelligent devices. The distributed optimal resource allocation system for a multi-agent system is used to execute the distributed optimal resource allocation method for a multi-agent system as described in any of the first aspects.

[0039] According to the specific embodiments provided in this application, this application has the following technical effects.

[0040] This application provides a distributed optimal resource allocation method and system for multi-agent systems. The method defines a distributed resource allocation problem with an unknown cost function and actuator saturation constraints by constructing a mathematical model for multi-agent resource allocation. It clarifies that each physical intelligent device can only obtain real-time measurements of its local cost function, and that the control input of the physical intelligent device is subject to nonlinear truncation constraints due to actuator saturation. This restores the real scenario of an unknown cost function and actuator saturation constraints, making it more practical and effective. It solves the problem that existing distributed optimization methods generally assume that the analytical expression of the local cost function and gradient information of the physical intelligent device are known, which leads to their inapplicability in practice. Furthermore, this application employs a distributed extreme value search optimization method based on an anti-saturation compensation mechanism and a dynamic event triggering mechanism. Combining the anti-saturation compensation mechanism with the dynamic event triggering mechanism effectively eliminates integral saturation caused by actuator saturation, thus ensuring the dynamic performance and steady-state accuracy of the multi-agent system under nonlinear input constraints. This solves the problem that most existing methods ignore nonlinear constraints or only perform simple truncation, failing to effectively eliminate integral saturation and leading to response delays and decreased steady-state accuracy in multi-agent systems, thereby significantly improving the performance of multi-agent systems. On the other hand, the dynamic event triggering mechanism enables on-demand information interaction between physical intelligent devices, solving the problem that existing methods cannot adaptively adjust the information interaction frequency according to the system state, resulting in low communication resource utilization, and thus improving communication resource utilization. As can be seen, this application, by combining multiple technologies such as the construction of a multi-agent resource allocation mathematical model and a distributed extreme value search optimization method based on anti-saturation compensation mechanism and dynamic event triggering mechanism, can achieve global optimal resource allocation under unknown cost function and input saturation constraints, and obtain a more accurate, reliable and reasonable resource allocation scheme, thereby improving the performance and communication resource utilization of the multi-agent system. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 An application environment diagram of a distributed optimal resource allocation method for a multi-agent system provided in an embodiment of this application; Figure 2 A flowchart illustrating a distributed optimal resource allocation method for a multi-agent system provided in an embodiment of this application; Figure 3This is a schematic diagram of the design process of a distributed optimal resource allocation method for a multi-agent system provided in an embodiment of this application; Figure 4 A schematic diagram illustrating the principle of a distributed extreme value search optimization method provided in an embodiment of this application; Figure 5 A communication diagram illustrating a multi-agent system using four agents is provided in one embodiment of this application. Figure 6 This is a schematic diagram of the structure of four intelligent agents provided in one embodiment of this application; Figure 7 A schematic diagram illustrating the triggering time of a dynamic event-triggered communication mechanism provided in an embodiment of this application; Figure 7 (a) in the diagram is a schematic diagram of the triggering time of the dynamic event-triggered communication mechanism of agent 1; Figure 7 (b) in the diagram is a schematic diagram of the triggering time of the dynamic event-triggered communication mechanism of agent 2; Figure 7 (c) in the diagram is a schematic diagram of the triggering time of the dynamic event-triggered communication mechanism of agent 3; Figure 7 (d) in the diagram represents the triggering moment of the dynamic event-triggered communication mechanism of agent 4; Figure 8 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0043] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0044] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0045] The distributed optimal resource allocation method for multi-agent systems provided in this application can be applied to, for example... Figure 1In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be set up independently, integrated into server 104, or placed in the cloud or on another server. Terminal 102 can send the actual system requirements, communication network topology data, status data, and resource allocation data of each physical intelligent device in the cluster control system to server 104. After receiving these data, server 104 constructs a multi-agent resource allocation mathematical model; it then uses a distributed extreme value search optimization method based on an anti-saturation compensation mechanism and a dynamic event triggering mechanism to solve the multi-agent resource allocation mathematical model and determine the optimal resource allocation scheme. Server 104 can then feed back the obtained optimal resource allocation scheme to terminal 102. Furthermore, in some embodiments, the distributed optimal resource allocation method for multi-agent systems can also be implemented separately by the server 104 or the terminal 102. For example, the terminal 102 can directly perform resource allocation processing based on the actual system requirements, communication network topology data, status data, and resource data to be allocated for each physical intelligent device in the cluster control system. Alternatively, the server 104 can obtain the actual system requirements, communication network topology data, status data, and resource data to be allocated for each physical intelligent device in the cluster control system from the data storage system, and perform resource allocation processing based on the actual system requirements, communication network topology data, status data, and resource data to be allocated for each physical intelligent device in the cluster control system.

