Distributed system stability optimization method, system and device and storage medium

Abstract

The embodiment of the invention provides a distributed system stability optimization method and system, equipment and a storage medium, and belongs to the technical field of distributed systems. The method comprises the following steps: a first agent device obtains initial decision variable data in one-to-one correspondence with agent devices in a distributed system at a first moment; based on the initial decision variable data and the communication time delay, control input quantity solving is carried out through a consistency protocol function, a cost gradient descent function and a control input function, and a target control input quantity is determined to update the initial decision variable data of the first agent equipment; the consistency protocol function represents the relationship of the consistency degree between the decision variables of the first intelligent agent device and the remaining intelligent agent devices at the second moment along with time change; the decision variable can be represented by a nonlinear function containing a second order; the second moment is determined based on the communication delay and the first moment. According to the embodiment of the invention, the stability of the distributed system can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of distributed systems, and in particular to a method, system, device and storage medium for optimizing the stability of a distributed system. Background Art

[0002] In distributed systems such as smart grids, power management, wireless networks, and mobile sensor networks, it is usually necessary for each agent in the distributed system to interact with each other and update strategies based on the interacted data (such as the trading strategy of the power grid, the power management strategy of the power supply, and other decision-making strategies). However, in the actual operation of distributed systems, communication delays are inevitable when agents share information or transmit data through dedicated wired or wireless communication networks. The existence of communication delays means that all agents in the distributed system will use outdated information to update their decision strategies. At this time, the stability and convergence of the distributed system may deteriorate due to the lack of accuracy of the updated decision strategy. Therefore, how to improve the stability of distributed systems is a technical problem that needs to be solved urgently. Summary of the invention

[0003] The main purpose of the embodiments of the present application is to propose a distributed system stability optimization method, system, device and storage medium, which can improve the stability of the distributed system.

[0004] To achieve the above object, a first aspect of an embodiment of the present application proposes a method for optimizing the stability of a distributed system, the method comprising:

[0005] Interacting with a second intelligent agent device in the same distributed system and in adjacent communication connection to perform decision data, and obtaining initial decision variable data corresponding to each intelligent agent device in the distributed system at a first moment;

[0006] Based on the initial decision variable data and communication delay of each of the distributed system, the control input is solved by a preset consistency protocol function, a cost gradient descent function and a control input function to determine the target control input of the first intelligent device; wherein the consistency protocol function represents the relationship between the decision change rate determined based on the consistency between the decision variable data of the first intelligent device at the second moment and the decision variable data of the remaining intelligent devices at the second moment and the change over time; the decision variable data can be represented by a nonlinear function including a second order; the cost gradient descent function represents the relationship between the auxiliary variable of the nonlinear function determined based on the cost gradient of the first intelligent device and the decision change rate and the change over time; the control input function represents the relationship between the control input determined based on the auxiliary variable, the cost gradient and the decision change rate and the change over time; the second moment is determined based on the communication delay and the first moment;

[0007] The initial decision variable data of the first intelligent agent device is updated according to the preset dynamics instance model and the target control input.

[0008] To achieve the above object, a second aspect of an embodiment of the present application proposes a distributed system, including:

[0009] Multiple intelligent agent devices, the multiple intelligent agent devices form a directed graph that is strongly connected, and the multiple intelligent agent devices can all be used as first intelligent agent devices to execute the distributed system stability optimization method described in any one of the first aspects.

[0010] To achieve the above-mentioned purpose, the third aspect of an embodiment of the present application proposes an intelligent body device, which includes a memory and a processor, the memory stores a computer program, and the processor implements the distributed system stability optimization method described in any one of the first aspects above when executing the computer program.

[0011] To achieve the above-mentioned purpose, the fourth aspect of an embodiment of the present application proposes a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the method for optimizing the stability of a distributed system as described in any one of the first aspects.

[0012] The optimization method, system, device and storage medium for the stability of a distributed system proposed in the present application can estimate the control input at the actual moment (i.e., the second moment) corresponding to the initial decision variable data in combination with the communication delay by constructing a consistency protocol function and a cost gradient descent function, and use it as the target control input. At this time, the initial decision variable data of the self can be updated based on the target control input and the dynamic instance model. At this time, the updated initial decision variable data comprehensively considers the communication delay with other intelligent devices in the distributed network system, and therefore, the decision strategy determined based on the decision variable is more accurate. Therefore, compared with the related art, the embodiment of the present application can improve the stability of the distributed network system. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 is a system schematic diagram of a distributed system provided in an embodiment of the present application;

[0014] Figure 2 The distributed system provided by the embodiment of the present application has a communication delay of τ 1 =1+0.4sin(t) The decision strategy x of each intelligent device under the setting i Trajectory simulation diagram;

[0015] Figure 3 The distributed system provided by the embodiment of the present application has a communication delay of τ1 =1+0.4sin(t) The state v of each intelligent device is set i Trajectory simulation diagram;

