Distributed coverage control method, storage medium and equipment
Through the distributed coverage control method, combined with gradient descent method and energy enhancement strategy, the state update of the agent is optimized, and the local optimal problem in non-convex area coverage is solved, and environmental monitoring performance and computing efficiency are improved.
Patent Information
- Application Number
- CN202411021623.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-29
- Publication Date
- 2024-11-08
AI Technical Summary
When covered by non-convex areas, the gradient descent method causes the system to fall into a local optimal state, making it difficult to effectively monitor the environment. The existing global optimization technology is computationally expensive and is not suitable for online optimization.
A distributed coverage control method is proposed. By constructing the objective function of the agent area coverage system, combining the gradient descent method and energy enhancement strategy, the state update of the agent is optimized, and the enhanced gradient and charging process are introduced to realize environmental disaster monitoring in non-convex areas .
Effectively avoid local optimal state, improve the system's coverage performance in non-convex environments, simplify calculations, and is suitable for online execution, and enhance the coverage performance and environmental monitoring capabilities of the agent.
Smart Images

Figure CN118916784A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of environmental monitoring, and in particular to a distributed coverage control method, storage medium, and equipment. Background Art
[0002] The increasing global warming has led to frequent extreme weather events, which in turn have triggered a series of natural or man-made disasters, seriously threatening human survival and development. In recent years, disasters such as mudslides, forest fires, and landslides have occurred frequently in mountainous areas, causing huge economic losses and casualties to society. However, with the rapid development of sensor technology, communication technology, and network technology, multi-robot systems have shown broad application prospects in the fields of regional monitoring, environmental monitoring, and disaster relief. Through these technical means, we can monitor the occurrence of disasters in advance, so as to take preventive measures and reduce losses and casualties.
[0003] However, when covering non-convex areas, the gradient descent method will cause the system to fall into a local optimal state, and thus fail to monitor the environment well. In this case, scholars use global optimization techniques, such as simulated annealing, genetic algorithms, or particle swarm algorithms. The common point of these methods is to introduce random elements in the control process so that the system escapes the equilibrium point and finds a better coverage position. However, these methods are computationally intensive and are usually not suitable for online optimization. Summary of the invention
[0004] The purpose of the present invention is to solve the problem that the system is prone to fall into a local optimal state in the current non-convex environment coverage, and propose a distributed coverage control method, comprising the following steps:
[0005] S1. Construct the objective function of the intelligent agent regional coverage system based on the probability distribution function of environmental disasters occurring in non-convex areas monitored by the intelligent agent;
[0006] S2. Obtain the local objective function of the agent according to the objective function, update the gradient state of the agent according to the gradient descent method, and obtain the gradient of the local objective function of the agent;
[0007] S3, optimize the objective function using the energy enhancement strategy according to the gradient of the local objective function of the agent;
[0008] S4. According to the optimization objective function obtained in S3, the control input of the intelligent agent regional coverage system is obtained to realize environmental disaster monitoring in non-convex areas.
[0009] Furthermore, the probability distribution function of the agent's detection environment is:
[0010] φ(x)=φ0-β||x-x0||
[0011] Among them, φ(x) is the probability distribution function of the agent's detection environment, φ0 represents a constant, β is a positive constant, ||x-x0|| represents the Euclidean distance between x and x0, x represents the position of the agent, and x0 represents a constant.
[0012] The objective function of the intelligent agent area coverage system is:
[0013] H(s)=∫ Ω φ(x)P(x,s)dx
[0014] Where H(s) is the objective function of the agent area coverage system, s represents the agent state, Ω represents the covered area, and P(x, s) represents the probability that at least one agent in the environment can recognize point x;
[0015]
[0016] Among them, p(x, s i ) represents the probability that the i-th agent in the environment can recognize point x, n is the number of agents, s i Represents the state of the ith agent.
