Distributed unknown time-varying area multi-agent optimal coverage method and system

Abstract

The present invention discloses a distributed unknown time-varying area multi-autonomous agent optimal coverage method and system. In the method of the present invention, a method for establishing a parameterized model of an unknown time-varying environment density function using basis functions is provided; a second-order kinematic model is established for the multi-autonomous agent system; a method for updating parameters of an unknown time-varying density function based on distributed forgetting factor recursive least squares estimation is designed; and a distributed optimal coverage control strategy for a multi-autonomous agent system based on online density function estimation is provided. Environmental data is continuously collected by the multi-autonomous agent system, and density function estimation parameters are updated to achieve optimal coverage control of the multi-autonomous agent system. The present invention enables the multi-autonomous agent system to autonomously obtain an environmental density function model with high accuracy through distributed interaction, thereby achieving optimal area coverage in an unknown time-varying environment.

Description

Technical Field

[0001] The present invention relates to a distributed unknown time-varying area multi-autonomous agent optimal coverage method and system, belonging to the field of multi-autonomous agent collaborative control. Background Art

[0002] In recent years, with the rapid development of communication technology, computer systems, and sensor performance, the cost of multi-agent systems has been continuously reduced, and technical functions have been continuously improved. The application of multi-agent system collaborative control in many fields such as joint search and rescue, intelligent transportation, and regional monitoring is becoming more and more extensive. Compared with single-agent systems, multi-agent systems have greatly improved the adaptability and robustness of tasks. Coverage control, as the core work of multi-agent system collaborative control, has been widely studied in recent years. Coverage control can be applied in many important environmental monitoring and information collection scenarios. Its main goal is to propose a control law that enables the multi-agent system to converge to the optimal coverage configuration within the task area. Centralized coverage control requires the integration of all autonomous data information, centralized processing, and assignment of tasks to each subject. This places high demands on the communication and interaction capabilities of the autonomous body and the storage and computing capabilities of the centralized node. At the same time, centralized data transmission and task assignment also bring greater security and robustness risks to the multi-agent system. Distributed coverage control can well avoid these problems. Each subject collects data and processes information separately, and completes the coverage task through communication interaction.

[0003] The problem of multi-agent coverage control based on distributed processing has great research significance, and many existing studies have made significant contributions to this topic. The classic distributed coverage control method is to use Voronoi partitioning to divide the task area into corresponding Voronoi unit sets, and to achieve coverage control by driving each autonomous agent to track the centroid of its corresponding Voronoi unit. Based on this method, a large number of related studies have been carried out. However, most existing studies assume that the environmental density function of the task area is known, which ignores the fact that the environmental density function is usually unknown in actual situations. Subsequently, the situation where the environmental density function of the task area is unknown has received a lot of attention and research. These studies proposed estimation algorithms for approximating the unknown environmental density function, but most of the estimation algorithms have the problems of complex models, high computational costs or many assumptions. Furthermore, there are relatively few existing studies on the task area with unknown and time-varying density functions. Therefore, coverage control of the task area with unknown and time-varying density functions is a challenging problem that deserves further study. Summary of the invention

[0004] Purpose of the invention: In view of the problems and improvement needs of the current coverage control method based on distributed multi-autonomous agent system in unknown time-varying environment, the present invention proposes a distributed unknown time-varying area multi-autonomous agent optimal coverage method and system, which estimates the unknown time-varying environmental density function through the distributed forgetting factor recursive least squares algorithm, and then designs the corresponding coverage controller to drive the multi-autonomous agent system to achieve the optimal area coverage control task.

[0005] Technical solution: In order to achieve the above-mentioned purpose, the distributed unknown time-varying area multi-autonomous agent optimal coverage method proposed in the present invention comprises the following steps:

[0006] Based on the distributed forgetting factor recursive least squares estimation algorithm model, each subject performs a fusion calculation on the perceived environmental information and the environmental information obtained by communication to estimate the unknown time-varying environmental density function;

[0007] Each agent designs a corresponding second-order kinematic model-based covering controller according to its estimated density function, generates control input, and the autonomous agent performs corresponding actions according to the control input.

