Multi-agent distributed online load segmentation method and system for ground disaster monitoring
By constructing a mathematical model of area scanning coverage and agent partition control law, combined with neural network algorithm, the problem that multi-agent systems cannot obtain global information in geodisaster monitoring is solved, and the quality of geodisaster monitoring and monitoring efficiency is improved. .
Patent Information
- Application Number
- CN202410960092.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-17
- Publication Date
- 2024-11-08
AI Technical Summary
In geological disaster monitoring, due to the limited perception range, multi-agent systems cannot obtain global information about disaster events, making it difficult to evenly allocate work in areas with uncertain probability of disaster events, thus affecting the monitoring quality.
The distributed online load segmentation method of multi-agents is adopted to construct a mathematical model of area scanning coverage and the control law of the agent partition along the short side direction, combined with neural network algorithm, the probability information of the geodisaster distribution is obtained, and the load equalization between agents is realized. Task allocation to improve monitoring efficiency.
In areas with uncertain distribution probability, uniform segmentation and task allocation of sub-region loads are achieved, which improves the quality of geological disaster monitoring and improves monitoring efficiency.
Smart Images

Figure CN118917533A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of geological disaster monitoring, and more specifically, to a multi-agent distributed online load segmentation method and system for geological disaster monitoring. Background Art
[0002] In recent years, as the problems that intelligent agents need to solve have become increasingly complex, scholars have gradually shifted from studying single agents to studying multi-agent systems. Multi-agent systems are often used in scenarios such as environmental monitoring, regional coverage, and unmanned search and rescue due to their distributed, flexible, and autonomous characteristics. For multi-agent monitoring systems, if the distribution information of geological disasters in each region is known in advance, it is easy to assign tasks to each agent. However, in practical applications, due to the limited perception range of the agent, it is usually impossible to obtain global information on the probability of disaster events. Therefore, how to evenly distribute work in areas with uncertain probability of disaster events to maximize the quality of monitoring is an important research direction for multi-agent systems. Summary of the invention
[0003] The purpose of the present invention is to provide a multi-agent distributed online load partitioning method and system for geological disaster monitoring, which can improve the quality of geological disaster monitoring.
[0004] The present invention provides a multi-agent distributed online load partitioning method for geological disaster monitoring, comprising the following steps: S1: according to the working form of the multi-agent system in geological disaster monitoring, using the basic principle of sweeping coverage, constructing a regional scanning coverage mathematical model of the area to be monitored; S2: according to the regional scanning coverage mathematical model, constructing the agent partitioning control law along the short side direction, and obtaining the dynamic coordinates of the multi-agent system moving along the short side direction; S3: according to the agent partitioning control law along the short side direction, obtaining the dynamic coordinates of the multi-agent system moving in the direction perpendicular to the short side; S4: according to the agent partitioning control law along the short side direction and the dynamic coordinates of the multi-agent system moving in the direction perpendicular to the short side, using a neural network algorithm to obtain geological disaster distribution probability information.
[0005] Furthermore, step S1 of the multi-agent distributed online load partitioning method for geological disaster monitoring specifically includes: according to the working form of the multi-agent system in geological disaster monitoring, using the basic principle of sweeping coverage, constructing a regional scanning coverage mathematical model, such as formula:
[0006] L μ = {q∈R 2 |q=μ+lσ(θ),0≤l≤l b}
[0007] σ(θ)=[cosθ,sinθ] T
[0008] p i =(p ix ,p iy ) T
[0009] Among them, L μ is a line segment parallel to the short side of the monitored area D, q represents the geographical coordinate information of the monitored area, and μ is the line segment L μ The endpoint coordinate on the x-axis, l is L μ The distance from the detection area to μ, σ(θ) is an auxiliary matrix for calculating the coordinate information of q, and θ is the angle between the long side and the short side of the area D; p i Indicates the position of the i-th agent, and multiple agents are distributed in L μ Up, p ix is the coordinate of the ith agent on the x-axis, p iy is the coordinate of the ith agent on the y-axis.
[0010] Furthermore, step S2 of the multi-agent distributed online load partitioning method for geological disaster monitoring specifically includes: constructing a partitioning control law of the agent along the short side direction according to the regional scanning coverage mathematical model, such as formula:
[0011] u i =m i -m i+1
[0012]
[0013] Then we can get the dynamic coordinates of the multi-agent system moving along the short side, as shown in the formula:
[0014] p i ((k+1)T)=p i (kT)+u i (kT)*T
[0015] Among them, p iy (kT) represents the y-axis coordinate information of the i-th agent at a certain moment, p ix ((k+1)T) represents the y-axis coordinate information of the agent at the next moment; u i For along L μ The partition control law of the direction is used. Under the action of the partition control law, the intelligent agent continuously adjusts its μ Direction of position; i-1,i Indicates p i-1 and p i The area between, ρ(x,y) is the probability of geological disasters in the sub-area, m i (t) and m i Represents the cumulative probability of geological disasters in the agent's neighboring area.