[0016] Figure 4 The distributed system provided by the embodiment of the present application has a communication delay of τ 2 ≤h=0.7 The decision strategy x of each intelligent device i Trajectory simulation diagram;

[0017] Figure 5 The distributed system provided by the embodiment of the present application has a communication delay of τ 2 ≤h=0.7 The state v of each intelligent device i Trajectory simulation diagram;

[0018] Figure 6 It is a flow chart of a method for optimizing the stability of a distributed system provided in an embodiment of the present application;

[0019] Figure 7 It is a schematic diagram of a process for each intelligent device to achieve Nash equilibrium in the distributed system stability optimization method provided in an embodiment of the present application;

[0020] Figure 8 It is a schematic diagram of decision variables of an embodiment of a method for optimizing the stability of a distributed system provided in an embodiment of the present application;

[0021] Fig. 9 It is a structural diagram of the hardware structure corresponding to the distributed system stability optimization method provided in the embodiment of the present application. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0023] It should be noted that, although the functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first", "second", etc. in the specification, claims and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0024] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0025] In distributed systems such as smart grids, power management, wireless networks, and mobile sensor networks, it is usually necessary for each agent in the distributed system to interact with data and update strategies based on the interactive data (such as power grid trading strategies, power supply power management strategies, and other decision strategies). With the development of cyber-physical systems (CPS), physical systems have been widely used in distributed systems. At the same time, there are more and more distributed algorithms combined with physical systems. However, in the prior art, all are based on ideal communication conditions to solve distributed Nash equilibrium to ensure the stability of the distributed system, such as a generalized Nash equilibrium solution algorithm designed based on a non-cooperative game problem with second-order dynamics and inequality constraints; another example is a distributed Nash equilibrium solution algorithm based on a game system with a second-order integrator type dynamic model, etc. However, in actual operation, when agents share information or transmit data through dedicated wired or wireless communication networks, communication delays will inevitably occur. The existence of communication delays means that all agents in the distributed system will use outdated information to update their decision strategies. At this time, it is easy to cause the stability and convergence of the distributed system to deteriorate due to the low accuracy of the updated decision strategy. Therefore, how to improve the stability of distributed systems is a technical problem that needs to be solved urgently. Based on this, the embodiments of the present application provide a method, system, device and storage medium for optimizing the stability of a distributed system, which can improve the stability of the distributed system.

[0026] Reference Figure 1 As shown, the distributed system of the embodiment of the present application includes multiple intelligent agent devices, and the multiple intelligent agent devices can all perform the following steps as the first intelligent agent device:

[0027] Interacting with a second intelligent agent device in the same distributed system and in adjacent communication connection to perform decision data, and obtaining initial decision variable data corresponding to each intelligent agent device in the distributed system at a first moment;

[0028] Based on the initial decision variable data and communication delay of each of the distributed system, the control input is solved by a preset consistency protocol function, a cost gradient descent function and a control input function to determine the target control input of the first intelligent device; wherein the consistency protocol function represents the relationship between the decision change rate determined based on the consistency between the decision variable data of the first intelligent device at the second moment and the decision variable data of the remaining intelligent devices at the second moment and the change over time; the decision variable data can be represented by a nonlinear function including a second order; the cost gradient descent function represents the relationship between the auxiliary variable of the nonlinear function determined based on the cost gradient of the first intelligent device and the decision change rate and the change over time; the control input function represents the relationship between the control input determined based on the auxiliary variable, the cost gradient and the decision change rate and the change over time; the second moment is determined based on the communication delay and the first moment;

[0029] The initial decision variable data of the first intelligent agent device is updated according to the preset dynamics instance model and the target control input.

[0030] The dynamic instance model represents the relationship between the decision variables and the target control input, and is constructed according to the decision variables in the actual application scenario. The dynamic instance model can be dynamically configured to dynamically adapt to different application scenarios. Therefore, after inputting the target control input and the initial decision variable data at the first moment, the decision variables of the agent itself combined with the communication delay and the decision quantity estimates of the remaining agents can be obtained.

[0031] The first moment is the moment when the first intelligent agent is triggered to solve the target control input. For the first intelligent device, the initial decision variable data of the remaining intelligent devices in the distributed system obtained by the first intelligent device is obtained by the first intelligent device from the second intelligent device at the first moment. Each intelligent device can forward the initial decision variable data of its adjacent intelligent device. The second intelligent device is an intelligent device that directly communicates with the first intelligent device, such as a direct physical connection or a wireless connection.

[0032] The initial decision variable data is the decision variable data at the first moment. Each decision variable data includes its own decision variable and the estimated decision variable values ​​of the remaining intelligent devices. When the distributed system is initialized, the decision variable data of the remaining intelligent devices in each intelligent device are updated based on the real value of the interaction, and are subsequently updated based on the consistency protocol function.