[0008] During the perception process, for the subject i, i = 1, ..., N at time k, its position coordinates and collected environmental data are recorded as (p i,k ,φ(p i,k ), where N is the number of agents, p i,k ∈R n is the position vector of agent i at time k, n is the dimension of task area Q, φ(p i,k )∈R is the environmental data collected by agent i at time k, that is, the density function value;

[0009] For the unknown time-varying ambient density function φ(q,t), it is assumed that there exists an ideal parameter vector satisfy:

[0010] φ(q,t)=K(q) T a(t)

[0011] Where q∈Q is any point in the task area, K: is the basis function vector, a(t) is the unknown time-varying parameter, represents an m-dimensional positive real vector;

[0012] The agent i, i = 1, ..., N uses the distributed forgetting factor recursive least squares estimation algorithm to estimate and update the parameter a(t) of the unknown time-varying density function, and then obtains the estimated density function in is the estimated parameter of agent i, specifically:

[0013] For the autonomous agent i, i = 1, ..., N, it is assumed that the position coordinates at time k and the collected environmental data (p i,k ,φ(p i,k )), using this data and combining it with stored data information and M i (k-1) Perform local estimation of the unknown time-varying density function parameters and calculate the generated and Z i (k):

[0014]

[0015] in is the estimated parameter of the unknown time-varying density function of subject i at time k-1, λ is the forgetting factor, 0.9≤λ<1;

[0016] The subjects communicate with each other, and the subject i obtains data information through communication interaction. and j∈I N Perform information fusion and computational generation and

[0017]

[0018] Among them I N = {1,…,N} represents the set of agents, B = {b ij} N×N is a weighted adjacency matrix. When agents i and j can communicate and interact, b ij > 0, otherwise b ij = 0, and satisfies b ij =b ji and M i (0) Initialize to K M I, where K M is a positive real number, and the value interval can be selected as 10 6 ~10 10 .

[0019] Furthermore, each subject designs a corresponding coverage controller based on the coverage control Lloyd algorithm according to its estimated density function, specifically:

[0020] The coverage control objective of the multi-agent system is to minimize the following coverage performance function:

[0021]

[0022] Where P = {p i, i=1,...,N} is the current position vector of N autonomous bodies, W={W i , i=1,...,N} is the arbitrary division of the convex task area Q by each subject, and W i is the area that autonomous agent i is responsible for. Assuming that the perception models of all autonomous agents are consistent, they are all omnidirectional perception models. f: is a non-decreasing differentiable function used to evaluate the perception performance of the autonomous agent on each point in the divided responsibility area;

[0023] Further, perform Voronoi partition on Q V = {V i ,i=1,...,N}:

[0024]

[0025] Weno Area V i The generalized quality Polar moment of inertia and centroid Given by:

[0026]

[0027] When selecting When , the coverage performance function under Voronoi partitioning can be written as:

[0028]

[0029] in It can represent the uncertainty of the perception of the autonomous subject i about the point q in the task area; the coverage performance function H(P) is calculated with respect to p i ,i∈I N Differentiate , and we get:

[0030]

[0031] The second-order kinematic model of the autonomous body in the present invention satisfies:

[0032]

[0033]

[0034] where v i is the velocity vector of the ith autonomous body, u i is the control input of the ith autonomous agent, i.e., the acceleration vector, k v is a positive constant; according to Lyapunov stability analysis and LaSalle invariant set theory, it can be analyzed that when p i ,i∈I N Both are in the Voronoi region V i ,i∈IN The centroid of the Voronoi configuration, that is, when the centroid-Voroi configuration is realized, the coverage performance function has a local optimal value; the Voronoi centroid is Then we have:

[0035]

[0036] When the environment information is unknown and time-varying, each subject estimates the density function based on its recursive least squares estimation algorithm using the distributed forgetting factor Calculate the corresponding Voronoi region V i Estimated generalized mass Estimated polar moment of inertia and estimated centroid

[0037]

[0038] When each agent is located at its corresponding estimated center of mass When , the coverage performance function has a local optimal value;

[0039] Furthermore, a coverage controller is designed based on the coverage control Lloyd algorithm to make the coverage performance converge to the local optimum:

[0040]

[0041] in is a uniformly positive definite gain matrix.