[0017] Furthermore, according to the objective function, the local objective function of the agent is expressed as:
[0018]
[0019] in, represents the state of the neighboring agent of the ith agent at the kth step; Indicates the status is is the local objective function of the ith agent; Indicates the status is The probability that at least one agent can recognize point x when
[0020] The specific method of updating the agent state using the gradient descent method is as follows:
[0021] S i,k+1 =s i,k +α i,k G i,k
[0022]
[0023] Among them, s i,k+1 represents the state of the i-th agent at the k+1th step, s i,k represents the state of the i-th agent at the k-th step, α i,k represents the step length of the kth step of the local objective function of the ith agent, G i,k represents the gradient of the local objective function of the ith agent at the kth step.
[0024] Furthermore, the energy enhancement strategy includes mode B, mode N, and mode C; when the gradient G of the local objective function of the agent i is zero, the local objective function H i To reach equilibrium, proceed to mode B: the gradient G i Replaced with enhanced gradient until is zero. Perform mode N: Mode N uses gradient G i Until G i is zero, and the local objective function is the enhanced local objective function Will and H i The largest one is denoted as H i If the energy of the intelligent body is greater than the charging standard energy, continue with mode B; if the energy of the intelligent body is less than or equal to the charging standard energy, execute mode C: go to the charging pile for charging. The local objective function during the charging process is Constantly H i The largest one is denoted as H i , until the agent energy is greater than the charging standard energy, continue with mode B.
[0025] Furthermore, an enhancement function is constructed to represent the enhanced gradient, and the gradient of the local objective function of the i-th agent is expressed as G i =[G iX , G iY ] T ,in,
[0026]
[0027]
[0028] Among them, G iX represents the gradient in the X direction of the local objective function of the ith agent, G iY represents the gradient in the Y direction of the local objective function of the ith agent, represents the state of the neighboring agent of the ith agent, Indicates the status is The local objective function of the ith agent is s iX Indicates state s i The state component in the X direction, s iY Indicates state s i The state component in the Y direction, V i represents the monitoring area of the ith agent, M i represents the neighbor set of the ith agent, p i (x,si ) indicates the state is s i The detection probability of the i-th agent for point x in the environment at time ;
[0029] G iX The formula is rewritten as:
[0030] in,
[0031] The enhancement function is constructed in the following form:
[0032]
[0033] The enhanced gradient is expressed as:
[0034]
[0035] in, γ1, γ2 are two positive constants, k is a constant, Φ i (x) represents the probability that the i-th agent does not detect the event occurring at x.
[0036] Further,
[0037] According to the optimization objective function obtained from S3, the control input of the intelligent area coverage system is:
[0038] Find the nominal input to the optimization objective function:
[0039]
[0040] Among them, α is a constant, represents the nominal input of the optimization objective function;
[0041] The control input of the final intelligent area coverage system is obtained through quadratic programming:
[0042]
[0043] st[E min ,E max ]
[0044] Among them, u * represents the control input of the final intelligent regional coverage system, u represents the control input of the intelligent regional coverage system, δ=[δ1,...,δ N ] T is a vector of relaxation parameters, the matrix K = diag(κ i ) is the diagonal positive definite weight matrix of δ, E min Indicates the minimum value of the intelligent body’s power, E maxIndicates the maximum value of the intelligent body's power.
[0045] The present invention further proposes a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned distributed coverage control method is implemented.
[0046] The present invention also proposes an electronic device, including a processor and a memory, wherein the processor and the memory are interconnected, wherein the memory is used to store a computer program, the computer program includes computer-readable instructions, and the processor is configured to call the computer-readable instructions to execute the above-mentioned distributed coverage control method.