[0033] The decision change rate is the derivative of the decision variable data.

[0034] It is understandable that each intelligent device in the distributed system acts as a first intelligent device to solve its own target control quantity. When each intelligent device in the distributed system has completed the update of its own initial decision quantity, the distributed system is in a Nash equilibrium state.

[0035] It is understandable that in the embodiment of the present application, each first intelligent device can only obtain the initial decision variable data of each remaining intelligent device in the distributed system from its adjacent second intelligent device. Thus, each intelligent device can be used as the initial decision variable data synchronously calculated by the first intelligent device to ensure that the distributed system is stable.

[0036] The cost gradient can be determined by derivation of the decision variable data based on the preset cost function, and the auxiliary variable can be used to simplify the constraints of the nonlinear function or the expression of the nonlinear function. Both the decision variable and the cost gradient can be expressed by functions.

[0037] For example, assume that the following content is defined in a distributed system:

[0038] Setting up a directed graph for a distributed network system in, is a node set, ε is the edge set, is the adjacency matrix, denoted as A[α ij ] N*N ; where N represents the number of intelligent devices in the distributed system, α ij The matrix elements in the i-th row and j-th column of the adjacency matrix represent the relationship between intelligent device i and intelligent device j. For example, they can all be set to 1 / N. When the communication relationship between the intelligent devices in the distributed system is determined, the adjacency matrix is ​​also determined.

[0039] Define each agent device i as belonging to The decision variables x i And the local cost function J i (x i ,x -i ), define the policy variable data X in the intelligent device i i =col(x i1 ,....,x iN ) belong to For intelligent device i in X i The pseudo gradient at x ij represents the estimated value of the strategy variable of the overall device to the agent device j, and x ii =x i ;x ijDetermined by the consistency protocol function;

[0040] Define the remaining decision variables for the remaining agents Define the total set of decision variables of all agent devices in agent i as The goal of each intelligent device is to optimize the local cost function J i (x i ,x -i ). Where n represents the state dimension and n is 1.

[0041] The general dynamics model of each intelligent device is assumed to have the following second-order nonlinear dynamics function with uncertain parameters, wherein the general dynamics model is set to formula (1) and formula (2); formula (1): Formula (2): in, and is a decision variable x i The derivative of (t), that is, the rate of change of decision; is the unknown constant of the intelligent device i The nonlinear function of . Can be set according to actual needs; i (t) is the control input of intelligent device i.

[0042] Then assume that Assumption 1: For each intelligent device, its local cost function J i (x i ,x -i ) in x i is continuously differentiable, and x -i The corresponding x i is a convex function, then we have

[0043] Assumption 2: F(x) is is strongly monotonic on ω, and on l F The first Lipschitz condition is satisfied, that is, for a positive constant ω, (y 1 -y 2 ) T [F(y 1 )-F(y 2 )]≥ω||y 1 -y 2 || 2 ; For a normal number l F There is ||F(y 1 )-F(y 2 )||≤l F ||y 1 -y 2 ||. Among them, y 1,y 2 are any two independent variable parameters of F(x), y 1 ,y 2 All belong to At this time, ω is a strongly monotone operator, l F As the first Lipschitz constant, it can be used to solve Nash equilibrium with second-order nonlinear functions.

[0044] definition It is called game The Nash equilibrium of In the case of , then both assumptions 1 and 2 are valid, and Then there is x * is the only existing Nash equilibrium for the non-cooperative game G. X=col(X 1 ,....,X N ).

[0045] Assumption 3: There is a directed path between any two nodes in the graph and it is a weight-balanced directed graph. L is The Laplace matrix of . To meet and Among them, 1 N I represents a column matrix with all values ​​being 1. N Represents the identity matrix.

[0046] Assumption 4: Nonlinear function g i () satisfies the second Lipschitz condition, where the second Lipschitz condition is set as follows:

[0047]

[0048] Among them, l v and l x are the second and third Lipschitz constants, respectively, which are positive constants. i , v′ i 、x i ,

[0049] The consistency protocol function, cost gradient descent function, and control input function are set as follows: Formula (3) to Formula (5):

[0050]

[0051] Among them, in the case of non-cooperative games that meet assumptions 1 to 4:

[0052] Assumption 5: τ 1 (t)∈[0,d],τ 1(t)≤σ<1; where d is the constant upper bound of the time-varying delay, and σ is the upper bound of the rate of change of the time-varying delay, that is, the rate of change threshold. The constants in formula (3) to formula (5) satisfy the following conditions:

[0053] The first constant satisfies

[0054] The second constant satisfies

[0055] The third constant satisfies

[0056] The fourth constant satisfies

[0057] Among them, l v The second Lipschitz constant is determined when the nonlinear function satisfies the preset second Lipschitz condition; x is the third Lipschitz constant determined when the nonlinear function satisfies the preset second Lipschitz condition; ε is the fifth constant; ω is the strong monotone operator when the pseudo gradient mapping function of the first intelligent device satisfies the preset first Lipschitz condition; l F is the first Lipschitz constant determined when the pseudo gradient mapping function of the first intelligent device satisfies the preset first Lipschitz condition; 2 is the second smallest eigenvalue of the adjacency matrix recorded in the first agent device.