[0042] Based on the same inventive concept, the present invention proposes a distributed unknown time-varying area multi-autonomous agent optimal coverage system, which contains multiple autonomous agents that can communicate and interact in the task area, and each agent includes the following modules: an environmental density estimation module, which is used for the autonomous agent to fuse the perceived environmental information with the environmental information obtained by communication based on a distributed forgetting factor recursive least squares estimation algorithm model, and estimate the unknown time-varying environmental density function; an optimal coverage control module, which is used for the autonomous agent to design a corresponding coverage controller based on a second-order kinematic model according to its estimated density function, generate control input, and the autonomous agent performs corresponding actions according to the control input.

[0043] Based on the same inventive concept, the present invention provides a distributed unknown time-varying area multi-autonomous agent optimal coverage system, which includes multiple autonomous agents that can communicate and interact in the task area; the autonomous agent includes a sensor for collecting environmental data, a memory, a processor, and a computer program stored in the memory and run on the processor. When the computer program is loaded into the processor, it implements the distributed unknown time-varying area multi-autonomous agent optimal coverage method.

[0044] Based on the same inventive concept, the present invention provides a computer system, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the computer program is loaded into the processor, the distributed unknown time-varying area multi-autonomous agent optimal coverage method is implemented.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] 1) Compared with the centralized multi-autonomous agent coverage control method, the distributed unknown time-varying area multi-autonomous agent optimal coverage method proposed in the present invention does not need to integrate all autonomous agent data information and centrally process them. Instead, each agent collects data and processes information separately, and completes the coverage task through communication interaction. This reduces the requirements for the communication interaction capabilities and storage and computing capabilities of the autonomous agents, and improves the security and robustness of the multi-autonomous agent system.

[0047] 2) Compared with most coverage control methods of multi-autonomous agent systems that are only applicable to mission area environment density functions that are known, the distributed unknown time-varying area multi-autonomous agent optimal coverage method proposed in the present invention is more practical and more adaptable, enabling the multi-autonomous agent system to not only efficiently complete coverage control tasks in unknown environments, but also effectively handle coverage control tasks in time-varying environments.

[0048] 3) Most of the existing multi-autonomous system coverage control methods use a first-order kinematic model. The present invention establishes a second-order kinematic model for the multi-autonomous system. The second-order kinematic model takes into account the acceleration of the autonomous body, making the movement of the autonomous body safer and smoother.

[0049] 4) Compared with the existing estimation algorithms for unknown environmental density functions, the distributed forgetting factor recursive least squares estimation algorithm further proposed in the present invention has the characteristics of low storage requirements, low computational complexity and fast convergence speed. It can effectively estimate the unknown time-varying environmental density function and has high estimation accuracy, which helps the multi-autonomous system to efficiently complete the coverage control task.

[0050] 5) The distributed unknown time-varying area multi-autonomous agent optimal coverage method and system proposed in the present invention are highly adaptable. They are not only applicable to autonomous agents equipped with omnidirectional perception model sensors in convex environment areas, but can also be extended to autonomous agents equipped with other perception model sensors and non-convex environment areas. There are also many types of autonomous agents that are applicable, including aerial, land, and underwater robots, etc., and they have broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 Schematic diagram of the steps of the distributed unknown time-varying area multi-autonomous agent optimal coverage method according to an embodiment of the present invention;

[0052] Figure 2 It is a schematic diagram of the initial position of multiple autonomous bodies gathering in the simulation experiment of the present invention;

[0053] Figure 3 It is a state diagram of multiple autonomous agents in the covering process under the initial distribution of multiple autonomous agents in the simulation experiment of the present invention;

[0054] Figure 4 It is a schematic diagram of the coverage control result of the initial distribution situation of multiple autonomous agents gathering in an unknown time-varying environment in the simulation experiment of the present invention;

[0055] Figure 5 It is a schematic diagram of the estimation effect of the unknown time-varying density function under the situation of initial distribution of multi-agent aggregation in the simulation experiment of the present invention;

[0056] Figure 6 It is a schematic diagram of the initial position dispersion of multiple autonomous agents in the simulation experiment of the present invention;

[0057] Figure 7 It is a state diagram of multiple autonomous agents in the covering process under the condition of dispersed initial distribution of multiple autonomous agents in the simulation experiment of the present invention;

