Cooperative catastrophe detection method and system for master-slave robot group

Abstract

The invention discloses a catastrophe cooperative detection method for a child-mother robot group, and relates to the technical field of coal mine safety. The catastrophe cooperative detection method comprises the steps that a three-dimensional grid model of a mine space is established, and each node stores the path pheromone concentration Cp, the dangerous pheromone concentration Cd and the exploration pheromone concentration Ce of the corresponding node position; the method comprises the following steps: acquiring environment data by utilizing a sub-robot, and generating a path pheromone increment delta Cp and a danger level pheromone increment delta Cd; acquiring path pheromone concentration Cp and dangerous pheromone concentration Cd of the current node and the adjacent node by using the sub-robot, calculating local path pheromone gradient delta Cp and dangerous pheromone gradient delta Cd, and generating a navigation path according to delta Cp and delta Cd; the sub-robot updates the exploration pheromone concentration Ce passing through the node according to the navigation path; and the mother robot generates a task allocation instruction by using the pheromone update data uploaded by all the child robots. According to the application, the further expansion of the disaster can be prevented, and great help is provided for timely rescue of the accident, disaster avoidance of underground personnel, post-disaster control measures and the like.

Description

Technical Field

[0001] The present application relates to the field of coal mine safety technology, and in particular to a method and system for collaborative disaster detection by a parent-child robot group. Background Art

[0002] Mine safety is crucial for the development of the coal industry. However, mine operating environments are complex and present multiple catastrophic risks. Once a mine disaster occurs, the underground environment undergoes drastic changes, posing a serious threat to the lives of trapped personnel. The first 72 hours after a disaster are the golden rescue window. Rapidly and accurately detecting the disaster environment, locating trapped personnel, and planning safe rescue routes are key to maximizing rescue success rates.

[0003] In recent years, robotics technology has been applied in mine rescue operations. Deploying rescue robots into disaster zones to conduct environmental surveys can effectively reduce risks to rescuers. However, existing mine rescue robots often use pre-existing map-based path planning methods, relying on pre-disaster mine tunnel maps for positioning and navigation.

[0004] A key characteristic of mine disasters is the dramatic change in the structural environment. The shockwave from a gas explosion can cause widespread tunnel collapse, completely blocking existing passageways with rocks, coal, and other debris. Water intrusion can flood tunnels, altering the navigable area. High temperatures from fires can cause the roof to deform and collapse, creating new obstacles. These structural changes can severely deviate from pre-stored mine maps and the actual environment, rendering path planning methods based on these maps ineffective.

[0005] In existing technologies, some solutions attempt to build maps in real time through the sensors of a single robot (SLAM technology), but they face many challenges in mine disaster environments: first, the exploration efficiency of a single robot is low, making it difficult to complete large-scale detection within the golden rescue period; second, smoke and dust in the disaster environment seriously affect visual sensors, and the performance of traditional SLAM algorithms decreases; third, if a single robot fails during the exploration process, the acquired environmental information may be lost.

[0006] Other approaches employ multi-robot collaborative exploration, but most still rely on centralized map construction and path planning, requiring stable communication between robots to share map data. However, communication conditions in mine disaster environments are poor. Factors such as tunnel collapse and metal support structures cause severe wireless signal attenuation, making reliable data transmission difficult. Furthermore, centralized planning methods are computationally complex and struggle to adapt to dynamic environmental changes.

[0007] Therefore, there is an urgent need for a robot swarm collaborative detection method that can adapt to the dynamic environment of mine disasters, does not rely on pre-stored maps, and has distributed autonomous decision-making capabilities, so as to improve the efficiency and reliability of disaster environment detection and provide technical support for mine emergency rescue. Summary of the Invention

[0008] In response to mine disasters that cause changes in environmental structure and render pre-stored maps invalid, the present application provides a method and system for collaborative disaster detection by a group of mother and child robots. The environmental perception data of the child robots is converted into pheromone increments in real time and updated to the corresponding nodes. The pheromone diffusion algorithm is used to form a continuous pheromone field distribution, enabling the child robots to perform map-free autonomous navigation based on local pheromone gradients. At the same time, the mother robot integrates the pheromone update data of all child robots to generate a global pheromone field, and identifies unexplored safe areas based on this to assign tasks, realizing adaptive collaborative exploration in disaster environments without relying on pre-stored maps.