[0058] Positive definite matrix Among them, Q 11 , Q 22 , Q 33 , Q 44 , Q 55 , Q 66 , Q 13 , Q 23 , Q 46 , Q 56 The following conditions are met respectively:

[0059]

[0060] Assumption 6: τ 2 (t)∈[0,d], which means a time-varying delay with a fast change, and there is no constraint on its rate of change; then the constants in formulas (3) to (5) satisfy the following conditions:

[0061] The first constant satisfies

[0062] The second constant satisfies

[0063] The third constant satisfies

[0064] The fourth constant satisfies

[0065] Among them, l v The second Lipschitz constant is determined when the nonlinear function satisfies the preset second Lipschitz condition; x is the third Lipschitz constant determined when the nonlinear function satisfies the preset second Lipschitz condition; ε is the fifth constant; ω is the strong monotone operator when the pseudo gradient mapping function of the first intelligent device satisfies the preset first Lipschitz condition; l F The first Lipschitz constant is determined when the pseudo gradient mapping function of the first intelligent device satisfies the preset first Lipschitz condition. Nn is the identity matrix.

[0066] Positive definite matrix Middle Q 11 , Q 22 , Q 33 , Q 44 , Q 55 , Q 66 , Q 13 , Q 23 , Q 46 , Q 56 The following conditions are met respectively:

[0067]

[0068] In the above assumptions 5 and 6, ε is a positive constant, P 2 , P 3 , P 4 , P 5 are all positive definite matrices. Wherein, * indicates that the matrix element can take the value of 0. 2 , P 3 , P 4 , P 5 There is no limitation on how to set it, and those skilled in the art can selectively set it according to actual needs.

[0069] For formula (3) to formula (5), based on the general dynamics model, the following process formula can be obtained:

[0070]

[0071] At this time, after the dynamic model and the cost function are instantiated (such as in the power system, the power cost function and the dynamic instance model of the power system are input), the corresponding process formula can be substituted for theoretical derivation to determine the Since the positive definite matrix Q is an auxiliary matrix constructed in the process of designing the effectiveness of the Lyapunov function analysis algorithm, the constructed Lyapunov function includes the system state error variable and the auxiliary state error variable considering the delay. The Lyapunov function is a stability judgment method that can make the above The value determined in the theoretical derivation process makes the entire distributed system converge to the Nash equilibrium. Since the use of Lyapunov function for theoretical derivation is a common method, we will not go into details about how to use the positive definite matrix Q. Once determined, the target control input can be determined through formulas (4) and (5).

[0072] At this time, Figure 1 Taking the distributed system shown in the figure as an example, there are 5 intelligent devices in the distributed system, namely, intelligent device 1, intelligent device 2, intelligent device 3, intelligent device 4 and intelligent device 5. The dynamics instance model of each intelligent device is set to include as well as The local cost function of agent device i is Among them, (b 1,1 ,...,b 1,5 )=(5,8,6,9,7);(b 2,1 ,...,b 2,5 )=(12,10,11,11,13);(b 3,1 ,...,b 3,5 )=(1,0.5,0.8,0.7,1.1); p 0 =200; p 1 =0.1. That is, the cost functions of Agents 1 to 5 are

[0073] as well as Taking agent 1 as an example, when agent 1 satisfies assumptions 1-2 and the strong monotonicity condition in assumption 4 and the first Lipschitz condition, it can be determined that ω = 1.2; l F =2.4; l x =1; l v =1. Similarly, the corresponding ω and l can be determined for each other intelligent device. F , l x , l v .

[0074] When each intelligent device in the distributed system is running, when the decision variables of each intelligent device are stable, the Nash equilibrium processing process is started, and each intelligent device is used as the first intelligent device to update its own decision strategy. Taking intelligent device 1 as an example, for intelligent device 1, when τ1 =1+0.4sin(t), that is, agent 1 satisfies assumption 5. According to assumption 5 and ω=1.2; l F =2.4; l x =1; l v =1 to determine the constant coefficients of formulas (3) to (5) and the positive definite matrix used for solving. When τ 1 ≤h=0.7, which means that Assumption 6 is satisfied. Based on Assumption 6 and ω=1.2; l F =2.4; l x =1; l v =1 to determine the constant coefficients of formulas (3) to (5) and the positive definite matrix used for solving. Then, based on formulas (3) to (5), positive definite matrices, general dynamics example models, and cost functions, we can solve the target control input of agent 1 to be 68.3735. Similarly, agents 2 to 5 can refer to agent 1 to obtain their respective target input control quantities. Agents 1 to 5 reach Nash equilibrium, and the analytical solution of Nash equilibrium in the distributed system is (68.3735, 132.3495, 85.0497, 96.3896, 61.9932). Among them, 132.3495 corresponds to agent 2; 85.0497 corresponds to agent 3, 96.3896 corresponds to agent 4, and 61.9932 corresponds to agent 5.