[0058] Figure 8 It is a schematic diagram of the coverage control results of the dispersed initial distribution situation of multiple autonomous agents in the simulation experiment of the present invention in an unknown time-varying environment;

[0059] Fig. 9 It is a schematic diagram of the estimation effect of the unknown time-varying density function under the condition of dispersed initial distribution of multiple autonomous agents in the simulation experiment of the present invention. DETAILED DESCRIPTION

[0060] In order to make the purpose of the invention, technical solutions and beneficial effects of the invention more clear and definite, the purpose of the invention, technical solutions and beneficial effects of the invention of the invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments are only used to explain and understand the present invention and are not used to limit the present invention.

[0061] like Figure 1 As shown, an embodiment of the present invention discloses a distributed unknown time-varying area multi-autonomous agent optimal coverage method. After the multi-autonomous agents accept the coverage control task in the unknown time-varying environment, each agent performs perception respectively, collects the environmental data of the current position, and then based on the distributed forgetting factor recursive least squares estimation algorithm model, the autonomous agent fuses the perceived environmental information with the environmental information obtained by communication interaction, and estimates the unknown and time-varying environmental density function. Then, each agent designs a corresponding second-order kinematic model-based coverage controller according to its estimated density function, generates control input, and executes corresponding actions according to the control input, thereby realizing the multi-autonomous agent optimal area coverage control in the unknown time-varying environment.

[0062] Specifically, in an embodiment of the present invention, a distributed unknown time-varying area multi-autonomous agent optimal coverage method mainly includes the following steps:

[0063] Step 1: Each subject senses and collects environmental data of the current location. The specific process is as follows:

[0064] For the autonomous agent i, i=1,…,N at time k, its position coordinates and collected environmental data are recorded as (p i,k ,φ(p i,k ), where N is the number of agents, p i,k ∈R n is the position vector of agent i at time k, n is the dimension of task area Q, φ(p i,k )∈R is the environmental data collected by agent i at time k, that is, the density function value.

[0065] In the embodiment of the present invention, the coordinate system model of the task area Q is:

[0066]

[0067] Where q is the position vector of any point in the n-dimensional task area Q.

[0068] Assume that in the mission area Q, the relationship between the environmental data detected and collected by the autonomous agent and the position vector can be represented by a density function model, which can be considered as a weighted function on the mission area to measure the relative importance of different locations in Q. For example, it can represent the probability of events occurring at different locations, the disaster situation at different locations in the disaster area, etc. Intuitively speaking, the goal of coverage control is to distribute autonomous agents more densely in areas with larger density function values, and to distribute fewer autonomous agents in areas with smaller density function values. In actual situations, the environmental density function is generally unknown a priori and may vary with time. For the unknown time-varying environmental density function φ(q,t), it is assumed that there is an ideal parameter vector satisfy:

[0069] φ(q,t)=K(q) T a(t)

[0070] Where K: is the basis function vector, a(t) is the unknown time-varying parameter, Represents an m-dimensional positive real vector. The above assumption is feasible, because as long as the appropriate basis function is selected, the above density function model can represent a large class of density functions. In addition, even if the density function cannot be accurately written in this form, a good approximation can be obtained by selecting appropriate parameters. The embodiment of the present invention selects the Gaussian radial basis function as K, which has a clear physical explanation and strong expressiveness.

[0071] Step 2: Each subject communicates and interacts with each other. Based on the distributed forgetting factor recursive least squares estimation algorithm model, the subject performs a fusion calculation on the perceived information and the information obtained by communication to estimate the unknown time-varying environment density function. The specific process is as follows:

[0072] The agent i, i = 1, ..., N uses the distributed forgetting factor recursive least squares estimation algorithm to estimate and update the parameter a(t) of the unknown time-varying density function, and then obtains the estimated density function in is the estimated parameter of agent i, specifically:

[0073] Assume that the agent i, i = 1, ..., N has obtained the position coordinates and collected environmental data at time k (p i,k ,φ(p i,k )), using this data and combining it with stored data information and M i (k-1) Perform local estimation of the unknown time-varying density function parameters and calculate the generated and Z i (k):

[0074]

[0075] in is the estimated parameter of the unknown time-varying density function of subject i at time k-1, λ is the forgetting factor, which needs to be a positive number close to 1, usually not less than 0.9;

