Unmanned cluster social heuristic cooperative crowd intelligence method for emergency rescue

Abstract

The invention discloses an unmanned cluster social heuristic collaborative crowd intelligence method for emergency rescue. A task matching and path planning framework with an auction game as a core is constructed. The method comprises the following steps: firstly, through a multi-capability modeling and auction mechanism, combining a coverage rate, a task balance rate and a capability matching degree to construct an initial path planning model; then, fire situation prediction and personnel dynamic evaluation are fused, dynamic adjustment of task requirements and regional state classification are realized, and a process path reconstruction mechanism is introduced to form a self-adaptive optimization model of a dual-region target. And finally, through an improved auction game and a mixed ADMM distributed solution strategy, efficient coordination and intelligent scheduling of the unmanned system in a complex dynamic environment are realized, and the intelligent scheduling capability and task execution collaboration of the unmanned system in a complex fire environment are remarkably enhanced.

Description

Technical Field

[0001] The present invention relates to the field of swarm intelligence and multi-agent collaborative control, and specifically to an unmanned cluster social heuristic collaborative swarm intelligence method for emergency rescue. The method belongs to the field of emergency rescue and intelligent scheduling, and is particularly suitable for the collaborative operation of multi-agent systems in complex emergency scenarios such as multi-task dynamic allocation, path planning, and task reconstruction optimization. Background Art

[0002] With the increasing frequency of natural disasters, especially widespread fires, posing an increasingly severe threat to life and property, traditional emergency response methods that rely on manual scheduling and single-device execution are no longer able to meet the demands of rapid, efficient, and coordinated rescue efforts. Although multi-agent systems have been increasingly applied to fire monitoring and task execution in recent years, current approaches still face numerous challenges in practical applications, including a lack of dynamic coordination mechanisms for task allocation, conflicts between path planning and task coverage, and delayed regional status updates. These challenges make it difficult to effectively address the uncertainty of fire environments and the real-time changes in task requirements. Furthermore, existing technologies generally overlook the resource matching challenges arising from the heterogeneous capabilities and diverse tasks of unmanned systems, lacking the ability to achieve efficient coordination and intelligent game-playing in complex dynamic scenarios. Therefore, there is an urgent need to develop an intelligent scheduling method that integrates multi-source information perception, capability modeling, regional evolution analysis, and task-playing decision-making to improve the comprehensive response efficiency and intelligent coordination of multi-agent systems in disaster environments. Summary of the Invention

[0003] Purpose of the invention: In order to overcome the deficiencies in the prior art, the present invention provides an unmanned cluster social heuristic collaborative group intelligence method for emergency rescue.

[0004] Technical solution: To achieve the above purpose, the technical solution adopted by the present invention is:

[0005] A social heuristic collaborative group intelligence method for unmanned swarms for emergency rescue, comprising the following steps:

[0006] Step 1: Construct the capability function C of the unmanned system for the three tasks of detection, fire fighting and rescue i (t), integrating sensor accuracy, extinguishing agent diffusion, and payload error parameters, combined with environmental modeling, the rescue area is discretized into grid cells to establish a basic task-space structure.

[0007] Step 2: After the unmanned system identifies the fire point or personnel, it dynamically generates a set of task points, each of which has a task requirement vector An equilibrium task matching model is established based on coverage, capability matching and overlapping penalty terms, and multi-objective task allocation is performed using an auction game mechanism.

[0008] Step 3: Construct a multi-factor bidding model consisting of differential coverage gain, capability demand matching function, and flight penalty function, and combine it with a nonlinear pricing mechanism and price fluctuation limit update rules to complete task bidding and initial scheduling in path planning.

[0009] Step 4: Use environmental data and model feedback to estimate the fire situation and personnel behavior trends, dynamically update task demand weights based on the prediction results, and perform regional status classification to improve the adaptability and real-time performance of task scheduling.

[0010] Step 5: Divide the task area into unknown areas and known areas, and establish non-equilibrium optimization objectives for maximizing coverage and minimizing task requirement satisfaction respectively. Dynamically adjust their matching weights and priorities through evolution functions, and construct a multi-region price difference model and multi-region price update rules.

[0011] Step 6: Set differentiated bidding strategies and price update rules for different regional task points. Bids in unknown areas are driven by coverage, while bids in known areas are driven by demand satisfaction. A hybrid auction-ADMM mechanism is used for local bidding, global multiplier updates, and scheduling conflict coordination to complete the process path reconstruction.

[0012] Step 7: When the task point reaches the preset coverage and balance index thresholds, the area to which it belongs changes from "unknown" to "known", and the price index decay is reset to clear the interference left by historical prices and adapt to the subsequent evolution of scheduling strategies.