[0046] The terminal 102 can be, but is not limited to, various desktop computers, laptops, smartphones, tablets, and IoT devices. The server 104 can be implemented using a standalone server or a server cluster consisting of multiple servers, or it can be a cloud server.

[0047] In one exemplary embodiment, such as Figure 2 As shown, a distributed optimal resource allocation method for a multi-agent system is provided. This method is executed by a computer device, specifically by a terminal or server alone, or by both a terminal and a server. In this embodiment, the method is applied to... Figure 1 Taking server 104 as an example, the following steps are included.

[0048] S1: Real-time acquisition of the actual system requirements, communication network topology data, status data, and resource data to be allocated for each physical intelligent device in the cluster control system.

[0049] S2: Based on the acquired basic attribute data and the actual system requirements, a multi-agent resource allocation mathematical model is constructed. This model represents a distributed resource allocation problem with an unknown cost function and actuator saturation constraints. In this distributed resource allocation problem, each physical intelligent device can acquire real-time measurements of its local cost function, and the control input of each physical intelligent device is subject to nonlinear truncation constraints due to actuator saturation.

[0050] S3: Based on the communication network topology data, the resource data to be allocated, and the state data, a distributed extreme value search optimization method based on anti-saturation compensation mechanism and dynamic event triggering mechanism is used to solve the multi-agent resource allocation mathematical model, determine the optimal resource allocation scheme, and generate control instructions.

[0051] S4: Send the control command to the cluster control system to control the cluster control system to allocate resources to each of the physical intelligent devices, thereby completing the distributed optimal resource allocation of the cluster control system.

[0052] By implementing steps S1 to S4 above, and combining various technologies such as the construction of a multi-agent resource allocation mathematical model, a distributed extreme value search optimization method based on anti-saturation compensation mechanism and dynamic event triggering mechanism, it is possible to achieve global optimal resource allocation under unknown cost function and input saturation constraints, thereby obtaining a more accurate, reliable and reasonable resource allocation scheme, which can improve the performance of multi-agent systems and the utilization rate of communication resources.

[0053] As an optional implementation, this distributed optimal resource allocation method for multi-agent systems can be applied to distributed resource allocation control scenarios in various types of multi-agent systems, where a multi-agent system refers to a cluster control system comprising multiple physical intelligent devices. The basic attribute data includes the number of physical intelligent devices, the unique identifier of each physical intelligent device, and its physical characteristic parameters. The communication network topology data is obtained by constructing a communication network between the physical intelligent devices, and includes the communication link relationships, communication bandwidth, and communication latency between the physical intelligent devices. The resource data to be allocated includes the total amount of resources to be allocated.

[0054] As an optional implementation, in step S3, based on the communication network topology data, the resource data to be allocated, and the state data, a distributed extreme value search optimization method based on anti-saturation compensation mechanism and dynamic event triggering mechanism is used to solve the multi-agent resource allocation mathematical model, determine the optimal resource allocation scheme, and generate control instructions, specifically including the following steps.

[0055] S31: Based on the communication network topology data, the state data, and the resource data to be allocated, and in conjunction with the local cost function characteristics of the multi-agent resource allocation mathematical model and the nonlinear truncation constraint boundary of actuator saturation, determine the range of values ​​for the parameters to be tuned; the parameters to be tuned include the small positive parameters of the extreme value search control (ESC) strategy, the threshold parameters of the dynamic event triggering mechanism, and the coefficients of the anti-saturation compensation term.

[0056] S32: Determine the tuned parameters based on the range of values ​​of the parameters to be tuned.

[0057] S33: Substitute the tuned parameters into the expression of the distributed extreme value search optimization method, use the output state of each physical intelligent device as the decision variable, and iteratively update the decision variables of each physical intelligent device until the decision variables of each physical intelligent device converge to the optimal solution of the multi-agent resource allocation mathematical model, thus obtaining the converged optimal decision variables of each physical intelligent device. Specifically, the distributed extreme value search optimization method estimates the gradient information of the unknown local cost function based on the real-time measured value of the local cost function of each physical intelligent device using an extreme value search control strategy, and uses an anti-saturation compensation term to eliminate the integral saturation phenomenon caused by actuator saturation; the dynamic event triggering mechanism determines the information exchange triggering condition based on the state estimation error of the physical intelligent device, and adaptively adjusts the information exchange frequency between adjacent physical intelligent devices based on the information exchange triggering condition to control the on-demand information interaction between the physical intelligent devices.