[0016] Furthermore, step S3 of the multi-agent distributed online load partitioning method for geological disaster monitoring specifically includes: according to the partition control law of the agent along the short side direction, the dynamic coordinates of the multi-agent system moving along the direction perpendicular to the short side are obtained, such as formula:
[0017] p ix ((k+1)T)=p ix (kT)+v*T
[0018] Among them, p ix (kT) represents the x-axis coordinate information of the ith agent at a certain moment, p ix ((k+1)T) represents the x-axis coordinate information of the agent at the next moment. The multi-agent system is stable from input to state; T is the time interval, and v is the speed of the multi-agent system moving along the direction perpendicular to the short side.
[0019] Furthermore, step S4 of the above-mentioned multi-agent distributed online load segmentation method for geological disaster monitoring specifically includes: S41: dividing the area to be detected into a preset number of sub-areas, and obtaining a label vector according to the preset number of sub-areas and the geological disaster distribution law; S42: using the label vector and the coordinates of the preset number of sub-areas of the area to be detected to train a neural network to obtain a trained neural network; S43: according to the agent partition control law along the short side direction and the dynamic coordinates of the multi-agent system moving along the direction perpendicular to the short side, using the trained neural network to obtain the geological disaster distribution probability information.
[0020] The present invention also provides a system, comprising the following modules: a scanning coverage model construction module, configured as: according to the working form of the multi-agent system in geological disaster monitoring, using the basic principle of sweeping coverage, constructing a regional scanning coverage mathematical model of the area to be monitored; a short side direction partitioning control law construction module, configured as: according to the regional scanning coverage mathematical model, constructing the agent partitioning control law along the short side direction, and obtaining the dynamic coordinates of the multi-agent system moving along the short side direction; a vertical short side direction dynamic coordinate acquisition module, configured as: according to the agent partitioning control law along the short side direction, obtaining the dynamic coordinates of the multi-agent system moving along the vertical short side direction; a geological disaster distribution probability acquisition module, configured as: according to the agent partitioning control law along the short side direction and the dynamic coordinates of the multi-agent system moving along the vertical short side direction, using a neural network algorithm to obtain geological disaster distribution probability information.
[0021] Furthermore, the scanning coverage model construction module of the above system is specifically configured as follows: according to the working form of the multi-agent system in geological disaster monitoring, using the basic principle of sweeping coverage, a regional scanning coverage mathematical model is constructed, such as the formula:
[0022] L μ= {q∈R 2 |q=μ+lσ(θ),0≤l≤l b}
[0023] σ(θ)=[cosθ,sinθ] T
[0024] p i =(p ix ,p iy ) T
[0025] Among them, L μ is a line segment parallel to the short side of the monitored area D, q represents the geographical coordinate information of the monitored area, and μ is the line segment L μ The endpoint coordinate on the x-axis, l is L μ The distance from the detection area to μ, σ(θ) is an auxiliary matrix for calculating the coordinate information of q, and θ is the angle between the long side and the short side of the area D; p i Indicates the position of the i-th agent, and multiple agents are distributed in L μ Up, p ix is the coordinate of the ith agent on the x-axis, p iy is the coordinate of the ith agent on the y-axis.
[0026] Furthermore, the short side direction partition control law construction module of the above system is specifically configured as follows: according to the area scanning coverage mathematical model, the partition control law of the intelligent agent along the short side direction is constructed, such as the formula:
[0027] u i =m i -m i+1
[0028]
[0029] Then we can get the dynamic coordinates of the multi-agent system moving along the short side, as shown in the formula:
[0030] p i ((k+1)T)=p i (kT)+u i (kT)*T
[0031] Among them, p iy (kT) represents the y-axis coordinate information of the i-th agent at a certain moment, p ix ((k+1)T) represents the y-axis coordinate information of the agent at the next moment; u i For along L μ The partition control law of the direction is used. Under the action of the partition control law, the intelligent agent continuously adjusts its μ Direction of position; i-1,iIndicates p i-1 and p i The area between, ρ(x,y) is the probability of geological disasters in the sub-area, m i (t) and m i Represents the cumulative probability of geological disasters in the agent's neighboring area.