[0013] Preferably: In step 1, the detection capability function is based on the distance between the unmanned system and the target The angle and sensor performance modeling are specifically expressed as follows:

[0014]

[0015] in, is the comprehensive effectiveness of the sensors of the unmanned system i, R d k is the maximum distance that the sensor can recognize, d To control the rate parameter at which detection performance decreases with distance, θ ij is the viewing angle between unmanned system i and target j, θ max is the maximum field of view of the unmanned system sensor, p i is the location of the unmanned system i, The location of the currently detected target.

[0016] The fire extinguishing capability is modeled based on the extinguishing agent diffusion and wind speed difference and is specifically expressed as:

[0017]

[0018] Among them, Vi fire is the initial capacity of the fire extinguishing agent of unmanned system i, r0 is the diffusion range of the fire extinguishing agent under windless conditions, v ω -v i is the difference between the real-time wind speed and the speed of unmanned system i, α f is the volatility factor of the extinguishing agent in the air, σ f To allow the wind speed to deviate from the unmanned system, t exe It is the time that the fire extinguishing agent acts in the air.

[0019] The delivery capability is modeled based on the error propagation function and is specifically expressed as:

[0020]

[0021] in, is the current payload mass of the unmanned system i, A j is the range of rescue area j, is the position of the unmanned system p i Location of the delivery destination The distance, σ r The accuracy variance of the rescue supplies delivered to the unmanned system i is specifically expressed as:

[0022]

[0023] Among them, h i is the height of the unmanned system i above the ground, is the velocity fluctuation of the unmanned system i during hovering, σ alt is the height measurement error, σ p Provides accurate positioning of the unmanned system.

[0024] Preferably: In step 2, the regional coverage structure estimates the coverage rate by dividing the task area into grids and calculating whether each grid is covered. The coverage rate, overlap penalty rate, and task balance rate together constitute the matching optimization target, which is specifically expressed as:

[0025]

[0026] Among them, λ b ∈[0,1] is the weight for controlling task balance and coverage, U i is the union coverage, O i is the overlap penalty term, B j is the equilibrium rate of a single task point, and M is the total number of tasks.

[0027] Union coverage U i Specifically expressed as:

[0028]

[0029] Where K is the number of discretized grid cells, a ik Indicates whether unmanned system i covers the kth grid.

[0030] Union Coverage O i Specifically expressed as:

[0031]

[0032] Among them, (x) + =max(x,0).

[0033] Task balance rate B j Specifically expressed as:

[0034]

[0035] Among them, x ij ∈{0,1} represents whether unmanned system i is assigned to task point t j , C ik Represented as unmanned system u i The capability of task type k∈{D,F,R} at time t, is the task requirement value of the multi-task, For the task point t j Total supply at capacity k.

[0036] Preferably: in step 3, the bidding function b ij Specifically expressed as:

[0037]

[0038] in, To cover the gain part, For the task matching part, This is the voyage penalty portion.

[0039] Override gain section Specifically expressed as:

[0040]

[0041] Among them, ΔU i (j) is the new effective gain, ΔO i (j) is the newly added overlap gain, μ0 is the overlap penalty coefficient, is an indicator function, which is 1 if the condition is met and 0 otherwise.

[0042] Task matching part Specifically expressed as:

[0043]

[0044] Among them, ω k is the capability type weight.

[0045] Range penalty section Specifically expressed as:

[0046]

[0047] Among them, κ c is the quadratic coefficient used to suppress high-risk mission selection close to the range limit, is the maximum operating distance of unmanned system i, q j is the location of the destination point j of the unmanned system, is the initial position of the unmanned system i.

[0048] Mission point price The dynamic change of follows a nonlinear increasing mechanism, as shown below:

[0049]

[0050] in, The current best competitor, is the overlap rate of the task-associated areas, as shown below:

[0051]

[0052] Where K is the number of discretized grid cells.

[0053] Preferably, in step 4, the dynamic correction mechanism of the task adjusts the fire extinguishing requirements of the task point based on the predicted results, as shown below:

[0054]

[0055] in, is the critical fire intensity for demand surge, σ k ∈[0,1] is the demand sensitivity coefficient, The intensity of the fire.

[0056] The adjustments to the rescue requirements for mission points are as follows:

[0057]

[0058] Among them, η R is the rescue demand sensitivity coefficient, n j =|P j | is the number of people affected, P j is located at the task point t j Gathering of people within the radius, is the survival probability of the trapped persons, which is specifically expressed as:

[0059]

[0060] Among them, ‖x p (τ)-x nearest-fire (τ)‖ is the Euclidean distance between the fire source and the person, x p is the position of the trapped person, x nearest-fire Indicates the nearest fire source location, L safe is the safety distance threshold, t rescue The time it takes for rescue to arrive. The survival probability can only be calculated before rescue arrives, otherwise it will be reset to zero.