[0009] One aspect of the present application provides a method for collaborative disaster detection by a group of parent-child robots, including: S1, establishing a three-dimensional grid model of the mine space, discretizing the space into grid nodes, and storing the path pheromone concentration C of the corresponding node position at each node. p , danger pheromone concentration C d and exploration pheromone concentration C e ; S2, using the sub-robot to collect environmental data, the environmental data includes position coordinates, passing status and disaster characteristic parameters, and converting the collected environmental data into corresponding pheromone data: when a successful passage is detected, a path pheromone increment ΔC is generated p When disaster characteristics are detected, the hazard level pheromone increment ΔC is generated according to the hazard level. d ; S3, increase the path pheromone by ΔC p and the hazard level pheromone increment ΔC d Write the grid node at the corresponding position, update the pheromone concentration of the corresponding node, and update the pheromone concentration of the adjacent nodes according to the diffusion algorithm; S4, use the sub-robot to obtain the path pheromone concentration C of the current node and the adjacent nodes p and danger pheromone concentration C d , calculate the local path pheromone gradient ▽C p and danger pheromone gradient ▽C d , and according to ▽C p and ▽C d Generate a navigation path; S5, the sub-robot moves according to the navigation path and updates the exploration pheromone concentration C of the node it passes during the movement e ; S6, the mother robot uses the pheromone update data uploaded by all the child robots to update the C of the corresponding node in the three-dimensional grid model p 、C dand C e value, generate global pheromone; S7, the mother robot generates global pheromone according to C e The value identifies the unexplored area, according to C d The value identifies the dangerous area, generates task assignment instructions, and guides the sub-robot to explore C first. e The value is lower than the threshold, and C d Areas with values ​​below the threshold.

[0010] Among them, the mother-and-child robot cluster is a heterogeneous robot group system with a layered architecture, consisting of a mother robot and multiple child robots, which complete complex disaster environment detection tasks through division of labor and cooperation. In this application, the mother robot integrates the pheromone data uploaded by all child robots, generates a global pheromone map, and performs task planning and allocation according to the global situation. The child robots perform specific environmental detection tasks, collect local environmental data, generate and release pheromones, and navigate autonomously according to the local pheromone gradient. The mother-and-child robot cluster realizes a collaborative mode of centralized decision-making and distributed execution, which is particularly suitable for the detection of mine disaster environments and can achieve efficient and safe collaborative detection in complex and dangerous environments.

[0011] Furthermore, the path pheromone concentration C p Indicates the cumulative degree of the sub-robot successfully passing the corresponding node position, C p The higher the value, the more reliable the corresponding path; the concentration of danger pheromone C d Indicates the environmental danger level of the corresponding node location, C d The higher the value, the more serious the disaster threat in the corresponding node area; explore the pheromone concentration C e represents the exploration coverage of the corresponding node, C e A higher value means that the sub-robot visits the corresponding node more frequently.

[0012] Among them, the path pheromone concentration C p It is a digital mark released and accumulated by the robot when it successfully passes a grid node, which is used to replace the "traversable path" information in the traditional map. In a mine disaster environment, the original tunnel may collapse and be blocked, and new channels may be formed due to structural damage. The path information in the pre-stored map is no longer valid. p By recording the successful passage events of the sub-robot in real time, a passability map of the current environment is dynamically constructed - High C p The value node indicates that the location has been successfully passed many times, and the path reliability is high; low C p A node with a value of zero or a zero value indicates that the location has not been verified or is inaccessible. This path marking mechanism based on actual detection results enables the robot swarm to gradually build a dynamic map that reflects the actual path conditions after the disaster without a pre-stored map.

[0013] Danger pheromone concentration C d It is a danger marker released by the robot when it detects disaster characteristics (high temperature, toxic gas, smoke, etc.). It is used to replace the dynamic danger information that cannot be reflected in traditional maps. After a mine disaster, fire may spread, gas may accumulate, and accumulated water may rise. These dynamic dangers cannot be predicted and marked by the pre-stored map. d By quantifying the degree of environmental danger in real time, it provides safe navigation information for robots. d The value node indicates that there is a serious disaster threat in the area and it should be avoided; low C d The value node indicates that it is relatively safe and can be passed. This dynamic danger marking mechanism enables the robot to adjust its navigation strategy according to the real-time environment status and avoid entering newly generated dangerous areas.

[0014] Exploring pheromone concentration C e It is an access mark left by a sub-robot when it passes through a grid node. It is used to record the exploration coverage and replace the traditional centralized exploration management. In a disaster environment with limited communication, it is difficult to maintain a global exploration record, and the pre-stored map cannot indicate which areas need priority search and rescue. e By directly marking the exploration history in the environment, distributed coverage management is achieved - high C e The value node indicates that it has been fully explored, and the low C e The value node indicates insufficient exploration or has not been reached. This environment-embedded exploration record allows the robot that arrives later to read the C e The mother robot can also understand the regional exploration status through C e Distributed identification explores blind spots to ensure that the entire disaster area is fully covered and no possible location of trapped people is missed.