[0047] The beneficial effects brought by the technical solution provided by the present invention are:
[0048] The solution of the present invention defines a new auxiliary local objective function, which changes the local objective function of the agent when equilibrium is reached, and is indirectly completed by converting the local gradient of the local objective into a new enhanced gradient. When the system reaches a new equilibrium point under the enhanced gradient, the original objective function is replaced by the objective function with the maximum value of the objective function of different equilibrium points. Compared with the random perturbation gradient, the enhancement function provides a systematic method to force each agent to move in a carefully selected direction, which is based on the structural characteristics of the objective function and the understanding of the feasible space and the current state of the agent, and further explores the feasible space to find a better target point. And the charging process of the agent is added to the algorithm, so that the system is equivalent to exploring the environment in the process of going to charge, so as to better find the target point. Compared with the previous method, the above-mentioned new coverage method introduces an enhancement function, and the calculation is simple and can be executed online, and combined with the charging process, it greatly improves the coverage performance of the agent, and can be used for persistent environmental monitoring in non-convex environments. BRIEF DESCRIPTION OF THE DRAWINGS
[0049] Figure 1 is a flow chart of a distributed coverage control method according to an embodiment of the present invention;
[0050] Figure 2 is a block diagram of an energy enhancement strategy according to an embodiment of the present invention;
[0051] Figure 3 is a block diagram of an electronic device in an exemplary embodiment of the present invention;
[0052] Figure 4 is an image sequence recorded during an environmental detection simulation experiment without using an energy enhancement strategy in an embodiment of the present invention, wherein: Figure 4(a) is the initial position of the agent, (b) is the system covering the target area according to the gradient descent method, (c) is the charging of one of the agents, and (d) is the completion of the charging of the agent and the finding of a target point according to the gradient descent method;
[0053] Figure 5 is an image sequence recorded during an environmental detection simulation experiment using an energy enhancement strategy in an embodiment of the present invention, wherein: Figure 5 (a) is the initial position of the agent, (b) is the target position found by the system using the enhancement function after the system covers the equilibrium point according to the gradient descent method, (c) is the charging of the agent, and (d) is the target position covered by the system after using the energy enhancement strategy of the present invention;
[0054] Figure 6 is a coverage performance curve of the embodiment of the present invention without using the energy enhancement strategy;
[0055] Figure 7 It is a coverage performance curve of the embodiment of the present invention using the energy enhancement strategy. DETAILED DESCRIPTION
[0056] To make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described below with reference to the accompanying drawings.
[0057] The flowchart of the distributed coverage control method according to the embodiment of the present invention is as follows: Figure 1 , specifically including the following steps:
[0058] S1. Construct the objective function of the intelligent agent regional coverage system based on the probability distribution function of environmental disasters occurring in non-convex areas monitored by the intelligent agent.
[0059] The agent model and environment model are designed, the agent is modeled using an affine control dynamics system, and the probability distribution function φ(x) of the environment is constructed based on the agent model. The probability distribution function enables the agent to adaptively update its position and collaboratively monitor a given uncertainty area.
[0060] The probability distribution function of the agent's detection environment is:
[0061] φ(x)=φ0-β||x-x0||
[0062] Among them, φ(x) is the probability distribution function of the agent's detection environment, φ0 represents a constant, β is a positive constant, ||x-x0|| represents the Euclidean distance between x and x0, x represents the position of the agent, and x0 represents a constant.
[0063] In the embodiment of the present invention, in a more specific scenario, such as a landslide monitoring system, φ(x) is the probability distribution of landslide occurrence in the area. When the intelligent agent monitors the area, it can identify the probability of an event occurring at point x with a certain probability, expressed as p(x, s i ). The intelligent agent in the embodiment of the present invention refers to a robot, wherein s i represents the position of robot i. Let P(x, s) represent the probability that at least one robot in the environment can recognize point q, expressed as:
[0064]
[0065] The objective function H(s) represents the coverage performance index of the robot in a certain area Ω:
[0066] H(s)=∫ Ω φ(x)P(x,s)dx
[0067] Among them, H(s) is the objective function of the robot area coverage system. The larger the value of H(s), the better the coverage performance of the system. s represents the robot state, Ω represents the covered area, P(x, s) represents the probability that at least one robot in the environment can recognize point x, and p(x, s i ) represents the probability that the i-th agent in the environment can recognize point x, n is the number of robots, s i Represents the state of the i-th robot.
[0068] S2. Obtain the local objective function of the agent according to the objective function, update the gradient state of the agent according to the gradient descent method, and obtain the gradient of the local objective function of the agent.