[0061] Preferably: in step 5, the task area is divided into an unknown area and a known area. For the unknown area, the objective function is specifically as follows:

[0062]

[0063] Among them, A u represents the range of the unknown area, λ u To control the equilibrium weight, For the task point t j The demand for the kth capability.

[0064] For a known area, the objective function is as follows:

[0065]

[0066] Among them, A k represents the range of the known area, λ u To control the equilibrium weight, For the task point t j The current dynamic needs of To be assigned to task point t j The sum of capabilities.

[0067] For unknown and known areas, the weights are dynamically adjusted through the evolution function as follows:

[0068]

[0069] Among them, κ u is the time attenuation coefficient, ν u is the sensitivity coefficient of the change rate of the unknown area, μ k To control the priority of urgent tasks, k is the demand threshold for determining urgent tasks, The priority of the rescue mission.

[0070] Based on the above weights and the objective functions for the two types of regions, the overall optimization objective function is constructed as follows:

[0071]

[0072] Among them, J k is the objective function of the known area, J u is the objective function of the unknown region, ω k (t) is the weight of the known area, ω u (t) is the weight of the unknown area.

[0073] Using the hybrid auction-ADMM framework, the local updates are as follows:

[0074]

[0075] Among them, ρ i is the ADMM penalty factor, z i Global shared variables. i is the Lagrange multiplier in ADMM, y i Assign variables to the current unmanned system.

[0076] In the global coordination stage, the alternating multiplier method is used to coordinate the task allocation between unmanned systems and resolve allocation conflicts. The global aggregation and multiplier update formulas are as follows:

[0077]

[0078] u i ←u i +y i -z

[0079] Where N is the number of unmanned systems.

[0080] Preferably: in step 7, the conditions for region conversion are specifically as follows:

[0081]

[0082] in, is the indicator function. When the corresponding value exceeds the threshold, the value of the indicator function is 1, c is the coverage threshold. b is the balance value threshold.

[0083] An unmanned swarm social heuristic collaborative crowd intelligence system for emergency rescue, used to implement the unmanned swarm social heuristic collaborative crowd intelligence method for emergency rescue, includes a basic task-space structure unit, a multi-objective task allocation, a task bidding and initial scheduling unit, a regional state classification unit, a multi-region optimization price difference update unit, a path reconstruction unit, and a reset unit, wherein:

[0084] The basic task-space structure unit is used to construct the capability function C of the unmanned system for the three tasks of detection, fire fighting and rescue. i (t), integrating sensor accuracy, extinguishing agent diffusion, and payload error parameters, combined with environmental modeling, the rescue area is discretized into grid cells to establish a basic task-space structure.

[0085] The multi-objective task allocation is used to dynamically generate a set of task points after the unmanned system identifies the fire point or personnel, and each task point has a task requirement vector An equilibrium task matching model is established based on coverage, capability matching and overlapping penalty terms, and multi-objective task allocation is performed using an auction game mechanism.

[0086] The task bidding and initial scheduling unit is used to construct a multi-factor bidding model consisting of differential coverage gain, capability demand matching function and flight penalty function, and combines the nonlinear pricing mechanism with the price fluctuation limit update rule to complete the task bidding and initial scheduling in path planning.

[0087] The regional status classification unit uses environmental data and model feedback to estimate the fire situation and personnel behavior trends, dynamically updates task demand weights based on the prediction results, and performs regional status classification to improve the adaptability and real-time performance of task scheduling.

[0088] The multi-region optimization price difference update unit is used to divide the task area into unknown areas and known areas, establish non-equilibrium optimization objectives for maximizing coverage and minimizing task requirement satisfaction, and dynamically adjust its matching weights and priorities through evolution functions to construct a multi-region price difference model and multi-region price update rules.

[0089] The path reconstruction unit sets differentiated bidding strategies and price update rules for different regional task points. Bidding in unknown areas is driven by coverage, while bidding in known areas is driven by demand satisfaction. A hybrid auction-ADMM mechanism is used to perform local bidding, global multiplier updates, and scheduling conflict coordination to complete the process path reconstruction.

[0090] The reset unit is used to change the area to which the task point belongs from "unknown" to "known" when the task point reaches the preset coverage and balance index thresholds, and reset the price index attenuation to clear the interference left by historical prices and adapt to the subsequent evolution of scheduling strategies.