[0032] Furthermore, the vertical short side direction dynamic coordinate acquisition module of the above system is specifically configured as follows: according to the agent partition control law along the short side direction, the dynamic coordinates of the multi-agent system moving along the vertical short side direction are obtained, such as the formula:
[0033] p ix ((k+1)T)=p ix (kT)+v*T
[0034] Among them, p ix (kT) represents the x-axis coordinate information of the ith agent at a certain moment, p ix ((k+1)T) represents the x-axis coordinate information of the agent at the next moment. The multi-agent system is stable from input to state; T is the time interval, and v is the speed of the multi-agent system moving along the direction perpendicular to the short side.
[0035] Furthermore, the geological disaster distribution probability acquisition module of the above-mentioned system is specifically configured as follows: dividing the area to be detected into a preset number of sub-areas, and obtaining a label vector according to the preset number of sub-areas and the geological disaster distribution law; using the label vector and the coordinates of the preset number of sub-areas of the area to be detected to train a neural network to obtain a trained neural network; according to the intelligent agent partition control law along the short side direction and the dynamic coordinates of the multi-agent system moving along the direction perpendicular to the short side, using the trained neural network to obtain the geological disaster distribution probability information.
[0036] The implementation of the multi-agent distributed online load partitioning method and system for geological disaster monitoring provided by the present invention has the following beneficial effects:
[0037] The present invention applies sweeping coverage to geological disaster monitoring. First, a mathematical model of distributed online load segmentation of a multi-agent system is established. Then, a corresponding motion control algorithm is designed to enable the agent to communicate with its nearest neighbor to achieve equal division of the sub-region load. Then, a neural network algorithm is used to perform high-quality dynamic monitoring of geological disasters. Through a small amount of communication between agents, online segmentation of sub-region loads is achieved when performing monitoring tasks in areas with uncertain distribution probabilities. This allows sub-region monitoring and task allocation to be carried out simultaneously, which can evenly distribute the workload, improve monitoring efficiency, and thus improve the quality of geological disaster monitoring. BRIEF DESCRIPTION OF THE DRAWINGS
[0038] The present invention will be further described below with reference to the accompanying drawings and embodiments, in which:
[0039] Figure 1 It is a flow chart of a multi-agent distributed online load partitioning method for geological disaster monitoring provided by the present invention;
[0040] Figure 2 It is a schematic diagram of multi-agent operation provided by the present invention;
[0041] Figure 3 It is an overview diagram of a neural network provided by the present invention;
[0042] Figure 4 The present invention provides a MATLAB simulation experiment on geological disaster distribution;
[0043] Figure 5 It is the motion trajectory diagram of the intelligent body provided by the present invention. DETAILED DESCRIPTION
[0044] In order to have a clearer understanding of the technical features, purposes and effects of the present invention, specific embodiments of the present invention are now described in detail with reference to the accompanying drawings.
[0045] Figure 1 A schematic diagram of a multi-agent distributed online load partitioning method for geological disaster monitoring in this embodiment is shown. In this embodiment, the multi-agent distributed online load partitioning method for geological disaster monitoring includes the following steps:
[0046] S1: According to the working form of the multi-agent system in geological disaster monitoring, using the basic principle of sweeping and covering, a regional scanning and covering mathematical model of the area to be monitored is constructed;
[0047] Specifically, step S1 of the multi-agent distributed online load partitioning method for geological disaster monitoring includes: according to the working form of the multi-agent system in geological disaster monitoring, using the basic principle of sweeping coverage, constructing a regional scanning coverage mathematical model, such as formula:
[0048] L μ = {q∈R 2 |q=μ+lσ(θ),0≤l≤l b}
[0049] σ(θ)=[cosθ,sinθ] T
[0050] p i =(p ix ,p iy ) T
[0051] Among them, L μ is a line segment parallel to the short side of the monitored area D, q represents the geographical coordinate information of the monitored area, and μ is the line segment L μ The endpoint coordinate on the x-axis, l is Lμ The distance from the detection area to μ, σ(θ) is an auxiliary matrix for calculating the coordinate information of q, and θ is the angle between the long side and the short side of the area D; p i Indicates the position of the i-th agent, and multiple agents are distributed in L μ Up, p ix is the coordinate of the ith agent on the x-axis, p iy is the coordinate of the ith agent on the y-axis;
[0052] S2: Based on the area scanning coverage mathematical model, the partition control law of the agent along the short side direction is constructed to obtain the dynamic coordinates of the multi-agent system moving along the short side direction;
[0053] Specifically, step S2 of the multi-agent distributed online load partitioning method for geological disaster monitoring specifically includes: constructing the partition control law of the agent along the short side direction according to the regional scanning coverage mathematical model, such as formula:
[0054] u i =m i -m i+1
[0055]
[0056] Then we can get the dynamic coordinates of the multi-agent system moving along the short side, as shown in the formula:
[0057] p i ((k+1)T)=p i (kT)+u i (kT)*T
[0058] Among them, p iy (kT) represents the y-axis coordinate information of the i-th agent at a certain moment, p ix ((k+1)T) represents the y-axis coordinate information of the agent at the next moment; u i For along L μ The partition control law of the direction is used. Under the action of the partition control law, the intelligent agent continuously adjusts its μ Direction of position; i-1,i Indicates p i-1 and p i The area between, ρ(x,y) is the probability of geological disasters in the sub-area, m i (t) and m i Represents the cumulative probability of geological disasters in the agent's neighboring area;
[0059] S3: According to the partition control law of the agent along the short side direction, the dynamic coordinates of the multi-agent system moving along the direction perpendicular to the short side are obtained;
[0060] Specifically, step S3 of the multi-agent distributed online load partitioning method for geological disaster monitoring specifically includes: according to the partition control law of the agent along the short side direction, the dynamic coordinates of the multi-agent system moving along the direction perpendicular to the short side are obtained, such as the formula:
[0061] p ix ((k+1)T)=p ix (kT)+v*T
[0062] Among them, p ix (kT) represents the x-axis coordinate information of the ith agent at a certain moment, p ix ((k+1)T) represents the x-axis coordinate information of the agent at the next moment. The multi-agent system is stable from input to state; T is the time interval, and v is the speed of the multi-agent system moving along the vertical short side;
[0063] S4: Based on the partition control law of the agent along the short side direction and the dynamic coordinates of the multi-agent system moving along the direction perpendicular to the short side, the neural network algorithm is used to obtain the probability information of geological disaster distribution;
[0064] Specifically, step S4 of the above-mentioned multi-agent distributed online load segmentation method for geological disaster monitoring specifically includes: S41: dividing the area to be detected into a preset number of sub-areas, and obtaining a label vector according to the preset number of sub-areas and the law of geological disaster distribution; S42: using the label vector and the coordinates of the preset number of sub-areas of the area to be detected to train a neural network to obtain a trained neural network; S43: according to the control law of the partitioning of the agent along the short side direction and the dynamic coordinates of the multi-agent system moving along the direction perpendicular to the short side, using the trained neural network, obtain the probability information of geological disaster distribution.
[0065] This embodiment provides a system, including the following modules:
[0066] The scanning coverage model building module is configured as follows: according to the working form of the multi-agent system in geological disaster monitoring, using the basic principle of sweeping coverage, a regional scanning coverage mathematical model of the area to be monitored is built;
[0067] Specifically, the scanning coverage model construction module of the above system is specifically configured as follows: according to the working form of the multi-agent system in geological disaster monitoring, using the basic principle of sweeping coverage, a regional scanning coverage mathematical model is constructed, such as the formula:
[0068] L μ = {q∈R 2 |q=μ+lσ(θ),0≤l≤l b}
[0069] σ(θ)=[cosθ,sinθ] T
[0070] p i =(p ix ,p iy ) T
[0071] Among them, L μ is a line segment parallel to the short side of the monitored area D, q represents the geographical coordinate information of the monitored area, and μ is the line segment L μ The endpoint coordinate on the x-axis, l is L μ The distance from the detection area to μ, σ(θ) is an auxiliary matrix for calculating the coordinate information of q, and θ is the angle between the long side and the short side of the area D; p i Indicates the position of the i-th agent, and multiple agents are distributed in L μ Up, p ix is the coordinate of the ith agent on the x-axis, p iy is the coordinate of the ith agent on the y-axis;
[0072] The module for building the partition control law in the short side direction is configured as follows: according to the mathematical model of area scanning coverage, the partition control law of the agent along the short side direction is built to obtain the dynamic coordinates of the multi-agent system moving along the short side direction;
[0073] Specifically, the short-side direction partition control law construction module of the above system is specifically configured as follows: according to the area scanning coverage mathematical model, the partition control law of the intelligent agent along the short side direction is constructed, such as the formula:
[0074] u i =m i -m i+1
[0075]
[0076] Then we can get the dynamic coordinates of the multi-agent system moving along the short side, as shown in the formula:
[0077] p i ((k+1)T)=p i (kT)+u i (kT)*T
[0078] Among them, p iy (kT) represents the y-axis coordinate information of the i-th agent at a certain moment, p ix ((k+1)T) represents the y-axis coordinate information of the agent at the next moment; u i For along L μ The partition control law of the direction is used. Under the action of the partition control law, the intelligent agent continuously adjusts its μ Direction of position; i-1,i Indicates p i-1 and pi The area between, ρ(x,y) is the probability of geological disasters in the sub-area, m i (t) and m i Represents the cumulative probability of geological disasters in the agent's neighboring area;