[0058] The parameter tuning process is essentially an optimization of the algorithm's performance. Within the range of values ​​for the parameters to be tuned, different values ​​will lead to differences in the algorithm's search path and convergence accuracy, which can be seen as generating multiple potential allocation schemes. Substituting the tuned parameters into the expression of the distributed extreme value search optimization method involves comparing the global cost value ultimately achieved by the system under different parameter configurations to determine the specific parameter values ​​that minimize cost and maximize stability (especially when dealing with executor saturation constraints). The final output resource allocation scheme is the convergent solution calculated under the drive of this set of optimal parameters.

[0059] The decision variables of physical intelligent devices refer to the state information of the physical intelligent devices themselves. Their specific physical meaning varies from scenario to scenario. For example, in the power system scenario, it usually refers to the active power output value; in the robot collaboration scenario, it can refer to the relative position coordinates with respect to the center of the robot group, the shared load, or the moving speed.

[0060] The real-time measured value of the local cost function of a physical intelligent device refers to the data directly measured and read by the device's built-in measurement modules (such as ammeters, flow meters, and fuel meters) during task execution. Its significance lies in the fact that the algorithm does not need to know complex cost mathematical formulas in advance; instead, it directly uses the measured real-time physical quantities as feedback, continuously adjusts through extreme value search, and ultimately automatically finds the lowest-cost and most energy-efficient resource allocation scheme.

[0061] S34: Determine the optimal resource allocation scheme based on the convergent optimal decision variables of each of the physical intelligent devices.

[0062] S35: Generate corresponding control instructions based on the optimal resource allocation scheme.

[0063] As an optional implementation, in step S31, the communication network topology data is undirected strongly connected topology data, which includes a set of nodes, a set of edges, and topology matrix information characterizing the network topology; wherein, each node in the set of nodes represents a physical intelligent device, and each edge in the set of edges represents a communication connection between two physical intelligent devices.

[0064] As an optional implementation, in step S3, the distributed extreme value search optimization method adopts a dual-time-scale dynamic adjustment structure.

[0065] The dual-timescale dynamic adjustment structure specifically includes the following:

[0066] (1) Adjust the dynamic consistency estimator based on the first time scale to achieve fast estimation of global information.

[0067] (2) Adjust the output state of the physical intelligent device based on the second time scale so as to realize that the decision variables of the physical intelligent device converge to the optimal solution, and the convergence speed is affected by the cost function and the constraint conditions.

[0068] The adjustment rate of the first time scale is greater than the adjustment rate of the second time scale.

[0069] The first time scale drives the dynamic consensus estimator, whose adjustment gain is configured to pre-calculate the convergence estimation of global information of the cluster before the output of the physical intelligent device deviates significantly. The second time scale drives the evolution of the decision variables of the physical intelligent device, which smooths the global information obtained by the first time scale through an integral term to guide each physical intelligent device to converge toward the optimal resource allocation scheme. In the multi-agent system of this application embodiment, the dynamic consensus estimator is an auxiliary algorithm structure deployed locally on each physical intelligent device. Its core function is to adjust based on the first time scale in a decentralized architecture to achieve rapid estimation of global information. It enables physical intelligent devices with only local observation capabilities to indirectly obtain the macroscopic state of the entire cluster control system through network communication with neighboring nodes.

[0070] As an optional implementation, after step S34, the multi-agent system distributed optimal resource allocation method further includes the following steps.

[0071] Based on singular perturbation theory, a convergence analysis is performed on the dual-timescale dynamic adjustment structure to verify the effectiveness of the distributed extreme value search optimization method in converging to the optimal solution of the distributed resource allocation problem.

[0072] The convergence analysis employs singular perturbation theory, decoupling the system into a boundary layer subsystem at the first time scale (responsible for consensus estimation) and a simplified layer subsystem at the second time scale (responsible for optimal solution finding). Lyapunov stability analysis proves the exponential convergence of the boundary layer system to global information at the first time scale, and averaging theory ensures the asymptotic convergence of the simplified layer system to the global optimum at the second time scale. This guarantees the steady-state accuracy and dynamic performance of the overall algorithm under dual-scale adjustment.

[0073] As an optional implementation, in step S3, the distributed extreme value search optimization method satisfies the following assumptions: the local cost function of each physical intelligent device is quadratically continuous and differentiable and uniformly strong convex with respect to the decision variables, and the gradient of the local cost function satisfies the Lipschitz continuity property; the distributed resource allocation problem has an optimal solution.

[0074] The distributed optimal resource allocation method for multi-agent systems proposed in this embodiment can be widely applied to various scenarios such as UAV swarms, smart microgrids, forestry data acquisition, and water conservancy natural disasters. It can achieve accurate, reliable, and reasonable distributed resource allocation for the corresponding swarm control systems. The implementation process of the technical solution in this embodiment will be illustrated below using UAV swarm scenarios and smart microgrid scenarios as examples.