[0091] Compared with the prior art, the present invention has the following beneficial effects:

[0092] The present invention constructs an integrated scheduling framework with capability modeling, auction game and dynamic optimization as the core. The method first establishes a task-space mapping structure by functional modeling of detection, fire extinguishing and rescue capabilities, combining parameters such as sensor accuracy, fire extinguishing agent diffusion characteristics and load error. Subsequently, based on multi-factor bidding functions and indicators such as coverage, capability matching, and task balance rate, a nonlinear auction game mechanism is used to complete multi-objective task matching and path planning. The system integrates the prediction results of fire situation and personnel behavior, dynamically updates task requirements and regional status classification, and realizes the "unknown" and "known" conversion of task areas. And by introducing a multi-region price difference model and an evolutionary weight adjustment mechanism, the adaptability and response efficiency of task scheduling are improved. Finally, with the help of the auction-ADMM hybrid optimization algorithm, local bidding, global conflict coordination and process path reconstruction are completed, significantly enhancing the intelligent scheduling capability and task execution coordination of unmanned systems in complex fire environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0093] Figure 1 It is a flow chart of the present invention.

[0094] Figure 2 It is a simulation scene visualization diagram of the present invention.

[0095] Figure 3 It is a visualization diagram of the movement trajectory of the unmanned system of the present invention. DETAILED DESCRIPTION

[0096] The present invention is further illustrated below with reference to the accompanying drawings and specific embodiments. It should be understood that these examples are only used to illustrate the present invention and are not used to limit the scope of the present invention. After reading the present invention, modifications of various equivalent forms of the present invention made by those skilled in the art all fall within the scope defined by the claims attached to this application.

[0097] This paper designs an unmanned cluster social heuristic collaborative group intelligence method for emergency rescue, the process of which is as follows: Figure 1-3 The specific steps are as follows:

[0098] Step S101: Construct the capability function C of the unmanned system for the three tasks of detection, fire fighting and rescue i (t), integrating parameters such as sensor accuracy, extinguishing agent diffusion, and load error, combined with environmental modeling, the rescue area is discretized into grid cells to establish a basic task-space structure. The detection capability function is based on the distance between the unmanned system and the target. The angle and sensor performance modeling are specifically expressed as follows:

[0099]

[0100] in, is the comprehensive effectiveness of the sensors of the unmanned system i, R d k is the maximum distance that the sensor can recognize, d To control the rate parameter at which detection performance decreases with distance, θ ij is the viewing angle between unmanned system i and target j, θ max is the maximum field of view of the unmanned system sensor, p i is the location of the unmanned system i, The location of the currently detected target.

[0101] The fire extinguishing capability is modeled based on the extinguishing agent diffusion and wind speed difference and is specifically expressed as:

[0102]

[0103] Among them, V i fire is the initial capacity of the fire extinguishing agent of unmanned system i, r0 is the diffusion range of the fire extinguishing agent under windless conditions, v ω -v i is the difference between the real-time wind speed and the speed of unmanned system i, α f is the volatility factor of the extinguishing agent in the air, σ f To allow the wind speed to deviate from the unmanned system, t exe It is the time that the fire extinguishing agent acts in the air.

[0104] The delivery capability is modeled based on the error propagation function and is specifically expressed as:

[0105]

[0106] in, is the current payload mass of the unmanned system i, A j is the range of rescue area j, is the position of the unmanned system p i Location of the delivery destination The distance, σ r The accuracy variance of the rescue supplies delivered to the unmanned system i is specifically expressed as:

[0107]

[0108] Among them, h i is the height of the unmanned system i above the ground, is the velocity fluctuation of the unmanned system i during hovering, σ alt is the height measurement error, σ p Provides accurate positioning of the unmanned system.

[0109] Step S102: After the unmanned system identifies the fire point or personnel, it dynamically generates a set of task points, each of which has a task requirement vector A balanced task matching model is established based on coverage, capability matching, and overlap penalty, and an auction game mechanism is used for multi-objective task allocation. The regional coverage structure estimates coverage by dividing the task area into grids and calculating whether each grid is covered. The combined coverage, overlap penalty, and task balance ratio constitute the matching optimization objective, which can be expressed as:

[0110]

[0111] Among them, λ b ∈[0,1] is the weight for controlling task balance and coverage, U i is the union coverage, O i is the overlap penalty term, B j is the equilibrium rate of a single task point, and M is the total number of tasks.

[0112] Union coverage U i Specifically expressed as:

[0113]

[0114] Where K is the number of discretized grid cells, a ik Indicates whether unmanned system i covers the kth grid.