[0075] At this time, the input control amount of each intelligent device is updated to the target input control amount, then refer to Figure 2 and Figure 3 The simulation results in the introduction of the slow-changing time-varying communication delay that satisfies Assumption 5 are as follows: i , Can converge to Nash equilibrium, the variable v of each agent device i can converge to 0. Similarly, refer to Figure 4 and Figure 5 The simulation results in the introduction of the fast-changing time-varying communication delay satisfying Assumption 6, the decision variable x of each agent i , Can converge to Nash equilibrium, the decision change rate of each intelligent device v i can converge to 0.

[0076] Refer to the following Figure 6 As shown, according to the distributed system stability optimization method provided in the embodiment of the present application, for a first intelligent agent device, the first intelligent agent device is located in the distributed system, and the method includes:

[0077] Step S100, exchanging decision data with a second intelligent agent device located in the same distributed system and in adjacent communication connection, to obtain initial decision variable data corresponding to each intelligent agent device in the distributed system at a first moment;

[0078] Step S200, based on each initial decision variable data and communication delay of the distributed system, the control input quantity is solved through the preset consistency protocol function, cost gradient descent function and control input function to determine the target control input quantity of the first intelligent device; wherein the consistency protocol function represents the relationship between the decision change rate determined based on the consistency degree between the decision variable data of the first intelligent device at the second moment and the decision variable data of the remaining intelligent devices at the second moment and the change of the decision variable over time; the decision variable can be represented by a nonlinear function including a second order; the cost gradient descent function represents the relationship between the auxiliary variable of the nonlinear function determined based on the cost gradient of the first intelligent device and the decision change rate and the change of the decision over time; the control input function represents the relationship between the control input quantity determined based on the auxiliary variable, the cost gradient and the decision change rate and the change of the decision over time; the second moment is determined based on the communication delay and the first moment;

[0079] Step S300: Update the initial decision variable data of the first intelligent agent device according to the preset dynamics instance model and the target control input.

[0080] Therefore, by constructing a consistency protocol function and a cost gradient descent function, the control input at the actual moment (i.e., the second moment) corresponding to the initial decision variable data can be estimated in combination with the communication delay, and used as the target control input. At this time, based on the target control input and the dynamic instance model, the decision variable itself can be updated. At this time, the decision variable comprehensively considers the communication delay with other intelligent devices in the distributed network system. Therefore, the decision strategy determined based on the decision variable is more accurate. The embodiment of the present application can improve the stability of the distributed network system.

[0081] It is understandable that the decision variable data can be represented by a second-order nonlinear function to construct a heterogeneous second-order nonlinear dynamic model with unknown terms. At the same time, the communication delay is introduced into the consistency protocol function, so that the method of the embodiment of the present application can take into account the heterogeneous second-order nonlinear dynamic model with unknown terms, and solve the influence of communication delay on the stability and convergence of the distributed system.

[0082] It is understandable that the introduction of the cost gradient of each intelligent device can make it possible to combine the total cost of the entire distributed system in the process of solving the target control input, and further ensure that Nash equilibrium can be reached after each intelligent device executes the above steps.

[0083] For example, refer to Figure 7 As shown, at the first moment, each intelligent device obtains the initial decision variable data of the remaining intelligent devices in the distributed system. Set the participant index i=1; select a first intelligent device from the distributed system as participant i; for participant i, determine the initial decision quantity data of participant i and other (n-1) participants (i.e., the remaining intelligent devices) for their own target control quantity input. Each intelligent device in the distributed system determines the target control input quantity. When each intelligent device updates its own target control input quantity strategy based on its own target control input quantity, it determines whether there is a participant who has updated the decision variable. If there is a participant who has updated the decision variable, each intelligent device is re-determined until the decision variables of each intelligent device in the distributed network system are no longer updated. At this time, the distributed system reaches Nash equilibrium, and all participant sets reach Nash equilibrium.

[0084] It can be understood that the directed graph formed by the distributed system is a strongly directed graph; the decision data is interacted with the second intelligent agent device located in the same distributed system and in adjacent communication connection, and the initial decision variable data corresponding to each intelligent agent device in the distributed system is obtained, including:

[0085] Sending its own initial decision variable data to the second intelligent agent device in the distributed system;

[0086] Receive the initial decision variable data of the second intelligent device sent by the second intelligent device and the initial decision variable data of the remaining intelligent devices forwarded by the second intelligent device.