[0075] For drone swarm scenarios, the main issue involves resource allocation for inspection tasks. In this scenario, the drone swarm control system acts as the swarm control system, and each drone is a physical intelligent device. Resources to be allocated include inspection area, task load, and energy consumption quota. The core problem in this scenario is the saturation constraint of drone motors / actuators and the unknown cost function of each drone (energy consumption characteristics vary with the environment), requiring the achievement of distributed optimal task allocation. In practical applications, firstly, each drone uses its own sensors, onboard computer, and communication module to collect real-time data on actual system requirements, communication topology, status, and resources to be allocated. Actual system requirements include total inspection area, total task duration, and total energy consumption limit. Communication topology data includes inter-drone communication links, communication bandwidth, and latency. Status data includes location, remaining battery power, current energy consumption, speed, and task progress. Resource data to be allocated includes the total inspection area and total task load to be partitioned. Furthermore, known basic attribute data includes the number of drones, ID, maximum thrust, motor saturation power, maximum range, and energy consumption coefficient. The cost function is the energy cost corresponding to the i-th drone performing an inspection task for a certain inspection area. Then, based on the acquired basic attribute data and actual system requirements, a multi-agent resource allocation mathematical model is constructed. Next, an undirected strongly connected topology is constructed based on the communication range between drones, with nodes representing drones and edges representing communication links between them. Then, based on the communication network topology data, the data of resources to be allocated, and the state data of each drone, a distributed extreme value search optimization method based on anti-saturation compensation and dynamic event triggering mechanisms is used to solve the multi-agent resource allocation mathematical model. During the solution process, when the drone motor output reaches saturation, the anti-saturation compensation term is activated to eliminate integral saturation. The dynamic event triggering mechanism is used to determine when to communicate based on the state estimation error, reducing energy consumption. When all drones converge to the global optimal solution, satisfying the global equality constraints in the expression of the multi-agent resource allocation mathematical model, the optimal resource allocation scheme is generated, and control commands are generated. In this optimal resource allocation scheme, drone 1 is allocated inspection area A1, and drone 2 is allocated inspection area A2. Control commands are sent from the backend control center to the drone swarm control system, controlling each drone in the drone swarm control system to execute its corresponding inspection task at the lowest cost.

[0076] For smart microgrid scenarios, the main focus is on power resource allocation. In this scenario, the smart microgrid consists of power generation equipment such as photovoltaic (PV) devices, wind power devices, and energy storage devices, as well as controllable load devices. These devices are physically intelligent (PMI) devices, and the resources to be allocated are power quotas. The core objective is to minimize the total power generation cost while satisfying grid power balance constraints, and there is a saturation limit for device power output. In practical applications, firstly, the aforementioned devices collect real-time data on actual system demand, communication topology, status data, and resources to be allocated through their own monitoring modules. Actual system demand includes total load power and total grid-connected power requirements. Communication topology includes communication links, bandwidth, and latency between devices. Status data includes real-time PV / wind power output, energy storage SOC, and load demand changes. Resources to be allocated include the total power generation quota. Known basic attribute data includes the number of devices, ID, maximum power generation, saturation power, and cost coefficient. The cost function is the power generation cost of the i-th power generation device, and the actual system demand is the total system load demand. Then, an undirected strongly connected communication network is constructed to represent the internal communication relationships of microgrid devices such as photovoltaic equipment, wind power equipment, energy storage equipment, and controllable load equipment. Next, based on the acquired basic attribute data and actual system requirements, a multi-agent resource allocation mathematical model is constructed. Then, based on the communication network topology data, resource data to be allocated, and state data, a distributed extreme value search optimization method based on anti-saturation compensation and dynamic event triggering mechanisms is used to solve the multi-agent resource allocation mathematical model, generating an optimal resource allocation scheme and control commands. This optimal resource allocation scheme is the optimal power scheme, specifying the output power of photovoltaic equipment, wind power equipment, and energy storage equipment. Control commands are sent from the backend control center to the intelligent microgrid cluster control system, controlling each power generation device in the intelligent microgrid cluster control system to execute its corresponding power generation task at the lowest cost.

[0077] In an exemplary embodiment, to illustrate the technical solutions provided in the embodiments of this application in detail, the following are provided: Figure 3 This paper presents a distributed optimal resource allocation method for multi-agent systems. It aims to combine extreme value search control strategies, anti-integral saturation mechanisms, and dynamic event triggering mechanisms to achieve globally optimal resource allocation under unknown cost functions and input saturation constraints. This results in a more accurate, reliable, and reasonable resource allocation scheme, thereby improving the performance and communication resource utilization of the multi-agent system. Figure 3 As shown, the design process of this method mainly includes the following steps: proposing a resource allocation problem, designing the connectivity of the communication network, designing a distributed extreme value search optimization method, designing a dynamic event-triggered communication scheme, designing the algorithm parameter range, and adjusting the parameters to complete the optimization. The specific steps are as follows.