[0076] The subjects communicate with each other, and the subject i obtains data information through communication interaction and j∈I N Perform information fusion and computational generation and

[0077]

[0078] Among them I N = {1,…,N} represents the set of agents, B = {b ij} N×N is a weighted adjacency matrix. When agents i and j can communicate and interact, b ij > 0, otherwise b ij = 0, and satisfies b ij =b ji and M i (0) Initialize to K M I, where K M is a positive real number, and the value interval can be selected as 106 ~10 10 , is the estimated parameter of the unknown time-varying density function of agent i at time k. Agent i obtains the estimated parameter according to the above estimation algorithm After that, the corresponding estimated density function can be further obtained

[0079] Step 3: Each subject designs a corresponding coverage controller based on its estimated density function, generates control input, and the self-agent performs corresponding actions based on the control input. The specific process is as follows:

[0080] The coverage control objective of the multi-agent system is to minimize the following coverage performance function:

[0081]

[0082] Where P = {p i , i=1,...,N} is the current position vector of N autonomous bodies, W={W i , i=1,...,N} is the arbitrary division of the convex task area Q by each subject, and W i is the area that autonomous agent i is responsible for. Assuming that the perception models of all autonomous agents are consistent, they are all omnidirectional perception models. f: is a non-decreasing differentiable function used to evaluate the perception performance of the autonomous subject on each point in the divided responsibility area. Without loss of generality, we select

[0083] Perform Voronoi partition on Q V = {V i ,i=1,...,N}:

[0084]

[0085] Furthermore, the Voronoi region V i The generalized quality Polar moment of inertia and centroid Given by:

[0086]

[0087] When selecting When , the coverage performance function under Voronoi partitioning can be written as:

[0088]

[0089] in It can represent the uncertainty of the perception of agent i for the midpoint q in the mission area. Combined with the coverage goal of deploying agents more densely in areas with large density function values, it is easy to understand that the coverage control goal of the multi-agent system is to minimize the coverage performance function H(P). In order to minimize the coverage performance function, the coverage performance function H(P) is calculated with respect to p i ,i∈I N Differentiate , and we get:

[0090]

[0091] In the embodiment of the present invention, the second-order kinematic model of the autonomous body satisfies:

[0092]

[0093]

[0094] where v i is the velocity vector of the ith autonomous body, u i ,i∈I N is the control input of the ith autonomous agent, i.e., the acceleration vector, k v is a positive constant.

[0095] Consider the Lyapunov function:

[0096]

[0097] Taking the time derivative of V we get:

[0098]

[0099] Therefore, V>0, According to LaSalle's invariant set principle, we know This means that when t→∞, there is v i =0 and Then the coverage performance function H reaches a local optimal value, at this time p i ,i∈I N Both are in the corresponding Voronoi region V i ,i∈I N The centroid of , that is, the centroid-Voroy configuration is a local optimal covering configuration. The Voronoi centroid is Then we have:

[0100]

[0101] When the environment information is unknown and time-varying, each subject estimates the density function based on its recursive least squares estimation algorithm using the distributed forgetting factor Calculate the corresponding Voronoi region V iEstimated generalized mass Estimated polar moment of inertia and estimated centroid

[0102]

[0103] When each agent is located at its corresponding estimated center of mass When , the coverage performance function has a local optimal value.

[0104] In order to make the coverage performance converge to the local optimum, the following coverage controller is designed based on the coverage control Lloyd algorithm:

[0105]

[0106] in is a uniformly positive gain matrix. The autonomous agent i generates the control input u according to the covering controller i Execute the corresponding overwriting action.