[0069] According to the objective function, the local objective function of the agent is expressed as:
[0070]
[0071] in, represents the state of the neighboring agent of the ith agent at the kth step; Indicates the status is is the local objective function of the ith agent; Indicates the status is The probability that at least one agent can recognize point x.
[0072] The specific method of updating the agent state using the gradient descent method is as follows:
[0073] s i,k+1 =s i,k +α i,k G i,k
[0074]
[0075] Among them, s i,k+1 represents the state of the i-th agent at the k+1th step, s i,k represents the state of the i-th agent at the k-th step, α i,k represents the step length of the kth step of the local objective function of the ith agent, G i,k represents the gradient of the local objective function of the ith agent at the kth step.
[0076] Construct an enhancement function to represent the enhanced gradient, and express the gradient of the local objective function of the i-th agent as G i =[G iX , G iY ] T ,in,
[0077]
[0078] Among them, G iX represents the gradient in the X direction of the local objective function of the ith agent, G iY represents the gradient in the Y direction of the local objective function of the ith agent, represents the state of the neighboring agent of the ith agent, Indicates the status is The local objective function of the ith agent is s iX Indicates state s i The state component in the X direction, s iY Indicates state s i The state component in the Y direction, V i represents the monitoring area of the ith agent, represents the neighbor set of the ith agent, and p i (x,s i ) indicates the state is s i The detection probability of the i-th agent for point x in the environment at time .
[0079] G iX The formula is rewritten as:
[0080] in,
[0081] The enhancement function is constructed in the following form:
[0082]
[0083] The enhanced gradient is expressed as:
[0084]
[0085] in, γ1, γ2 are two positive constants, k is a constant, Φ i (x) represents the probability that the i-th agent does not detect the event occurring at x.
[0086] The embodiment of the present invention proposes an energy enhancement function design scheme, which is This method reduces the weight value of the point x covered by the agent and increases the weight value of the uncovered point, thereby changing the objective function of the agent and making the agent move toward the uncovered point.
[0087] S3. Optimize the objective function using an energy enhancement strategy based on the gradient of the agent's local objective function.
[0088] Energy Enhancement Strategy Reference Figure 2 Including mode B, mode N, mode C; when the gradient G of the local objective function of the agent i is zero, the local objective function H i To reach equilibrium, proceed to mode B: the gradient G i Replaced with enhanced gradient until is zero, and mode N is performed: mode N uses the gradient G i Until G i is zero, and the local objective function is the enhanced local objective function Will and H i The largest one is denoted as H i If the energy of the intelligent body is greater than the charging standard energy, continue with mode B; if the energy of the intelligent body is less than or equal to the charging standard energy, execute mode C: go to the charging pile for charging. The local objective function during the charging process is Constantly H i The largest one is denoted as H i , until the agent energy is greater than the charging standard energy, continue with mode B.
[0089] Whenever the newly defined local objective function H i When equilibrium is reached, the local gradient G i is zero, changing the local objective function to an enhanced objective function Use non-zero boost gradient Instead of the original gradient G i , which forces the agent to move in the direction determined by the reinforcement function and further explore the feasible space. When a new equilibrium point is reached, the original objective function is updated to the objective function with the largest objective function value among all equilibrium points. Then the gradient-based algorithm converges to a new (possibly better, but not worse) equilibrium point.
[0090] S4. According to the optimization objective function obtained in S3, the control input of the intelligent agent regional coverage system is obtained to realize environmental disaster monitoring in non-convex areas.
[0091] Specifically:
[0092] Find the nominal input to the optimization objective function:
[0093]
[0094] Among them, α is a constant, represents the nominal input of the optimization objective function;
[0095] The control input of the final intelligent area coverage system is obtained through quadratic programming:
[0096]
[0097] st[E min ,E max ]
[0098] Among them, u * represents the control input of the final intelligent agent area coverage system, u represents the control input of the intelligent agent area coverage system, δ=[δ1,...,δ N ] T is a vector of relaxation parameters, the matrix κ=diag(κ i ) is the diagonal positive definite weight matrix of δ, E min Indicates the minimum value of the intelligent body’s power, E max Indicates the maximum value of the intelligent body's power.