[0079] The dynamic coordinate acquisition module in the direction perpendicular to the short side is configured to: obtain the dynamic coordinates of the multi-agent system moving in the direction perpendicular to the short side according to the partition control law of the agent in the direction of the short side;
[0080] Specifically, the vertical short side direction dynamic coordinate acquisition module of the above system is specifically configured as follows: according to the agent partition control law along the short side direction, the dynamic coordinates of the multi-agent system moving along the vertical short side direction are obtained, as shown in the formula:
[0081] p ix ((k+1)T)=p ix (kT)+v*T
[0082] Among them, p ix (kT) represents the x-axis coordinate information of the ith agent at a certain moment, p ix ((k+1)T) represents the x-axis coordinate information of the agent at the next moment. The multi-agent system is stable from input to state; T is the time interval, and v is the speed of the multi-agent system moving along the vertical short side;
[0083] The geological disaster distribution probability acquisition module is configured as follows: according to the partition control law of the agent along the short side direction and the dynamic coordinates of the multi-agent system moving along the direction perpendicular to the short side, the geological disaster distribution probability information is acquired by using a neural network algorithm;
[0084] Specifically, the geological disaster distribution probability acquisition module of the above-mentioned system is specifically configured as follows: dividing the area to be detected into a preset number of sub-areas, and obtaining a label vector according to the preset number of sub-areas and the geological disaster distribution law; using the label vector and the coordinates of the preset number of sub-areas of the area to be detected to train the neural network and obtain a trained neural network; according to the partition control law of the intelligent agent along the short side direction and the dynamic coordinates of the multi-agent system moving along the direction perpendicular to the short side, using the trained neural network to obtain the geological disaster distribution probability information.
[0085] In some embodiments, the above-mentioned multi-agent distributed online load partitioning method for geological disaster monitoring can also be implemented in the following manner.
[0086] In this embodiment, the multi-agent distributed online load partitioning method for geological disaster monitoring specifically includes the following steps:
[0087] Step 1) Based on the basic principle of sweeping and covering, analyze the working form of the multi-agent system in geological disaster monitoring. The regional scanning and covering model is as follows: Figure 2 As shown:
[0088] L μ = {q∈R 2 |q=μ+lσ(θ),0≤l≤l b}
[0089] σ(θ)=[cosθ,sinθ] T
[0090] μ=(μ x , μ y ) T
[0091] Among them, L μ is a line segment parallel to the short side of area D, q represents the geographical coordinate information of the area to be monitored, and μ is the line segment L μ The endpoint coordinate on the x-axis, l is L μ The distance from the detection area to μ, σ(θ) is an auxiliary matrix for calculating the coordinate information of q; multiple agents are distributed in L μ On, the position is p i express:
[0092] p i =(p ix ,p iy ) T
[0093] In vertical L μ The area D is divided into several strips in the direction of L. μ The position of the nearest neighbor on the network and the probability information on the corresponding interval, after the intelligent agent monitors the regional probability information, the load division can be realized according to the distributed control law;
[0094] Step 2) Design a distributed control method for online load division, so that the multi-agent system can divide the load of the next strip area while monitoring the sub-area of the current strip area, and reasonably assign tasks to each agent; the agents move along L μ Direction partition control is as follows:
[0095] u i =m i -m i+1
[0096]
[0097] Then we can get the dynamic coordinates of the multi-agent system moving along the short side, as shown in the formula:
[0098] p iy ((k+1)T)=p iy (kT)+ui (kT)*T
[0099] Among them, p iy (kT) represents the y-axis coordinate information of the i-th agent at a certain moment, then p ix ((k+1)T) represents the y-axis coordinate information of the agent at the next moment; u i For along L μ The partition control law of the direction is used. Under the action of the partition control law, the intelligent agent continuously adjusts its μ Direction of position; i-1,i Indicates p i-1 and p i The area between, ρ(x,y) is the probability of geological disasters in the sub-area, m i Represents the cumulative probability of geological disasters in the agent's neighboring area;
[0100] Step 3) Apply the partition control law to the multi-agent distributed online load partitioning system for geological disaster monitoring;
[0101] Divide the strip area into several sub-areas, and the agent can detect the geological disaster information of the corresponding n sub-areas between the nearest neighbors; at this time, the geological disaster information is discrete data, so the partition control law u i middle Δs k , k are the area and disaster probability of the kth sub-region, m i It still represents the cumulative probability of geological disasters in the vicinity of the agent; while monitoring the strip area, the agent moves forward at a speed v to achieve dynamic coverage monitoring of the area; the multi-agent system moves along the vertical L at a speed v μ The dynamic change process of the coordinates of the directional movement is as follows:
[0102] p ix ((k+1)T)=p ix (kT)+v*T