[0115] Union Coverage O i Specifically expressed as:

[0116]

[0117] Among them, (x) + =max(x,0).

[0118] Task balance rate B j Specifically expressed as:

[0119]

[0120] Among them, x ij ∈{0,1} represents whether unmanned system i is assigned to task point t j , C ik Represented as unmanned system u i The capability of task type k∈{D,F,R} at time t, is the task requirement value of the multi-task, For the task point t j Total supply at capacity k.

[0121] Step S103: Construct a multi-factor bidding model consisting of differential coverage gain, capability demand matching function and flight penalty function, and combine the nonlinear pricing mechanism with the price fluctuation limit update rule to complete the task bidding and initial scheduling in the path planning. ij Specifically expressed as:

[0122]

[0123] in, To cover the gain part, For the task matching part, This is the voyage penalty portion.

[0124] Override gain section Specifically expressed as:

[0125]

[0126] Among them, ΔU i (j) is the new effective gain, ΔO i (j) is the newly added overlap gain, μ0 is the overlap penalty coefficient, is an indicator function, which is 1 if the condition is met and 0 otherwise.

[0127] Task matching part Specifically expressed as:

[0128]

[0129] Among them, ω k is the capability type weight.

[0130] Range penalty section Specifically expressed as:

[0131]

[0132] Among them, κ c is the quadratic coefficient used to suppress high-risk mission selection close to the range limit, is the maximum operating distance of unmanned system i, q j is the location of the destination point j of the unmanned system, is the initial position of the unmanned system i.

[0133] Mission point price The dynamic change of follows a nonlinear increasing mechanism, as shown below:

[0134]

[0135] in, The current best competitor, is the overlap rate of the task-associated areas, as shown below:

[0136]

[0137] Where K is the number of discretized grid cells.

[0138] Step S104: Utilize environmental data and model feedback to estimate fire scene trends and personnel behavior trends, dynamically update task demand weights based on the prediction results, and perform regional status classification to improve the adaptability and real-time performance of task scheduling. The dynamic task correction mechanism adjusts the fire extinguishing requirements of the task point based on the prediction results, as shown below:

[0139]

[0140] in, is the critical fire intensity for demand surge, σ k ∈[0,1] is the demand sensitivity coefficient, The intensity of the fire.

[0141] The adjustments to the rescue requirements for mission points are as follows:

[0142]

[0143] Among them, η R is the rescue demand sensitivity coefficient, n j =|P j | is the number of people affected, P j is located at the task point t j Gathering of people within the radius, is the survival probability of the trapped persons, which is specifically expressed as:

[0144]

[0145] Among them, ‖x p (τ)-x nearest-fire (τ)‖ is the Euclidean distance between the fire source and the person, x p is the position of the trapped person, x nearest-fire Indicates the nearest fire source location, L safe is the safety distance threshold, t rescue The time it takes for rescue to arrive. The survival probability can only be calculated before rescue arrives, otherwise it will be reset to zero.

[0146] Step S105: Divide the task area into unknown areas and known areas, establish non-equilibrium optimization objectives for maximizing coverage and minimizing task requirement satisfaction, and dynamically adjust their matching weights and priorities through evolutionary functions to build a multi-region price difference model and multi-region price update rules. The task area is divided into unknown areas and known areas. For the unknown area, the objective function is as follows:

[0147]

[0148] Among them, A u represents the range of the unknown area, λ u To control the equilibrium weight, For the task point t j The demand for the kth capability.

[0149] For a known area, the objective function is as follows:

[0150]

[0151] Among them, A k represents the range of the known area, λ u To control the equilibrium weight, For the task point t j The current dynamic needs of To be assigned to task point t j The sum of capabilities.

[0152] For unknown and known areas, the weights are dynamically adjusted through the evolution function as follows:

[0153]

[0154] Among them, κ u is the time attenuation coefficient, ν u is the sensitivity coefficient of the change rate of the unknown area, μ k To control the priority of urgent tasks, k is the demand threshold for determining urgent tasks, The priority of the rescue mission.

[0155] Based on the above weights and the objective functions for the two types of regions, the overall optimization objective function is constructed as follows:

[0156]

[0157] Among them, J k is the objective function of the known area, J u is the objective function of the unknown region, ω k (t) is the weight of the known area, ωu (t) is the weight of the unknown area.

[0158] Using the hybrid auction-ADMM framework, the local updates are as follows:

[0159]

[0160] Among them, ρ i is the ADMM penalty factor, z i Global shared variables. i is the Lagrange multiplier in ADMM, y i Allocation variables for the current unmanned system.