[0107] The following is a simulation experiment of the optimal coverage method of distributed unknown time-varying area multi-agent designed by the present invention. Assume that there are 20 agents and the square task area is divided into 3×3 grids. Density function φ(q,t)=K(q) T a(t) is parameterized by a set of Gaussian radial basis functions, and K is chosen to be a vector of 9 Gaussian radial basis functions, K = [K(1), ..., K(9)] T , where K(j), j = 1, ..., 9 is given by:

[0108]

[0109] where δ j = 0.18 is an adjustable parameter, the task area is divided into 3×3 grids, μ j ,j=1,…,9 is selected as the center of each grid. In the initial coverage configuration of the autonomous system, the true density function parameter vector is selected as a=[100,a min ,a min ,a min ,a min ,a min ,a min ,a min ,100] T , where a min =0.1, which leads to a bimodal distribution of the density function in the task area. The total coverage time of the simulation experiment is set to 60s, and the iteration step is 0.1s. The present invention considers the time-varying density function and sets the parameter vector of the true density function to change to a=[a min ,a min,100,a min ,a min ,a min 100,a min ,a min ] T , changes to a=[a min ,100,a min ,a min ,a min ,a min ,a min ,100,a min ] T The estimated parameters of agent i are Initialized to [a min ,a min ,a min ,a min ,a min ,a min ,a min ,a min ,a min ] T The forgetting factor in the parameter estimation algorithm is set to λ = 0.96, and the uniformly positive gain matrix in the covering control law is set to k v =1.

[0110] When the environmental density function of the task area is unknown and time-varying, each agent needs to continuously measure and collect environmental information at its location to estimate the density function. Therefore, the initial position of the agent will have a certain impact on the final coverage configuration. Intuitively, when the agents are more dispersed in the task area, the coverage effect will be relatively good, thanks to the fact that the dispersed agents can measure environmental information at different locations in the task area, while the agents that are relatively concentrated in the task area can only measure local environmental information in the concentrated area. Simulation experiments are conducted for the cases where multiple agents are initially concentrated and initially dispersed in the task area.

[0111] The initial positions of the autonomous agents in the case of aggregated initial distribution and dispersed initial distribution are as follows: Figure 2 and Figure 6 As shown in the figure, * represents the Gaussian center of the density function, that is, the peak point of the density function. The step 1 to step 3 of the coverage method are continuously iterated until the condition for the termination of the iteration is reached, thereby completing the coverage task in the unknown time-varying environment. The condition for the termination of the iteration can be set to the maximum number of iterations or the moving distance of the autonomous subject is lower than a certain threshold. Figure 3 and Figure 7The coverage configurations of the multi-agent system at 15s, 30s and 60s (final moment) are given under the conditions of aggregated initial distribution and dispersed initial distribution. The coverage control results of the multi-agent system under the conditions of aggregated initial distribution and dispersed initial distribution are shown in Figure 2. Figure 4 and Figure 8 As shown in the figure, the motion trajectory of the autonomous body is depicted, and “+” indicates the final position of the autonomous body. Figure 4 and Figure 8 It can be seen that the multiple autonomous bodies with clustered initial distribution and dispersed initial distribution have reached the local optimal coverage configuration after using the coverage method described in the present invention. A large number of autonomous bodies are densely distributed in areas with larger density function values, and the coverage effect is significant. Figure 5 and Fig. 9 The average parameter errors of multiple agents in the case of clustered initial distribution and dispersed initial distribution are given respectively. As can be seen from the figure, the unknown time-varying density function estimation effect is good, and the parameter convergence speed is very fast, which verifies the effectiveness and excellent performance of the distributed forgetting factor recursive least squares estimation algorithm described in step 2.

[0112] From the above simulation experiment results, it can be seen that based on the distributed unknown time-varying area multi-agent optimal coverage method designed by the present invention, the multi-agent system efficiently estimates the unknown time-varying density function in both the case of aggregated initial distribution and dispersed initial distribution, and effectively realizes the multi-agent optimal area coverage control in the unknown environment. The satisfactory coverage effect verifies the effectiveness and excellent performance of the method designed by the present invention.

[0113] Based on the same inventive concept, an embodiment of the present invention discloses a distributed unknown time-varying area multi-autonomous agent optimal coverage system, wherein the task area contains multiple autonomous agents that can communicate and interact, and each agent includes the following modules: an environmental density estimation module, which is used for the autonomous agent to fuse the perceived environmental information with the environmental information obtained by communication based on a distributed forgetting factor recursive least squares estimation algorithm model, and estimate the unknown time-varying environmental density function; an optimal coverage control module, which is used for the autonomous agent to design a corresponding coverage controller based on a second-order kinematic model according to its estimated density function, generate control input, and the autonomous agent performs corresponding actions according to the control input.