[0103] Among them, p ix (kT) represents the x-axis coordinate information of the ith agent at a certain moment, then p ix ((k+1)T) represents the x-axis coordinate information of the agent at the next moment; the multi-agent system is stable from input to state;
[0104] Step 4) using a neural network algorithm to obtain geological disaster distribution probability information;
[0105] First, divide D into 100×100 sub-regions, and assign a label vector y of 0 or 1 to each sub-region; assuming that geological disasters are mainly distributed in the central area of D, based on this feature, it can be known that the probability of the label of the sub-region in the central row is higher than that of the label of the sub-region at the edge is higher. The code is designed to assign a label vector to each sub-region according to the above rule; then, Figure 3 As shown in the figure, the coordinate information of the sub-region is used as the input feature value, and the label vector y is used as the target output to train the neural network. Then, the trained neural network can be used to predict the probability of the new sub-region and successfully obtain the probability information of geological disaster distribution.
[0106] The constructed neural network has two input layer feature values, namely the x-axis coordinate and y-axis coordinate of the agent, and two hidden layers, containing 10 and 5 neurons respectively. The output layer outputs one value, namely the probability of geological disasters in the sub-region. The scaled conjugate gradient algorithm is used to optimize the neural network training process, and the cross entropy is used as the loss function to measure the difference between the neural network prediction output and the actual label.
[0107] Finally, using the trained neural network on the MATLAB platform, we designed a code to conduct a simulation experiment on the distribution of geological disasters, and at the same time displayed the probability distribution of geological disasters in the form of thermal imaging. Figure 4 The specific implementation steps are as follows:
[0108] 1. Initialization parameters: set the number of rows and columns of the sub-region, and the size of the hidden layer;
[0109] 2. Generate sub-region coordinates: Use the meshgrid function to generate sub-region coordinate matrices x1 and x2, and convert the coordinate matrices into column vector form and combine them into X_all;
[0110] 3. Initialize the label vector: create an all-zero column vector y to store the label of each sub-region;
[0111] 4. Assign labels: traverse each grid, set different probabilities according to the coordinate range to generate sub-region vector labels, assign high-probability labels to sub-regions in the specified range, and assign low-probability labels to other sub-regions;
[0112] 5. Build a neural network: Use the patternnet function to create a neural network with a specified hidden layer size;
[0113] 6. Train the neural network: Use the train function to train the neural network, with the input data as X_all and the target data as the label vector y;
[0114] 7. Predict the probability of new coordinates: Use the trained neural network to predict the original coordinate matrix X_all to obtain the probability of geological disasters;
[0115] 8. Visualize predicted probability: Use the imagesc function to display the predicted disaster probability as a heat map;
[0116] Step 5) Design code on the MATLAB platform to implement the simulation of the multi-agent distributed online load partitioning method for geological disaster monitoring, such as Figure 5 The specific implementation steps are as follows:
[0117] 1. Initialization parameters: Set the number of agents N = 3, the number of iterations Step = 800, the length and width of area D La = 8, Lb = 5, Lc = 1, and the vertical L μ The speed in the direction is v=1, the time step Ts=0.01, the initial coordinate position of the multi-agent p=[0,2; 0,3; 0,4], and the empty matrix vp=[0,0; 0,0; 0,0] for storing the changed coordinates of the multi-agent;
[0118] 2. Draw the initial graph: Use the plot function to draw an empty coordinate axis to display the motion trajectory of multiple agents later;
[0119] 3. Iterative motion simulation: Design a nested loop function and use the outer for loop function to perform Step iterations. In each outer iteration, the x-axis of the restricted area D is 0~La and the y-axis is Lc~Lb. Each outer iteration uses the first inner for loop function to perform N inner iterations to update the distribution law u of each agent. i and position p i , and stores the new position in vp. Each outer iteration uses the second inner for loop function to perform N inner iterations, and updates the p matrix using the new position information in vp;
[0120] The first inner for loop function updates the distribution law u of each agent i and position p i The specific idea is as follows: in each inner iteration, first use the if loop function to collect the probability of geological disasters in the neighboring area of the agent one by one, and calculate the product of the probability of each sub-area in the adjacent area and the area of the sub-area according to the discrete formula and accumulate them to get the corresponding m i and m i+1 , and then we get the distribution law u i , and according to p iy ((k+1)T)=p iy (kT)+u i (kT)*T updates the y-axis coordinate of the agent according to p ix ((k+1)T)=p ix (kT)+v*T updates the x-axis coordinate of the agent;
[0121] by Figure 4 For example, area D is divided into 100×100 sub-areas, and multiple agents are initially evenly distributed on the y-axis, that is, there are 25 sub-areas between every two adjacent agents. Due to coordinate updates, the number and position of sub-areas between adjacent agents may change. The round function can be used to update the number and position of sub-areas between adjacent agents in a rounded form.