[0015] Furthermore, S2 generates a path pheromone increment ΔC when a successful passage is detected. p When disaster characteristics are detected, the hazard level pheromone increment ΔC is generated according to the hazard level. d , including: obtaining the current position coordinates (x, y, z), speed v and obstacle collision signal of the sub-robot; when v>v min , and there is no collision signal, it is judged as a successful passing state, and the path pheromone increment ΔC is generated p =α p ×f(v), where α p is the path pheromone release intensity coefficient, f(v) is the speed-related function; the sub-robot collects disaster characteristic parameters through sensors, including temperature T, toxic gas concentration G and smoke concentration S. When any parameter exceeds the safety threshold, disaster detection is triggered.

[0016] Calculate the hazard level based on disaster characteristic parameters Among them, w1, w2, w3 are weight coefficients, T max ,G max ,S max is the upper limit of the danger level of the corresponding parameter; the danger pheromone increment ΔC is generated according to the danger level D d =α d ×D, where α d is the intensity coefficient of danger pheromone release;

[0017] Among them, the obstacle collision signal is an alarm signal generated when the sub-robot detects contact or imminent contact with an obstacle through a contact sensor (such as a collision switch, a pressure sensor) or a short-range sensor (such as ultrasound, a lidar). In a mine disaster environment, a previously unobstructed tunnel may generate new obstacles due to collapse, falling rocks, equipment tipping, etc. These obstacles do not exist in the pre-stored map. The collision signal is a real-time criterion for path passability. When a collision signal is detected, it indicates that there is a physical obstacle on the current path and the path pheromone should not be released; on the contrary, there is no collision signal and the normal moving speed (v>v min ) proves that the path is indeed passable and should release path pheromones to mark it. This verification mechanism based on actual passability ensures that path pheromones can accurately reflect the actual passability conditions after the disaster.

[0018] Path pheromone release intensity coefficient α p It is a regulatory parameter that controls the amount of path pheromones released by the sub-robot when it successfully passes a node, and determines the accumulation speed of path reliability information.

[0019] Danger pheromone release intensity coefficient α d It is the regulating parameter that controls the amount of danger pheromones released by the sub-robot when it detects disaster characteristics, and directly affects the marking intensity and warning range of the dangerous area.

[0020] Further, S3, the path pheromone increment ΔC p and the hazard level pheromone increment ΔC d Write the grid node at the corresponding position, update the pheromone concentration of the corresponding node, and update the pheromone concentration of the adjacent nodes according to the diffusion algorithm, including: according to the path pheromone increment ΔC p and the hazard level pheromone increment ΔC d , update the pheromone concentration of the corresponding node using the time series update method; set the path pheromone concentration C p and danger pheromone concentration C d The spatial diffusion range of C p Adopt 6-neighborhood local diffusion, which only diffuses to the 6 adjacent nodes: up, down, left, right, front, and back; Cd The 26-neighborhood omnidirectional diffusion is used to diffuse to all adjacent nodes. For the nodes within the diffusion range, the path pheromone diffusion amount is calculated based on the spatial distance d between the current node and the adjacent node. and the diffusion of danger pheromones Utilize path pheromone diffusion and the diffusion of danger pheromones Update the path pheromone concentration C of adjacent nodes respectively p and danger pheromone concentration C d .

[0021] Furthermore, S4 uses the sub-robot to obtain the path pheromone concentration C of the current node and adjacent nodes p and danger pheromone concentration C d , calculate the local path pheromone gradient and danger pheromone gradients And according to and Generate a navigation path, including: obtaining the path pheromone concentration C of the sub-robot's current node and adjacent nodes p and danger pheromone concentration C d ; Use the finite difference method to calculate the local pheromone concentration of the current node: path pheromone gradient Danger pheromone gradient Among them, the partial derivatives in each direction are calculated by the difference of pheromone concentrations of adjacent nodes; according to and Generate navigation vector V nav ; Among them, the navigation vector V nav It is the direction guidance vector calculated by the sub-robot in real time based on the local pheromone gradient, which is used to indicate the next moving direction without a pre-stored map. In the mine disaster environment, the traditional map-based path planning method fails, and the navigation vector is calculated by integrating the path pheromone gradient. and danger pheromone gradients Provides an autonomous navigation mechanism that adapts to dynamic environments. According to the navigation vector V nav Determine the next target node and generate a navigation path from the current node to the target node.

[0022] In particular, the navigation vector is calculated solely based on the pheromone concentration data of the current node and its neighborhood, without requiring global map information. This local decision-making mechanism enables each sub-robot to make independent navigation decisions even when communication is interrupted and global information is unavailable. When the mine environment changes (such as a new collapse or the spread of fire), the pheromone distribution is updated in real time, and the navigation vector is dynamically adjusted accordingly. For example, if a previously navigable path collapses, the path pheromone in that area will decrease due to volatilization, and the navigation vector will automatically point to another navigable direction; newly emerging dangerous areas will quickly accumulate dangerous pheromones, and the navigation vector will immediately adjust to avoid that direction.