[0161] In the global coordination stage, the alternating multiplier method is used to coordinate the task allocation between unmanned systems and resolve allocation conflicts. The global aggregation and multiplier update formulas are as follows:

[0162]

[0163] u i ←u i +y i -z

[0164] Step S106: Differentiated bidding strategies and price update rules are set for different regional task points. Bidding in unknown areas is driven by coverage, while bidding in known areas is driven by demand satisfaction. A hybrid auction-ADMM mechanism is used for local bidding, global multiplier updates, and scheduling conflict coordination to complete process path reconstruction.

[0165] Step S107: When a task point reaches the preset coverage and balance index thresholds, the area it belongs to is changed from "unknown" to "known", and the price index decay is reset to clear the interference left by historical prices and adapt to the subsequent scheduling strategy evolution. The conditions for area conversion are as follows:

[0166]

[0167] Among them, c is the coverage threshold. b is the balance value threshold.

[0168] Another embodiment of the present invention provides an unmanned swarm social heuristic collaborative crowd intelligence system for emergency rescue, which is used to implement the unmanned swarm social heuristic collaborative crowd intelligence method for emergency rescue, including a basic task-space structure unit, a multi-objective task allocation, a task bidding and initial scheduling unit, a regional status classification unit, a multi-region optimization price difference update unit, a path reconstruction unit, and a reset unit, wherein:

[0169] The basic task-space structure unit is used to construct the capability function C of the unmanned system for the three tasks of detection, fire fighting and rescue. i (t), integrating sensor accuracy, extinguishing agent diffusion, and payload error parameters, combined with environmental modeling, the rescue area is discretized into grid cells to establish a basic task-space structure.

[0170] The multi-objective task allocation is used to dynamically generate a set of task points after the unmanned system identifies the fire point or personnel, and each task point has a task requirement vector An equilibrium task matching model is established based on coverage, capability matching and overlapping penalty terms, and multi-objective task allocation is performed using an auction game mechanism.

[0171] The task bidding and initial scheduling unit is used to construct a multi-factor bidding model consisting of differential coverage gain, capability demand matching function and flight penalty function, and combines the nonlinear pricing mechanism with the price fluctuation limit update rule to complete the task bidding and initial scheduling in path planning.

[0172] The regional status classification unit uses environmental data and model feedback to estimate the fire situation and personnel behavior trends, dynamically updates task demand weights based on the prediction results, and performs regional status classification to improve the adaptability and real-time performance of task scheduling.

[0173] The multi-region optimization price difference update unit is used to divide the task area into unknown areas and known areas, establish non-equilibrium optimization objectives for maximizing coverage and minimizing task requirement satisfaction, and dynamically adjust its matching weights and priorities through evolution functions to construct a multi-region price difference model and multi-region price update rules.

[0174] The path reconstruction unit sets differentiated bidding strategies and price update rules for different regional task points. Bidding in unknown areas is driven by coverage, while bidding in known areas is driven by demand satisfaction. A hybrid auction-ADMM mechanism is used to perform local bidding, global multiplier updates, and scheduling conflict coordination to complete the process path reconstruction.

[0175] The reset unit is used to change the area to which the task point belongs from "unknown" to "known" when the task point reaches the preset coverage and balance index thresholds, and reset the price index attenuation to clear the interference left by historical prices and adapt to the subsequent evolution of scheduling strategies.

[0176] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A social heuristic collaborative group intelligence method for unmanned swarms for emergency rescue, characterized by: The following steps are involved: Step 1: Construct the capability function C of the unmanned system for the three tasks of detection, fire fighting and rescue i (t), integrating relevant parameters and combining environmental modeling to discretize the rescue area into grid cells and establish a basic task-space structure; Step 2: After the unmanned system identifies the fire point or personnel, it dynamically generates a set of task points, establishes an equilibrium task matching model, and uses an auction game mechanism to perform multi-objective task allocation; Step 3: Build a multi-factor bidding model, combining a nonlinear pricing mechanism with price fluctuation limit update rules to complete task bidding and initial scheduling in path planning; Step 4: Use environmental data and model feedback to estimate the fire situation and personnel behavior trends, dynamically update task demand weights based on the prediction results, and perform regional status classification; Step 5: Divide the mission area into unknown and known areas, establish non-equilibrium optimization objectives for maximizing coverage and minimizing mission requirement satisfaction, dynamically adjust matching weights and priorities through evolutionary functions, and construct a multi-region price difference model and multi-region price update rules. Step 6: Set differentiated bidding strategies and price update rules for different regional task points. Bids in unknown areas are driven by coverage, while bids in known areas are driven by demand satisfaction. A hybrid auction-ADMM mechanism is used for local bidding, global multiplier updates, and scheduling conflict coordination to complete process path reconstruction.