[0087] By acquiring the decision variable data of the remaining intelligent devices only from the adjacent second intelligent device, the calculation of the target control input quantity under the condition of communication delay can be further simplified, thereby improving the calculation efficiency.

[0088] It is understandable that the constant coefficients of the consistency protocol function, the cost gradient descent function, and the control input function are determined by the following steps:

[0089] Obtain a cost function constructed based on the decision variables and the set of remaining decision variables;

[0090] In the cost function, the decision variables x i is continuously differentiable, and for the remaining decision variable set x -i For x i In the case of a convex function, the gradient mapping function is determined according to the cost function;

[0091] According to a preset first Lipschitz condition, a strong monotone operator and a first Lipschitz constant in the gradient mapping function are determined;

[0092] According to the preset second Lipschitz condition, determining the second Lipschitz constant and the third Lipschitz constant of the nonlinear function;

[0093] Obtain coefficient range determination conditions corresponding one-to-one to each constant coefficient of the consistency protocol function, the cost gradient descent function, and the control input function;

[0094] According to the range of each coefficient, the conditions, the strong monotone operator, the first Lipschitz constant, the second Lipschitz constant and the third Lipschitz constant are determined to respectively determine the constant coefficients of the consistency protocol function, the cost gradient descent function and the control input function.

[0095] It is understandable that, assuming the cost function is J i (x i ,x -i ), then the gradient mapping function is set to Among them, x i is the decision variable, and the remaining decision variable set x -i :=(x 1 ,...,x i-1 ,...,x N ), which represents the set of decision changes of the remaining intelligent devices except the first intelligent device in the distributed system.

[0096] By dynamically determining the coefficient range determination conditions of the consistency protocol function, the cost gradient descent function, and the constant coefficients of the control input function, a general model can be constructed, thereby making the method of the embodiment of the present application more adaptable.

[0097] The coefficient range determination condition defines the value function of each coefficient boundary; based on the coefficient range determination condition, each coefficient range can be determined, and a value can be randomly selected from the coefficient range, or a value can be selected from the boundary of the coefficient range determination condition. In this regard, the embodiments of the present application do not restrict how the values ​​of each constant coefficient are selected.

[0098] It is understood that the method further comprises:

[0099] Determine a positive definite matrix based on a strongly monotone operator, a first Lipschitz constant, a second Lipschitz constant, and a third Lipschitz constant;

[0100] Based on each initial decision variable data in the distributed system, the control input quantity is solved through a preset consistency protocol function, a cost gradient descent function, and a control input function to determine the target control input quantity of the first intelligent device, including:

[0101] Based on each initial decision variable data in the distributed system, the control input quantity is solved through the positive definite matrix, the consistency protocol function, the cost gradient descent function and the control input function to determine the target control input quantity of the first intelligent device.

[0102] It can be understood that since the decision variable data can be represented as a dynamic model containing a second-order nonlinear function; therefore, by increasing the positive definite matrix, the efficiency of the solution can be ensured.

[0103] It is understandable that dynamically determining the positive definite matrix through a strong monotone operator, a first Lipschitz constant, a second Lipschitz constant, and a third Lipschitz constant can provide the same solution method, thereby improving the adaptability of the method.

[0104] It can be understood that a positive definite matrix is ​​determined by the following steps:

[0105] When the communication delay satisfies a preset slow delay condition, a positive definite matrix matching the slow delay condition is determined from multiple candidate positive definite matrices; wherein the slow delay condition indicates that the communication delay changes based on the delay change rate, and the delay change rate is less than a preset change rate threshold;

[0106] When the communication delay satisfies a preset fast delay condition, a positive definite matrix matching the fast delay condition is determined from a plurality of candidate positive definite matrices; the fast delay condition indicates that the communication delay does not change based on the delay change rate.

[0107] It is understandable that by setting the positive definite matrix under two conditions, the problem caused by communication delay is also considered in the solution process, thereby further improving the reliability of the target control input.

[0108] It is understandable that the consistency protocol function is The cost gradient descent function is Control Input Function

[0109] in, is the decision change rate, X i () represents the decision variable data of intelligent device i; t represents the first moment, τ(t) represents the communication delay at moment t, and t-τ(t) represents the second moment; X j () represents the decision variable data of intelligent device j; N represents the number of intelligent devices in the distributed system; α is the first constant, v i (t) represents the decision change rate at time t; Indicates that the intelligent device i is at x i Pseudo gradient of (t); is an auxiliary variable; k 1 , k 3 , k2 , α are the first constant, the second constant, the third constant and the fourth constant respectively; u i (t) represents the control input, α ij Represents the element in the i-th row and j-th column of the adjacency matrix in the first agent device.

[0110] It is understandable that the decision variables can be determined using a general dynamics model, which is as follows:

[0111]

[0112] in, It is an expression of a nonlinear general function. When solving the actual decision variation, it can be set as a specific nonlinear function based on the actual application scenario to form a dynamic example model.