[0078] S100, Problem Definition: Regarding... A cluster control system composed of physical intelligent devices is defined as a constrained optimization problem, which represents the global optimization objective of the system and is expressed as follows.

[0079] (1).

[0080] in, For the global cost function, For the first Local cost function of a physical intelligent device The number of physical smart devices.

[0081] Based on equation (1), global equality constraints need to be satisfied, as shown in the following equation.

[0082] (2).

[0083] in, For the first The output status of each physical smart device, and , Represents an n-dimensional real vector space. For the first The actual system requirements of a physical intelligent device.

[0084] The output state of a physical intelligent device is essentially its own state information, and its specific physical meaning varies depending on the scenario. For example, in a power system scenario, the output state of a physical intelligent device is usually the active power output value; in a robot collaboration scenario, the output state of a physical intelligent device is usually its relative position coordinates relative to the center of the cluster, its shared load, or its movement speed. The actual system demand corresponds to the total amount of resources that the entire system currently needs to allocate. In the mathematical model, the sum of the resources output by all physical intelligent devices must equal the total demand. For example, in a microgrid dispatching scenario, the actual system demand refers to the total power load of the entire network; in a task allocation scenario, the actual system demand refers to the total amount of tasks. The task of this algorithm is to ensure that, while meeting the total demand, the sharing ratio of each physical intelligent device is optimal (i.e., the lowest cost).

[0085] S200, Design the communication network: Design the connectivity of the communication network to ensure that the communication network between physical intelligent devices is an undirected connected graph structure with strong connectivity, i.e., an undirected strongly connected topology.

[0086] S300, Design Allocation Method: (e.g.) Figure 4 As shown in the figure, this application embodiment designs a distributed extreme value search optimization method based on anti-saturation compensation mechanism and dynamic event triggering mechanism, which is expressed as the following formula.

[0087] (3).

[0088] in, Indicates the first Real-time rate of change of decision variables for each physical intelligent device For the saturation function of the actuator, For the first Input from a physical smart device for The first derivative with respect to time, All are adjustable gain. For the first Anti-saturation compensation term for each physical intelligent device For time, For the adjacent first The first physical smart device and the first Communication weights between physical intelligent devices Indicates the first Lagrange multiplier signals transmitted by a physical intelligent device For the small positive parameter of the extreme value search control strategy, For the first Real-time measurement of the local cost function of a physical intelligent device For the first Lagrange multipliers corresponding to the coupling constraints of each physical intelligent device for The first derivative with respect to time, For the first The accompanying variables of each physical intelligent device are used to ensure the Lagrange multiplier consistency; for The first derivative with respect to time.

[0089] S400, add an event triggering mechanism: design a dynamic event triggering mechanism, expressed as the following formula.

[0090] (4).

[0091] in, All are adjustable gain. In the dynamic event triggering mechanism, the first A physical smart device in Internal state variables at any given time; for The first derivative with respect to time, For the first A physical smart device in The state estimation error at time t, and , This indicates the time of the k-th trigger, when the condition is met. Under certain conditions, the frequency of information exchange between neighbors is adaptively adjusted.

[0092] Among them, the internal state variable is the core of the dynamic event triggering mechanism. It represents the logical variable used to adjust the communication frequency. Its role is as a "buffer pool" for dynamic thresholds. Through the evolution of this variable, the system can adaptively decide when to exchange data.

[0093] S500, Determine the range of algorithm parameters: Design the range of algorithm parameters.

[0094] S600, Optimized Distribution: Substitute the parameters into the algorithm to minimize the error between the solution and the optimal solution, thus achieving optimal resource allocation.

[0095] In this embodiment of the application, the characteristic of the S100 resource allocation problem is that the physical intelligent device is unaware of... The specific analytical expression or gradient expression depends only on real-time function value measurements. and control input Due to physical limitations, it exhibits nonlinear truncation characteristics. .

[0096] In this embodiment of the application, the construction method of the S200 physical intelligent device communication network is as follows: Each physical intelligent device is defined to be able to communicate with its neighboring physical intelligent devices on an undirected strongly connected topology. For an undirected strongly connected topology, The set of nodes corresponding to each physical intelligent device is used The edge set corresponding to the communication between physical intelligent devices is denoted as... , recorded as In the figure Neighbors of physical smart devices It is a Laplace matrix.

[0097] In this embodiment of the application, the S300 method satisfies the following assumptions.