2. The unmanned swarm social heuristic collaborative intelligence method for emergency rescue according to claim 1 is characterized by: In step 1, the detection capability function is based on the distance between the unmanned system and the target The angle and sensor performance modeling are specifically expressed as follows: in, is the comprehensive effectiveness of the sensors of the unmanned system i, R d k is the maximum distance that the sensor can recognize, d To control the rate parameter at which detection performance decreases with distance, θ ij is the viewing angle between unmanned system i and target j, θ max is the maximum field of view of the unmanned system sensor, p i is the location of the unmanned system i, The location of the currently detected target; The fire extinguishing capability is modeled based on the extinguishing agent diffusion and wind speed difference and is specifically expressed as: Among them, V i fire is the initial capacity of the fire extinguishing agent of unmanned system i, r0 is the diffusion range of the fire extinguishing agent under windless conditions, v ω -v i is the difference between the real-time wind speed and the speed of unmanned system i, α f is the volatility factor of the extinguishing agent in the air, σ f To allow the wind speed to deviate from the unmanned system, t exe is the time the extinguishing agent remains active in the air; The delivery capability is modeled based on the error propagation function and is specifically expressed as: in, is the current payload mass of the unmanned system i, A j is the range of rescue area j, is the position of the unmanned system p i Location of the delivery destination The distance, σ r The accuracy variance of the rescue supplies delivered to the unmanned system i is specifically expressed as: Among them, h i is the height of the unmanned system i above the ground, is the velocity fluctuation of the unmanned system i during hovering, σ alt is the height measurement error, σ p Provides accurate positioning of the unmanned system.

3. The unmanned swarm social heuristic collaborative intelligence method for emergency rescue according to claim 2 is characterized by: In step 2, the regional coverage structure estimates the coverage rate by dividing the task area into grids and calculating whether each grid is covered. The union coverage rate, overlap penalty rate, and task balance rate together constitute the matching optimization objective, which is specifically expressed as: Among them, λ b ∈[0,1] is the weight for controlling task balance and coverage, U i is the union coverage, O i is the overlap penalty term, B j is the equilibrium rate of a single task point, M is the total number of tasks; Union coverage U i Specifically expressed as: Where K is the number of discretized grid cells, a ik Indicates whether the unmanned system i covers the kth grid; Union Coverage O i Specifically expressed as: Among them, (x) + =max(x,0); Task balance rate B j Specifically expressed as: Among them, x ij ∈{0,1} represents whether unmanned system i is assigned to task point t j , C ik Represented as unmanned system u i The capability of task type k∈{D,F,R} at time t, is the task requirement value of the multi-task, For the task point t j Total supply at capacity k.

4. The unmanned swarm social heuristic collaborative intelligence method for emergency rescue according to claim 3 is characterized by: In step 3, the bidding function b ij Specifically expressed as: in, To cover the gain part, For the task matching part, For the penalty portion of the voyage; Override gain section Specifically expressed as: Among them, ΔU i (j) is the new effective gain, ΔO i (j) is the newly added overlap gain, μ0 is the overlap penalty coefficient, is an indicator function, which is 1 when the condition is met and 0 otherwise; Task matching part Specifically expressed as: Among them, ω k is the capability type weight; Range penalty section Specifically expressed as: Among them, κ c is the quadratic coefficient used to suppress high-risk mission selection close to the range limit, is the maximum operating distance of unmanned system i, q j is the location of the destination point j of the unmanned system, is the initial position of the unmanned system i; Mission point price The dynamic change of follows a nonlinear increasing mechanism, as shown below: in, The current best competitor, is the overlap ratio of the task-associated areas, as shown below: Where K is the number of discretized grid cells.

5. The unmanned swarm social heuristic collaborative intelligence method for emergency rescue according to claim 4 is characterized by: In step 4, the dynamic correction mechanism of the task adjusts the fire extinguishing requirements of the task point based on the predicted results, as shown below: in, is the critical fire intensity for demand surge, σ k ∈[0,1] is the demand sensitivity coefficient, is the fire intensity; The adjustments to the rescue requirements for mission points are as follows: Among them, η R is the rescue demand sensitivity coefficient, n j =|P j | is the number of people affected, P j is located at the task point t j Gathering of people within the radius, is the survival probability of the trapped persons, which is specifically expressed as: Among them, ‖x p (τ)-x nearest-fire (τ)‖ is the Euclidean distance between the fire source and the person, x p is the position of the trapped person, x nearest-fire Indicates the nearest fire source location, L safe is the safety distance threshold, t rescue The time it takes for rescue to arrive. The survival probability can only be calculated before rescue arrives, otherwise it will be reset to zero.