[0113] It can be understood that, where the first constant satisfies

[0114] The second constant satisfies

[0115] Among them, the third constant satisfies

[0116] Among them, the fourth constant satisfies

[0117] Among them, l v The second Lipschitz constant is determined when the nonlinear function satisfies the preset second Lipschitz condition; x is the third Lipschitz constant determined when the nonlinear function satisfies the preset second Lipschitz condition; ε is the fifth constant; ω is the strong monotone operator when the pseudo gradient mapping function of the first intelligent device satisfies the preset first Lipschitz condition; l F The first Lipschitz constant is determined when the pseudo gradient mapping function of the first intelligent agent device satisfies a preset first Lipschitz condition.

[0118] It is understandable that, in some embodiments, the above constants may be directly configured in the system, or may be dynamically determined by setting specific dynamic model related parameters and cost functions, and the embodiments of the present application do not impose any limitation on this.

[0119] For example, refer to Figure 8 As shown, there are a total of ni intelligent devices in the distributed system, namely participants i~n. There is a communication delay between every two adjacent participants. Each participant executes steps 1 to 4. Taking participant i as an example, the specific steps are as follows:

[0120] Step 1: Transmit the initial decision data to the neighbors (e.g., participant i-1, participant i-2);

[0121] Step 2: Receive the initial decision-making quantity data of the neighbors (there are multiple initial decision-making quantity data, which are the initial sentence measurement data of the remaining participants).

[0122] Step 3: Use the consistency method to estimate the strategies of all participants, that is, estimate the decision change of each participant based on the consistency agreement function in step 2;

[0123] Step 4: Considering the communication delay, the gradient descent method is used to locally update its own decision, that is, the target control input is determined based on the cost gradient descent function and the control input function, and the initial decision variable data is updated based on the target control input.

[0124] The embodiment of the present application also provides an intelligent device, which includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the above-mentioned distributed system stability optimization method when executing the computer program. The intelligent device can be any intelligent terminal including a tablet computer, a car computer, etc.

[0125] See also Fig. 9 , Fig. 9 The hardware structure of an intelligent agent device of another embodiment is shown, and the intelligent agent device includes:

[0126] The processor 701 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application;

[0127] The memory 702 may be a NAND flash, and the relevant program code is stored in the memory 702, and is called by the processor 701 to execute the distributed system stability optimization method of the embodiment of the present application;

[0128] Input / output interface 703, used to implement information input and output;

[0129] Communication interface 704, used to realize communication interaction between the device and other devices, which can be realized through wired mode (such as USB, network cable, etc.) or wireless mode (such as mobile network, WIFI, Bluetooth, etc.);

[0130] A bus 705 that transmits information between the various components of the device (e.g., the processor 701, the memory 702, the input / output interface 703, and the communication interface 704);

[0131] The processor 701 , the memory 702 , the input / output interface 703 and the communication interface 704 are connected to each other in communication within the device via a bus 705 .

[0132] An embodiment of the present application also provides a computer-readable storage medium, which is a computer-readable storage medium that stores a computer program. When the computer program is executed by a processor, the above-mentioned method for optimizing the stability of the distributed system is implemented.

[0133] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0134] The embodiments described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0135] Those skilled in the art will appreciate that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0136] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0137] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0138] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0139] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0140] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the above units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0141] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0142] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0143] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes multiple instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store programs.

[0144] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but the scope of the rights of the present invention is not limited thereto. Any modification, equivalent substitution and improvement made by a person skilled in the art without departing from the scope and essence of the present invention should be within the scope of the rights of the present invention.

Claims

1. A method for optimizing the stability of a distributed system, characterized in that: Applied to a first intelligent agent device, the first intelligent agent device is located in a distributed system, the method includes: Interacting with a second intelligent agent device in the same distributed system and in adjacent communication connection to perform decision data, and obtaining initial decision variable data corresponding to each intelligent agent device in the distributed system at a first moment; Based on the initial decision variable data and communication delay of each of the distributed system, the control input is solved by a preset consistency protocol function, a cost gradient descent function and a control input function to determine the target control input of the first intelligent device; wherein the consistency protocol function represents the relationship between the decision change rate determined based on the consistency between the decision variable data of the first intelligent device at the second moment and the decision variable data of the remaining intelligent devices at the second moment and the change over time; the decision variable data can be represented by a nonlinear function including a second order; the cost gradient descent function represents the relationship between the auxiliary variable of the nonlinear function determined based on the cost gradient of the first intelligent device and the decision change rate and the change over time; the control input function represents the relationship between the control input determined based on the auxiliary variable, the cost gradient and the decision change rate and the change over time; the second moment is determined based on the communication delay and the first moment; The initial decision variable data of the first intelligent agent device is updated according to the preset dynamics instance model and the target control input.