[0114] Based on the same inventive concept, an embodiment of the present invention discloses a distributed unknown time-varying area multi-autonomous agent optimal coverage system, which includes multiple autonomous agents that can communicate and interact in the task area; the autonomous agent includes a sensor for collecting environmental data, a memory, a processor, and a computer program stored in the memory and run on the processor, and when the computer program is loaded into the processor, it implements the distributed unknown time-varying area multi-autonomous agent optimal coverage method.

[0115] Based on the same inventive concept, an embodiment of the present invention discloses a computer system, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the computer program is loaded into the processor, the distributed unknown time-varying area multi-autonomous agent optimal coverage method is implemented.

[0116] It is worth noting that the above is only a more preferred embodiment of the present invention, which is only used to verify and explain the present invention, and does not constitute a limitation on the present invention. Various changes, equivalent substitutions and modifications made by relevant technical personnel without departing from the technical ideas of the present invention are all included in the protection scope of the present invention.

Claims

1. A distributed unknown time-varying area multi-agent optimal coverage method, characterized in that: The following steps are involved: Based on the distributed forgetting factor recursive least squares estimation algorithm model, each subject performs a fusion calculation on the perceived environmental information and the environmental information obtained by communication to estimate the unknown time-varying environmental density function; Each subject designs a corresponding second-order kinematic model-based covering controller according to its estimated density function, generates control input, and the self-agent performs corresponding actions according to the control input; Assume the unknown time-varying ambient density function φ(q,t) = K(q) T a(t), where q∈Q is any point in the task area Q, K is the basis function vector, and a(t) is an unknown time-varying parameter; the agent i, i=1,…,N uses the distributed forgetting factor recursive least squares estimation algorithm to estimate and update the parameter a(t) of the unknown time-varying density function, and then obtains the estimated density function in is the estimated parameter of agent i, N is the number of agents, specifically: For the autonomous agent i, i = 1, ..., N, the position coordinates at time k and the collected environmental data (p i,k ,φ(p i,k )) and combined with the stored data information and M i (k-1) Perform local estimation of unknown time-varying density function parameters and calculate the generated and Z i (k): in is the estimated parameter of the unknown time-varying density function of subject i at time k-1, λ is the forgetting factor, 0.9≤λ<1; The subjects communicate with each other, and the subject i obtains data information through communication interaction. and j∈I N Perform information fusion and computational generation and Among them I N = {1,…,N} represents the set of agents, B = {b ij } N×N is a weighted adjacency matrix. When agents i and j can communicate and interact, b ij > 0, otherwise b ij = 0, and satisfies b ij =b ji and M i (0) Initialize to K M I, where K M is a positive real number, and the value interval is 10 6 ~10 10 ; Perform Voronoi partition on Q V = {V i ,i=1,...,N}: Voronoi region V of subject i i The generalized quality Polar moment of inertia and centroid for: When selecting When , the coverage performance function under Voronoi partitioning is recorded as: in It can represent the perceived uncertainty of the self-agent i about the point q in the task area; The second-order kinematic model of the autonomous body satisfies: where v i is the velocity vector of the ith autonomous body, u i is the control input of the ith autonomous agent, i.e., the acceleration vector, k v is a positive constant; When the environmental information is unknown and time-varying, each subject estimates the density function based on its recursive least squares estimation algorithm using the distributed forgetting factor Calculate the corresponding Voronoi region V i Estimated generalized mass Estimated polar moment of inertia and estimated centroid When each agent is located at its corresponding estimated center of mass When , the coverage performance function has a local optimal value; Design a coverage controller based on the coverage control Lloyd algorithm to make the coverage performance converge to the local optimum: in is a uniformly positive definite gain matrix.

2. The distributed unknown time-varying area multi-agent optimal coverage method according to claim 1 is characterized in that: Each subject designs the corresponding coverage controller based on the coverage control Lloyd algorithm according to its estimated density function.

3. The distributed unknown time-varying area multi-agent optimal coverage method according to claim 2 is characterized in that: The coverage control objective of the multi-agent system is to minimize the following coverage performance function: Where P = {p i , i=1,...,N} is the current position vector of N autonomous bodies, W={W i , i=1,...,N} is the arbitrary division of the convex task area Q by each subject, and W i is the area that subject i is responsible for, f: is a non-decreasing differentiable function used to evaluate the perception performance of the autonomous agent on each point in the divided responsibility area.