6. The unmanned swarm social heuristic collaborative intelligence method for emergency rescue according to claim 5 is characterized by: In step 5, the task area is divided into unknown areas and known areas. For the unknown area, the objective function is as follows: Among them, A u represents the range of the unknown area, λ u To control the equilibrium weight, For the task point t j The demand for the kth capability; For a known area, the objective function is as follows: Among them, A k represents the range of the known area, λ u To control the equilibrium weight, For the task point t j Current dynamic needs; To be assigned to task point t j The sum of the capabilities; For unknown and known areas, the weights are dynamically adjusted through the evolution function as follows: Among them, κ u is the time attenuation coefficient, v u is the sensitivity coefficient of the change rate of the unknown area, μ k To control the priority of urgent tasks, k is the demand threshold for determining urgent tasks, The priority of the rescue mission; Based on the above weights and the objective functions for the two types of regions, the overall optimization objective function is constructed as follows: Among them, J k is the objective function of the known area, J u is the objective function of the unknown region, ω k (t) is the weight of the known area, ω u (t) is the weight of the unknown area; Using the hybrid auction-ADMM framework, the local updates are as follows: Among them, ρ i is the ADMM penalty factor, z i Global shared variables; u i is the Lagrange multiplier in ADMM, y i Assign variables to the current unmanned system; In the global coordination stage, the alternating multiplier method is used to coordinate the task allocation between unmanned systems and resolve allocation conflicts. The global aggregation and multiplier update formulas are as follows: you i ←u i +y i -z Where N is the number of unmanned systems.

7. The unmanned swarm social heuristic collaborative intelligence method for emergency rescue according to claim 6 is characterized by: Including step 7, the method of step 7 is as follows: when the task point reaches the preset coverage and balance index thresholds, the area to which it belongs is changed from "unknown" to "known", and the price index decay is reset to clear the interference left by historical prices and adapt to the subsequent evolution of scheduling strategies.

8. The unmanned swarm social heuristic collaborative intelligence method for emergency rescue according to claim 7 is characterized by: In step 7, the conditions for region conversion are as follows: in, is the indicator function. When the corresponding value exceeds the threshold, the value of the indicator function is 1, c is the coverage threshold; b is the balance value threshold.

9. A system for implementing the unmanned swarm social heuristic collaborative group intelligence method for emergency rescue as described in claim 1, characterized by: It includes basic task-spatial structure unit, multi-objective task allocation, task bidding and initial scheduling unit, regional status classification unit, multi-region optimization price difference update unit, path reconstruction unit, and reset unit, among which: The basic task-space structure unit is used to construct the capability function C of the unmanned system for the three tasks of detection, fire fighting and rescue. i (t), integrating sensor accuracy, extinguishing agent diffusion, and load error parameters, and combining environmental modeling to discretize the rescue area into grid cells and establish a basic task-space structure; The multi-objective task allocation is used to dynamically generate a set of task points after the unmanned system identifies the fire point or personnel, and each task point has a task requirement vector An equilibrium task matching model is established based on coverage, capability matching and overlapping penalty terms, and an auction game mechanism is used to allocate multi-objective tasks. The task bidding and initial scheduling unit is used to construct a multi-factor bidding model consisting of differential coverage gain, capability demand matching function and flight penalty function, and combines a nonlinear pricing mechanism with a price fluctuation limit update rule to complete task bidding and initial scheduling in path planning; The regional status classification unit uses environmental data and model feedback to estimate the fire situation and personnel behavior trends, dynamically updates task demand weights based on the prediction results, and performs regional status classification to improve the adaptability and real-time performance of task scheduling; The multi-region optimization price difference update unit is used to divide the task area into unknown areas and known areas, establish non-equilibrium optimization objectives for maximizing coverage and minimizing task requirement satisfaction, and dynamically adjust its matching weights and priorities through an evolution function to construct a multi-region price difference model and multi-region price update rules; The path reconstruction unit is used to set differentiated bidding strategies and price update rules for different regional task points. In unknown areas, bidding is driven by coverage, while in known areas, bidding is driven by demand satisfaction. A hybrid auction-ADMM mechanism is used to perform local bidding, global multiplier updates, and scheduling conflict coordination to complete process path reconstruction. The reset unit is used to change the area to which the task point belongs from "unknown" to "known" when the task point reaches the preset coverage and balance index thresholds, and reset the price index attenuation to clear the interference left by historical prices and adapt to the subsequent evolution of scheduling strategies.

Images