[0099] According to the control input of the intelligent agent area coverage system, the intelligent agent can monitor environmental disasters in non-convex areas.
[0100] In an exemplary embodiment, a computer-readable storage medium is included, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned distributed coverage control method is implemented.
[0101] See also Figure 3 In an exemplary embodiment, an electronic device is also included, including at least one processor, at least one memory, and at least one communication bus.
[0102] In order to verify the effectiveness of the method of the present invention, four homogeneous agents are arranged vertically. The image sequence recorded during the environment detection simulation experiment without the energy enhancement strategy in the embodiment of the present invention is used as a reference. Figure 4The 4 black dots in the figure represent 4 intelligent agents. The dotted circle around the energy body represents the monitoring range of the intelligent agent. The solid circle represents the charging pile. The value bar on the right side of the figure represents the probability distribution of the intelligent agent's monitoring environment. The larger the probability distribution value, the greater the weight, and the greater the possibility of an accident. Figure 4 (a) is the initial position of the agent, where the charging field is arranged at the four corners of the figure, (b) is the system covering the target area according to the gradient descent method, and it can be seen that the system is trapped in the local optimum. (c) is charging one of the agents. d is the completion of the charging of the agent, and then a target point is found according to the gradient descent method. Figure 4 As shown in (a), each iteration takes 0.01 seconds. The circular areas in the four corners represent the charging field. Figure 4 No energy enhancement strategy is added. It can be seen that three agents are covering the same high-density area and are trapped in the local optimum.
[0103] The image sequence recorded during the environment detection simulation experiment using the energy enhancement strategy in the embodiment of the present invention is referenced Figure 5 The 4 black dots in the figure represent 4 intelligent agents. The dotted circle around the energy body represents the monitoring range of the intelligent agent. The solid circle represents the charging pile. The value bar on the right side of the figure represents the probability distribution of the intelligent agent's monitoring environment. The larger the probability distribution value, the greater the weight, and the greater the possibility of an accident. Figure 5 In the process, the environment setting and the location and Figure 4 same, Figure 5 (a) is the initial position of the agent, and (b) is the target position found by the enhancement function after the system falls into the equilibrium point according to the gradient descent method. It can be seen that its coverage performance is better than Figure 4 (b) and (c) show the intelligent agent charging, and (d) is the coverage target location after the system applies the EBS strategy (the energy enhancement strategy of the present invention). Figure 5 (b) shows that the system finds a better balance point through the enhancement function. Figure 6 and Figure 7 From the comparison of the coverage performance curves, it can be seen that the coverage performance of the intelligent agent is obviously improved after using the energy enhancement strategy.
[0104] Through the simulation results shown above, we have verified that the intelligent agent is able to find a better target point and perform the coverage task, thereby also verifying the effectiveness of the control architecture designed by the present invention.
[0105] Wherein, a computer program is stored in the memory, and the computer program includes computer-readable instructions. The processor calls the computer-readable instructions stored in the memory through the communication bus to execute the above-mentioned distributed coverage control method.
[0106] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A distributed coverage control method, characterized in that: The following steps are involved: S1. Construct the objective function of the intelligent agent regional coverage system based on the probability distribution function of environmental disasters occurring in non-convex areas monitored by the intelligent agent; S2. Obtain the local objective function of the agent according to the objective function, update the gradient state of the agent according to the gradient descent method, and obtain the gradient of the local objective function of the agent; S3, optimize the objective function using the energy enhancement strategy according to the gradient of the local objective function of the agent; S4. According to the optimization objective function obtained in S3, the control input of the intelligent agent regional coverage system is obtained to realize environmental disaster monitoring in non-convex areas.