[0098] Assumption 1: For all physically intelligent devices, the local cost function It is capable of second-order continuous differentiation, is relatively uniform and strongly convex, and its gradient is Lipschitz continuous. Local inequality constraints apply. It is a continuously differentiable convex function.

[0099] Assumption 2: There exists an optimal solution to the resource allocation problem.

[0100] In this embodiment of the application, the design range of the S300 algorithm parameters should be determined by comprehensive consideration in order to ensure the effectiveness and convergence of the algorithm.

[0101] In this embodiment of the application, the S300-optimized distributed optimal resource allocation method with anti-integral saturation compensation mechanism has a dual-time-scale dynamic adjustment structure, specifically including the following:

[0102] (1) The dynamic consistency estimator is adjusted on a fast time scale to quickly estimate global information.

[0103] (2) The output of each physical smart device is adjusted on a relatively slow time scale, by adjusting It converges to the optimal solution, and its convergence speed is affected by the cost function and constraints.

[0104] In this embodiment, the S300 convergence analysis method is based on the following steps and principles.

[0105] make Thus, the boundary layer system is obtained, and its dynamic equation is as follows.

[0106] (5).

[0107] in, Indicates the input of a physical intelligent device. For adjustable gain, The amplitude of the cosine signal. These are the frequencies of the sine and cosine signals. The communication network between physical intelligent devices is an undirected, strongly connected topology.

[0108] When in a quasi-steady state, a simplified system is obtained, and its dynamic equation is as follows.

[0109] (6).

[0110] in, Represents the Laplace matrix, For the measurement error vector, This represents the law of variation of the decision estimates of physically intelligent devices.

[0111] The simplified system asymptotically converges to the optimal trajectory, i.e. .

[0112] For a positive existence, such that for any, the dynamic form of the first... The states of each physical intelligent device asymptotically converge to the optimal trajectory of the problem, that is, for any , we have .

[0113] This application example uses four intelligent agents (i.e., physical intelligent devices) to form a multi-agent system (i.e., a cluster control system), and its communication diagram is as follows. Figure 5As shown. The optimization objective is expressed as follows.

[0114] (6).

[0115] Then, the cost function for each agent is set as follows: , , , .

[0116] Then set the parameters to , , , , , , and .like Figure 6 As shown, the output state of each agent , All converge to the optimal solution of the resource allocation problem. The event triggering time of the dynamic event-triggered scheme is as follows: Figure 7 As shown, Figure 7 (a) in the diagram is a schematic diagram of the triggering time of the dynamic event-triggered communication mechanism of agent 1; Figure 7 (b) in the diagram is a schematic diagram of the triggering time of the dynamic event-triggered communication mechanism of agent 2; Figure 7 (c) in the diagram is a schematic diagram of the triggering time of the dynamic event-triggered communication mechanism of agent 3; Figure 7 (d) in the figure is a schematic diagram of the triggering time of the dynamic event-triggered communication mechanism of agent 4. It can be seen that the system has no Zeno behavior.

[0117] This application proposes a distributed optimal resource allocation method for a multi-agent system. The specific steps of this method are as follows: First, define the problem and construct a system containing... A network system of physical intelligent devices is proposed to solve the distributed resource allocation problem with unknown cost functions and actuator saturation constraints. In this problem, the physical intelligent devices can only measure local cost function values ​​and the control input is restricted by nonlinear truncation. The algorithm uses an extremum search control strategy to estimate gradient information and combines a classical anti-saturation compensation mechanism to eliminate integral saturation caused by actuator saturation. A novel distributed extremum search algorithm based on a dynamic event-triggered communication mechanism is proposed. The algorithm uses a state-dependent threshold function to adaptively adjust the information exchange frequency between neighbors. Analysis shows that the designed algorithm exhibits dual time-scale characteristics and is regarded as a singular perturbation system. The singular perturbation theory proves that the decision variables of all physical intelligent devices can converge asymptotically to the optimal solution in a semi-global reality. Finally, the effectiveness of the results is verified by an economic scheduling simulation example involving four distributed energy sources.

[0118] Compared with the prior art, the distributed optimal resource allocation method for multi-agent systems proposed in this application has the following advantages.

[0119] (1) Existing distributed optimization methods usually assume that the cost function is known or ignore the nonlinear effects caused by input saturation. The embodiments of this application use the extreme value search control strategy to estimate the gradient information without knowing the gradient expression, and introduce an anti-integral saturation mechanism to effectively eliminate the integral saturation phenomenon caused by actuator saturation, thereby ensuring the dynamic performance and steady-state accuracy of the system under nonlinear input constraints.