[0122] It can be seen that the motion trajectory of the intelligent agent is Figure 4 The distribution of geological disasters is consistent with that of the previous two studies, which verifies the performance of the method.
[0123] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation modes, which are merely illustrative rather than restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the present invention and the claims, all of which are within the protection of the present invention.
Claims
1. A multi-agent distributed online load partitioning method for geological disaster monitoring, characterized in that: The following steps are involved: S1: According to the working form of the multi-agent system in geological disaster monitoring, using the basic principle of sweeping and covering, a regional scanning and covering mathematical model of the area to be monitored is constructed; S2: According to the area scanning coverage mathematical model, a partition control law of the agent along the short side direction is constructed to obtain the dynamic coordinates of the multi-agent system moving along the short side direction; S3: According to the partition control law of the agent along the short side direction, the dynamic coordinates of the multi-agent system moving along the direction perpendicular to the short side are obtained; S4: According to the partition control law of the agent along the short side direction and the dynamic coordinates of the multi-agent system moving along the direction perpendicular to the short side, a neural network algorithm is used to obtain the geological disaster distribution probability information.
2. The multi-agent distributed online load partitioning method for geological disaster monitoring according to claim 1 is characterized in that: Step S1 specifically includes: according to the working form of the multi-agent system in geological disaster monitoring, using the basic principle of sweeping and covering, constructing a regional scanning and covering mathematical model, such as the formula: L μ ={q∈R 2 |q=μ+lσ(θ),0≤l≤lb} σ(θ)=[cosθ,sinθ]T pi=(pix,piy)T Among them, L μ is a line segment parallel to the short side of the monitored area D, q represents the geographical coordinate information of the monitored area, and μ is the line segment L μ The endpoint coordinate on the x-axis, l is L μ The distance from the detection area to μ, σ(θ) is an auxiliary matrix for calculating the coordinate information of q, and θ is the angle between the long side and the short side of the area D; p i Indicates the position of the i-th agent, and multiple agents are distributed in L μ Up, p ix is the coordinate of the ith agent on the x-axis, p iy is the coordinate of the ith agent on the y-axis.
3. The multi-agent distributed online load partitioning method for geological disaster monitoring according to claim 1 is characterized in that: Step S2 specifically includes: constructing a partition control law of the intelligent agent along the short side direction according to the area scanning coverage mathematical model, such as the formula: ui=mi-mi+1 Then we can get the dynamic coordinates of the multi-agent system moving along the short side, as shown in the formula: pi((k+1)T)=pi(kT)+ui(kT)*T Among them, p iy (kT) represents the y-axis coordinate information of the i-th agent at a certain moment, p ix ((k+1)T) represents the y-axis coordinate information of the agent at the next moment; u i For along L μ The partition control law of the direction is used. Under the action of the partition control law, the intelligent agent continuously adjusts its μ Direction of position; i-1,i Indicates p i-1 and p i The area between, ρ(x,y) is the probability of geological disasters in the sub-region, mi(t) and m i Represents the cumulative probability of geological disasters in the agent's neighboring area.
4. The multi-agent distributed online load partitioning method for geological disaster monitoring according to claim 1 is characterized in that: Step S3 specifically includes: according to the partition control law of the agent along the short side direction, the dynamic coordinates of the multi-agent system moving along the direction perpendicular to the short side are obtained, such as the formula: pix((k+1)T)=pix(kT)+v*T Among them, p ix (kT) represents the x-axis coordinate information of the ith agent at a certain moment, p ix ((k+1)T) represents the x-axis coordinate information of the agent at the next moment. The multi-agent system is stable from input to state; T is the time interval, and v is the speed of the multi-agent system moving along the direction perpendicular to the short side.