4. A distributed unknown time-varying area multi-agent optimal coverage system, characterized in that: The task area contains multiple autonomous entities that can communicate and interact, and each entity has the following modules: The environment density estimation module is used by the autonomous agent to fuse the perceived environment information with the environment information obtained through communication based on the distributed forgetting factor recursive least squares estimation algorithm model, and estimate the unknown time-varying environment density function; The optimal coverage control module is used for the autonomous agent to design a corresponding coverage controller based on the second-order kinematic model according to its estimated density function, generate control input, and the autonomous agent performs corresponding actions according to the control input; In the environment density estimation module, the unknown time-varying environment density function φ(q,t)=K(q) is assumed T a(t), where q∈Q is any point in the task area Q, K is the basis function vector, and a(t) is an unknown time-varying parameter; the agent i, i=1,…,N uses the distributed forgetting factor recursive least squares estimation algorithm to estimate and update the parameter a(t) of the unknown time-varying density function, and then obtains the estimated density function in is the estimated parameter of agent i, N is the number of agents, specifically: For the autonomous agent i, i = 1, ..., N, the position coordinates at time k and the collected environmental data (p i,k ,φ(p i,k )) and combined with the stored data information and M i (k-1) Perform local estimation of unknown time-varying density function parameters and calculate the generated and Z i (k): in is the estimated parameter of the unknown time-varying density function of subject i at time k-1, λ is the forgetting factor, 0.9≤λ<1; The subjects communicate with each other, and the subject i obtains data information through communication interaction and j∈I N Perform information fusion and computational generation and Among them I N = {1,…,N} represents the set of agents, B = {b ij } N×N is a weighted adjacency matrix. When agents i and j can communicate and interact, b ij > 0, otherwise b ij = 0, and satisfies b ij =b ji and M i (0) Initialize to K M I, where K M is a positive real number, and the value interval is 10 6 ~10 10 ; Perform Voronoi partition on Q V = {V i ,i=1,…,N}: Voronoi region V of subject i i The generalized quality Polar moment of inertia and centroid for: When selecting When , the coverage performance function under Voronoi partitioning is recorded as: in It can represent the perceived uncertainty of the self-agent i about the point q in the task area; The second-order kinematic model of the autonomous body satisfies: where v i is the velocity vector of the ith autonomous body, u i is the control input of the ith autonomous agent, i.e., the acceleration vector, k v is a positive constant; When the environmental information is unknown and time-varying, each subject estimates the density function based on its recursive least squares estimation algorithm using the distributed forgetting factor Calculate the corresponding Voronoi region V i Estimated generalized mass Estimated polar moment of inertia and estimated centroid When each agent is located at its corresponding estimated center of mass When , the coverage performance function has a local optimal value; Design a coverage controller based on the coverage control Lloyd algorithm to make the coverage performance converge to the local optimum: in is a uniformly positive definite gain matrix.

5. The distributed unknown time-varying area multi-agent optimal coverage system according to claim 4 is characterized in that: Each subject designs the corresponding coverage controller based on its estimated density function and the Lloyd algorithm for coverage control; the coverage performance function under Voronoi partitioning is: in It can represent the perceived uncertainty of the self-agent i about the point q in the task area; The second-order kinematic model of the autonomous body satisfies: where v i is the velocity vector of the ith autonomous body, u i is the control input of the ith autonomous agent, i.e., the acceleration vector, k v is a positive constant; when each autonomous body is located at its corresponding estimated center of mass When , the coverage performance function has a local optimal value; in is a uniformly positive definite gain matrix.

6. A distributed unknown time-varying area multi-agent optimal coverage system, characterized in that: The task area includes multiple autonomous agents that can communicate and interact with each other; the autonomous agent includes a sensor for collecting environmental data, a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is loaded into the processor, the distributed unknown time-varying area multi-autonomous agent optimal coverage method according to any one of claims 1 to 3 is implemented.

7. A computer system comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the computer program is loaded into a processor, it implements the distributed unknown time-varying area multi-autonomous agent optimal coverage method according to any one of claims 1-3.

Images