[0120] (2) The embodiments of this application design a dynamic event triggering mechanism and use singular perturbation theory for convergence analysis. Unlike existing algorithms based on ideal continuous communication, this application uses a state-dependent threshold function to adaptively adjust the information exchange frequency between neighbors, which significantly reduces the communication burden and avoids Zeno behavior while ensuring convergence. In addition, the embodiments of this application also have a dual-time-scale dynamic adjustment structure (fast anti-saturation dynamics and slow optimization dynamics), which is regarded as a singular perturbation system. The closed-loop system can be rigorously proven to converge to the optimal solution in a semi-global practical asymptotic manner through singular perturbation theory.

[0121] Based on the same inventive concept, this application also provides a multi-agent system distributed optimal resource allocation system for implementing the aforementioned multi-agent system distributed optimal resource allocation method. The solution provided by this multi-agent system distributed optimal resource allocation system is similar to the implementation scheme described in the above method. Therefore, the specific limitations in the embodiments of the multi-agent system distributed optimal resource allocation system provided below can be found in the limitations of the multi-agent system distributed optimal resource allocation method described above, and will not be repeated here.

[0122] In one exemplary embodiment, a multi-agent system distributed optimal resource allocation system is provided, applied to a distributed resource allocation control scenario of a cluster control system including multiple physical intelligent devices. The multi-agent system distributed optimal resource allocation system is used to execute the multi-agent system distributed optimal resource allocation method.

[0123] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 8As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs stored in the non-volatile storage media. The database stores the actual system requirements, communication network topology data, status data, and resource allocation data for each physical intelligent device in the cluster control system. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When executed by the processor, the computer program implements a distributed optimal resource allocation method for a multi-agent system.

[0124] Those skilled in the art will understand that Figure 8 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0125] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0126] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0127] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0128] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Moreover, the collection, use and processing of the relevant data are carried out in compliance with the relevant data protection laws and policies of the country where the location is located, and with the authorization granted by the owner of the corresponding device.

[0129] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0130] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0131] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0132] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A distributed optimal resource allocation method for a multi-agent system, applied to a distributed resource allocation control scenario in a cluster control system including multiple physical intelligent devices, characterized in that, The distributed optimal resource allocation method for multi-agent systems includes: The actual system requirements, communication network topology data, status data, and resource data to be allocated of each physical intelligent device in the cluster control system are collected in real time. Based on the acquired basic attribute data and the actual system requirements, a multi-agent resource allocation mathematical model is constructed. The multi-agent resource allocation mathematical model is used to represent a distributed resource allocation problem with an unknown cost function and subject to actuator saturation constraints. In the distributed resource allocation problem, each physical intelligent device can obtain the real-time measurement value of the local cost function, and the control input of the physical intelligent device is subject to nonlinear truncation constraints due to actuator saturation. Based on the communication network topology data, the resource data to be allocated, and the status data, a distributed extreme value search optimization method based on anti-saturation compensation mechanism and dynamic event triggering mechanism is used to solve the multi-agent resource allocation mathematical model, determine the optimal resource allocation scheme, and generate control commands. The control command is sent to the cluster control system to control the cluster control system to allocate resources to each of the physical intelligent devices, thereby completing the distributed optimal resource allocation of the cluster control system.

2. The distributed optimal resource allocation method for a multi-agent system according to claim 1, characterized in that, The basic attribute data includes the number of physical intelligent devices, the unique identifier of each physical intelligent device, and physical characteristic parameters. The communication network topology data includes the communication link relationship between each physical intelligent device, communication bandwidth, and communication latency. The resource data to be allocated includes the total amount of resources to be allocated.

3. The distributed optimal resource allocation method for a multi-agent system according to claim 1 or 2, characterized in that, The mathematical expression for the multi-agent resource allocation model is as follows: ； ； in, This represents the system's global optimization goal. For the global cost function, For the first Local cost function of a physical intelligent device The number of physical smart devices. This represents a global equality constraint. For the first The output status of each physical smart device, and , Represents an n-dimensional real vector space. For the first The actual system requirements of a physical intelligent device.