2. A distributed coverage control method according to claim 1, characterized in that: The probability distribution function of the agent's detection environment is: φ(x)=φ0-β‖x-x0‖ Among them, φ(x) is the probability distribution function of the agent's detection environment, φ0 represents a constant, β is a positive constant, ‖x-x0‖ represents the Euclidean distance between x and x0, x represents the position of the agent, and x0 represents a constant; The objective function of the intelligent area coverage system is: H(s)=∫ Ω φ(x)P(x,s)dx Where H(s) is the objective function of the agent area coverage system, s represents the agent state, Ω represents the covered area, and P(x,s) represents the probability that at least one agent in the environment can recognize point x; Among them, p(x,s i ) represents the probability that the i-th agent in the environment can recognize point x, n is the number of agents, s i Represents the state of the ith agent.
3. A distributed coverage control method according to claim 2, characterized in that: According to the objective function, the local objective function of the agent is expressed as: in, represents the state of the neighboring agent of the ith agent at the kth step; Indicates the status is is the local objective function of the ith agent; Indicates the status is The probability that at least one agent can recognize point x when The specific method of updating the agent state using the gradient descent method is as follows: s i,k+1 =s i,k +a i,k G i,k Among them, s i,k+1 represents the state of the i-th agent at the k+1th step, s i,k represents the state of the i-th agent at the k-th step, α i,k represents the step length of the kth step of the local objective function of the ith agent, G i,k represents the gradient of the local objective function of the ith agent at the kth step.
4. A distributed coverage control method according to claim 3, characterized in that: Energy enhancement strategies include mode B, mode N, and mode C; when the gradient G of the local objective function of the agent i is zero, the local objective function H i To reach equilibrium, proceed to mode B: the gradient G i Replaced with enhanced gradient until Zero; proceed to mode N: Mode N uses gradient G i Until G i is zero, and the local objective function is the enhanced local objective function Will and H i The largest one is denoted as H i If the energy of the intelligent body is greater than the charging standard energy, continue with mode B; if the energy of the intelligent body is less than or equal to the charging standard energy, execute mode C: go to the charging pile for charging. The local objective function during the charging process is Constantly H i The largest one is denoted as H i , until the agent energy is greater than the charging standard energy, continue with mode B.
5. A distributed coverage control method according to claim 4, characterized in that: Construct an enhancement function to represent the enhanced gradient, and express the gradient of the local objective function of the i-th agent as G i =[G iX ,G iY ] T ,in, Among them, G iX represents the gradient in the X direction of the local objective function of the ith agent, G iY represents the gradient in the Y direction of the local objective function of the ith agent, represents the state of the neighboring agent of the ith agent, Indicates the status is The local objective function of the ith agent is s iX Indicates state s i The state component in the X direction, s iY Indicates state s i The state component in the Y direction, V i represents the monitoring area of the ith agent, M i represents the neighbor set of the ith agent, p i (x,s i ) indicates the state is s i The detection probability of the i-th agent for point x in the environment at time ; G iX The formula is rewritten as: in, The enhancement function is constructed in the following form: The enhanced gradient is expressed as: in, γ1, γ2 are two positive constants, k is a constant, Φ i (x) represents the probability that the i-th agent does not detect the event occurring at x.
6. A distributed coverage control method according to claim 4, characterized in that: According to the optimization objective function obtained from S3, the control input of the intelligent area coverage system is: Find the nominal input to the optimization objective function: Among them, α is a constant, represents the nominal input of the optimization objective function; The control input of the final intelligent area coverage system is obtained through quadratic programming: s.t.[E min ,E max ] Among them, u * represents the control input of the final intelligent regional coverage system, u represents the control input of the intelligent regional coverage system, δ=[δ1,...,δ N ] T is a vector of relaxation parameters, the matrix K = diag(κ i ) is the diagonal positive definite weight matrix of δ, E min Indicates the minimum value of the intelligent body’s power, E max Indicates the maximum value of the intelligent body's power.
7. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.
8. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the processor and the memory are interconnected, wherein the memory is used to store a computer program, the computer program comprises computer-readable instructions, and the processor is configured to call the computer-readable instructions to execute the method according to any one of claims 1 to 6.