5. The multi-agent distributed online load partitioning method for geological disaster monitoring according to claim 1 is characterized in that: Step S4 specifically includes: S41: Divide the area to be detected into a preset number of sub-areas, and obtain a label vector according to the preset number of sub-areas and the distribution law of geological disasters; S42: training a neural network using the label vector and the coordinates of a preset number of sub-regions of the region to be detected to obtain a trained neural network; S43: According to the partition control law of the agent along the short side direction and the dynamic coordinates of the multi-agent system moving along the direction perpendicular to the short side, the geological disaster distribution probability information is obtained using the trained neural network.
6. A system, characterized in that: The system includes the following modules: The scanning coverage model building module is configured as follows: according to the working form of the multi-agent system in geological disaster monitoring, using the basic principle of sweeping coverage, a regional scanning coverage mathematical model of the area to be monitored is built; The short side direction partition control law construction module is configured to: construct the agent partition control law along the short side direction according to the area scanning coverage mathematical model, and obtain the dynamic coordinates of the multi-agent system along the short side direction; A dynamic coordinate acquisition module in the direction perpendicular to the short side, configured to: obtain the dynamic coordinates of the multi-agent system moving in the direction perpendicular to the short side according to the partition control law of the agent in the direction perpendicular to the short side; The geological disaster distribution probability acquisition module is configured to: obtain the geological disaster distribution probability information using a neural network algorithm based on the partition control law of the agent along the short side direction and the dynamic coordinates of the multi-agent system moving along the direction perpendicular to the short side.
7. A system according to claim 6, characterized in that: The scanning coverage model building module is specifically configured as follows: according to the working form of the multi-agent system in geological disaster monitoring, using the basic principle of sweeping coverage, a regional scanning coverage mathematical model is constructed, such as the formula: L μ ={q∈R 2 |q=μ+lσ(θ),0≤l≤lb} σ(θ)=[cosθ,sinθ]T pi=(pix,piy)T Among them, L μ is a line segment parallel to the short side of the monitored area D, q represents the geographical coordinate information of the monitored area, and μ is the line segment L μ The endpoint coordinate on the x-axis, l is L μ The distance from the detection area to μ, σ(θ) is an auxiliary matrix for calculating the coordinate information of q, and θ is the angle between the long side and the short side of the area D; p i Indicates the position of the i-th agent, and multiple agents are distributed in L μ Up, p ix is the coordinate of the ith agent on the x-axis, p iy is the coordinate of the ith agent on the y-axis.
8. A system according to claim 6, characterized in that: The short side direction partition control law construction module is specifically configured as follows: according to the area scanning coverage mathematical model, the agent partition control law along the short side direction is constructed, such as formula: ui=mi-mi+1 Then we can get the dynamic coordinates of the multi-agent system moving along the short side, as shown in the formula: pi((k+1)T)=pi(kT)+ui(kT)*T Among them, p iy (kT) represents the y-axis coordinate information of the i-th agent at a certain moment, p ix ((k+1)T) represents the y-axis coordinate information of the agent at the next moment; u i For along L μ The partition control law of the direction is used. Under the action of the partition control law, the intelligent agent continuously adjusts its μ Direction of position; i-1,i Indicates p i-1 and p i The area between, ρ(x,y) is the probability of geological disasters in the sub-region, mi(t) and m i Represents the cumulative probability of geological disasters in the agent's neighboring area.
9. A system according to claim 6, characterized in that: The vertical short side direction dynamic coordinate acquisition module is specifically configured as follows: according to the partition control law of the agent along the short side direction, the dynamic coordinates of the multi-agent system moving along the vertical short side direction are obtained, such as the formula: pix((k+1)T)=pix(kT)+v*T Among them, p ix (kT) represents the x-axis coordinate information of the ith agent at a certain moment, p ix ((k+1)T) represents the x-axis coordinate information of the agent at the next moment. The multi-agent system is stable from input to state; T is the time interval, and v is the speed of the multi-agent system moving along the direction perpendicular to the short side.
10. A system according to claim 6, characterized in that: The specific configuration of the geological disaster distribution probability acquisition module is as follows: Divide the area to be detected into a preset number of sub-areas, and obtain a label vector according to the preset number of sub-areas and the distribution law of geological disasters; Using the label vector and the coordinates of a preset number of sub-regions of the area to be detected, training a neural network to obtain a trained neural network; According to the partition control law of the agent along the short side direction and the dynamic coordinates of the multi-agent system moving along the direction perpendicular to the short side, the geological disaster distribution probability information is obtained by using the trained neural network.