4. The distributed optimal resource allocation method for a multi-agent system according to claim 1, characterized in that, Based on the communication network topology data, the resource data to be allocated, and the state data, a distributed extreme value search optimization method based on anti-saturation compensation mechanism and dynamic event triggering mechanism is used to solve the multi-agent resource allocation mathematical model, determine the optimal resource allocation scheme, and generate control commands, specifically including: Based on the communication network topology data, the state data, and the resource data to be allocated, and combined with the local cost function characteristics of the multi-agent resource allocation mathematical model and the nonlinear truncation constraint boundary of the actuator saturation, the range of values ​​for the parameters to be tuned is determined; the parameters to be tuned include the small positive parameters of the extreme value search control strategy, the threshold parameters of the dynamic event triggering mechanism, and the coefficients of the anti-saturation compensation term; Based on the value range of the parameter to be tuned, determine the tuned parameter; The tuned parameters are substituted into the expression of the distributed extreme value search optimization method, and the output state of each physical intelligent device is used as the decision variable. The decision variables of each physical intelligent device are iteratively updated until they converge to the optimal solution of the multi-agent resource allocation mathematical model, thus obtaining the converged optimal decision variables of each physical intelligent device. The distributed extreme value search optimization method estimates the gradient information of the unknown local cost function based on the real-time measured value of the local cost function of each physical intelligent device using an extreme value search control strategy, and uses an anti-saturation compensation term to eliminate the integral saturation phenomenon caused by actuator saturation. The dynamic event triggering mechanism determines the information exchange triggering condition based on the state estimation error of the physical intelligent device, and adaptively adjusts the information exchange frequency between adjacent physical intelligent devices based on the information exchange triggering condition to control the on-demand information interaction between the physical intelligent devices. Based on the convergent optimal decision variables of each of the aforementioned physical intelligent devices, the optimal resource allocation scheme is determined; Based on the optimal resource allocation scheme, corresponding control commands are generated.

5. The distributed optimal resource allocation method for a multi-agent system according to claim 4, characterized in that, The communication network topology data is undirected strongly connected topology data, which includes a set of nodes, a set of edges, and topology matrix information characterizing the network topology. Each node in the set of nodes represents a physical intelligent device, and each edge in the set of edges represents a communication connection between two physical intelligent devices.

6. The distributed optimal resource allocation method for a multi-agent system according to claim 4, characterized in that, The expression for the distributed extreme value search optimization method is: ； in, Indicates the first Real-time rate of change of decision variables for each physical intelligent device For the saturation function of the actuator, For the first Input from a physical smart device for The first derivative with respect to time, All are the first Adjustable gain of a physical smart device For the first Anti-saturation compensation term for each physical intelligent device For time, For the adjacent first The first physical smart device and the first Communication weights between physical intelligent devices Indicates the first Lagrange multiplier signals transmitted by a physical intelligent device For the small positive parameter of the extreme value search control strategy, For the first Real-time measurement of the local cost function of a physical intelligent device For the first Lagrange multipliers corresponding to the coupling constraints of each physical intelligent device for The first derivative with respect to time, For the first The accompanying variables of a physical intelligent device for The first derivative with respect to time; The expression for the dynamic event triggering mechanism is: ； in, All are the first Adjustable gain of a physical smart device In the dynamic event triggering mechanism, the first A physical smart device in Internal state variables at any given time; for The first derivative with respect to time, For the first A physical smart device in The state estimation error at time t, and , This indicates the time of the kth trigger.

7. The distributed optimal resource allocation method for a multi-agent system according to claim 4, characterized in that, The distributed extreme value search optimization method adopts a dual-time-scale dynamic adjustment structure. The dual-timescale dynamic adjustment structure specifically includes: The dynamic consistency estimator is adjusted based on the first time scale to achieve rapid estimation of global information; The output state of the physical intelligent device is adjusted based on the second time scale so that the decision variables of the physical intelligent device converge to the optimal solution, and the convergence speed is affected by the cost function and constraints. The adjustment rate of the first time scale is greater than the adjustment rate of the second time scale.

8. The distributed optimal resource allocation method for a multi-agent system according to claim 7, characterized in that, After substituting the tuned parameters into the expression of the distributed extreme value search optimization method, using the output state of each physical intelligent device as a decision variable, iteratively updating the decision variables of each physical intelligent device until the decision variables of each physical intelligent device converge to the optimal solution of the multi-agent resource allocation mathematical model, and obtaining the converged optimal decision variables of each physical intelligent device, the multi-agent system distributed optimal resource allocation method further includes: Based on singular perturbation theory, a convergence analysis is performed on the dual-timescale dynamic adjustment structure to verify the effectiveness of the distributed extreme value search optimization method in converging to the optimal solution of the distributed resource allocation problem.

9. The distributed optimal resource allocation method for a multi-agent system according to claim 1, characterized in that, The distributed extreme value search optimization method satisfies the following assumptions: the local cost function of each physical intelligent device is quadratically continuous and differentiable and uniformly strong convex with respect to the decision variables, and the gradient of the local cost function satisfies the Lipschitz continuity property; the distributed resource allocation problem has an optimal solution.

10. A distributed optimal resource allocation system for a multi-agent system, applied to a distributed resource allocation control scenario in a cluster control system including multiple physical intelligent devices, characterized in that, The multi-agent system distributed optimal resource allocation system is used to execute the multi-agent system distributed optimal resource allocation method according to any one of claims 1-9.

Images