2. The method for optimizing the stability of a distributed system according to claim 1, characterized in that: The constant coefficients of the consistency protocol function, the cost gradient descent function and the control input function are determined by the following steps: Obtain a cost function constructed based on the decision variables and the set of remaining decision variables; In the cost function, the decision variables x i is continuously differentiable, and for the remaining decision variable set x -i For x i In the case of a convex function, a gradient mapping function is determined according to the cost function; According to a preset first Lipschitz condition, determining a strong monotone operator and a first Lipschitz constant in the gradient mapping function; Determining a second Lipschitz constant and a third Lipschitz constant of the nonlinear function according to a preset second Lipschitz condition; Obtaining coefficient range determination conditions corresponding one-to-one to each constant coefficient of the consistency protocol function, the cost gradient descent function, and the control input function; According to the conditions for determining the range of each coefficient, the strong monotone operator, the first Lipschitz constant, the second Lipschitz constant and the third Lipschitz constant, the constant coefficients of the consistency protocol function, the cost gradient descent function and the control input function are determined respectively.

3. The method for optimizing the stability of a distributed system according to claim 2, characterized in that: The method further comprises: Determining a positive definite matrix according to the strongly monotone operator, the first Lipschitz constant, the second Lipschitz constant and the third Lipschitz constant; The method of solving the control input quantity based on the initial decision variable data and the communication delay of each of the distributed systems through a preset consistency protocol function, a cost gradient descent function and a control input function to determine the target control input quantity of the first intelligent device includes: Based on the initial decision variable data and communication delay of the distributed system, the control input quantity is solved through the positive definite matrix, the consistency protocol function, the cost gradient descent function and the control input function to determine the target control input quantity of the first intelligent device.

4. The method for optimizing the stability of a distributed system according to claim 3, characterized in that: The positive definite matrix is ​​determined by the following steps: In the case where the communication delay satisfies a preset slow delay condition, determining a positive definite matrix matching the slow delay condition from a plurality of candidate positive definite matrices; wherein the slow delay condition indicates that the communication delay changes based on a delay change rate, and the delay change rate is less than a preset change rate threshold; When the communication delay satisfies a preset fast delay condition, a positive definite matrix matching the fast delay condition is determined from a plurality of candidate positive definite matrices; the fast delay condition indicates that the communication delay does not change based on a delay change rate.

5. The method for optimizing the stability of a distributed system according to claim 1, characterized in that: The consistency protocol function is The cost gradient descent function is The control input function Among them, the is the decision change rate, the X i () represents the decision variable data of the intelligent device i; t represents the first moment, τ(t) represents the communication delay at moment t, and t-τ(t) represents the second moment; X j () represents the decision variable data of the intelligent device j; N represents the number of intelligent devices in the distributed system; α is the first constant, v i (t) represents the decision change rate at time t; Indicates that intelligent device i is in X i Pseudo gradient of (t); is the auxiliary variable; k1, k3, k2, and α are respectively the first constant, the second constant, the third constant, and the fourth constant; u i (t) represents the control input, α ij Represents the element in the i-th row and j-th column of the adjacency matrix in the first agent device.

6. The method for optimizing the stability of a distributed system according to claim 5, characterized in that: in, The first constant satisfies Wherein, the second constant satisfies Wherein, the third constant satisfies Wherein, the fourth constant satisfies Among them, the l v is the second Lipschitz constant determined when the nonlinear function satisfies the preset second Lipschitz condition; x is the third Lipschitz constant determined when the nonlinear function satisfies the preset second Lipschitz condition; ε is the fifth constant; ω is the strong monotone operator when the pseudo gradient mapping function of the first intelligent device satisfies the preset first Lipschitz condition; l F The first Lipschitz constant is determined when the pseudo gradient mapping function of the first intelligent agent device satisfies a preset first Lipschitz condition.

7. The method for optimizing the stability of a distributed system according to claim 1, characterized in that: The directed graph formed by the distributed system is a strongly directed graph; the decision data interaction with the second intelligent agent device located in the same distributed system and in adjacent communication connection is performed to obtain initial decision variable data corresponding to each intelligent agent device in the distributed system, including: Sending its own initial decision variable data to the second intelligent agent device in the distributed system; Receive the initial decision variable data of the second intelligent device itself sent by the second intelligent device and the initial decision variable data of the remaining intelligent devices forwarded by the second intelligent device.

8. A distributed system, characterized in that: include: Multiple intelligent agent devices, the multiple intelligent agent devices form a directed graph that is strongly connected, and the multiple intelligent agent devices can all be used as first intelligent agent devices to execute the distributed system stability optimization method described in any one of claims 1 to 7.

9. An intelligent agent device, characterized in that: The intelligent agent device includes a memory and a processor, the memory stores a computer program, and the processor implements the distributed system stability optimization method described in any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the distributed system stability optimization method according to any one of claims 1 to 7 is implemented.

Images