[0023] Furthermore, the navigation vector Among them, δ1 is the path attraction coefficient and δ2 is the hazard repulsion coefficient.

[0024] Furthermore, in S5, the sub-robot moves according to the navigation path and updates the exploration pheromone concentration C of the node it passes during the movement. e , including: the sub-robot moves along the navigation path, and collects the trajectory data of the sub-robot in real time to generate a grid node sequence; reads the current exploration pheromone concentration C of each node in the grid node sequence e Calculate the exploration pheromone increment ΔC based on the nonlinear cumulative function e ; The exploration pheromone increment ΔC e and the current exploration pheromone concentration C e The value is accumulated and the exploration pheromone concentration C of each grid node sequence is updated e .

[0025] Furthermore, S6, generating global pheromone, including: the mother robot receives the pheromone update data uploaded by all the child robots, and the update data includes the path pheromone concentration C in step S3 p and danger pheromone concentration C d , and the exploration pheromone concentration C updated in step S5 e ; Perform time sequence sorting and node clustering on the received pheromone update data; for multiple update data of the same grid node, select C p 、C e and C d The maximum value of is used as the fused pheromone concentration of the corresponding node; the fused pheromone concentration is used to update the pheromone concentration of the corresponding node in the three-dimensional grid model to obtain the global pheromone.

[0026] Furthermore, in S7, the mother robot uses the C e The value identifies the unexplored area, according to C d The value identifies the dangerous area, generates task assignment instructions, and guides the sub-robot to explore C first. eThe value is lower than the threshold, and C d The area with a value lower than the threshold includes: the exploration pheromone concentration C of each grid node extracted from the global pheromone e and danger pheromone concentration C d , build node attribute data set; select C e Less than the preset threshold and C d For nodes whose value is less than a preset threshold, a target node set is constructed; for each node in the target node set, the priority index of each node is calculated according to the multi-objective priority evaluation function; the target nodes are sorted according to the priority index, and combined with the current position and available number of each sub-robot, a task allocation instruction containing the target node coordinates is generated and sent to the corresponding sub-robot.

[0027] Multi-objective priority evaluation function:

[0028] Among them, w e is the exploration weight, w s is the safety weight, C e,max and C d,max are the maximum values ​​of exploration pheromone and danger pheromone, respectively.

[0029] Another aspect of the present application also provides a coordinated disaster detection system for a group of mother-and-child robots, including: a model building module, which establishes a three-dimensional grid model of the mine space, discretizes the space into grid nodes, and stores the path pheromone concentration C of the corresponding node position at each node. p , danger pheromone concentration C d and exploration pheromone concentration C e The environmental perception module is set in the sub-robot to collect environmental data, including position coordinates, passing status and disaster characteristic parameters, and convert the collected environmental data into corresponding pheromone data: when a successful passage is detected, a path pheromone increment ΔC is generated. p When disaster characteristics are detected, the hazard level pheromone increment ΔC is generated according to the hazard level. d ;Pheromone update module, the path pheromone increment ΔC p and the hazard level pheromone increment ΔC d Write the grid node at the corresponding position, update the pheromone concentration of the corresponding node, and update the pheromone concentration of the adjacent nodes according to the diffusion algorithm;

[0030] Gradient navigation module, set in the sub-robot, obtains the path pheromone concentration C of the current node and adjacent nodes p and danger pheromone concentration C d , calculate the local path pheromone gradient ▽C p and danger pheromone concentration ▽C d , and according to ▽Cp and ▽C d Generate a navigation path; the exploration recording module is set in the sub-robot, controls the sub-robot to move according to the navigation path, and updates the exploration pheromone concentration C of the node passed during the movement e ; The global fusion module is set in the mother robot, receives the pheromone update data uploaded by all child robots, and updates the C of the corresponding nodes in the three-dimensional grid model. p 、C d and C e value, generating global pheromone; the task assignment module is set in the mother robot, according to the C in the global pheromone e The value identifies the unexplored area, according to C d The value identifies the dangerous area, generates task assignment instructions, and guides the sub-robot to explore C e The value is lower than the threshold, and C d Areas with values ​​below the threshold.

[0031] Compared with the existing technology, the advantages of this application are:

[0032] This application fundamentally solves the problem of failure of pre-stored maps in mine disaster environments by converting the environmental perception data of the sub-robots into spatially distributed pheromone fields in real time. Specifically, each sub-robot releases a path pheromone increment ΔC when it successfully passes a path. p And update to the corresponding grid node to form the path pheromone concentration C p The spatial distribution of the hazard pheromone; when the disaster characteristics are detected, the increase of the danger pheromone ΔC d , forming the danger pheromone concentration C d This pheromone update mechanism allows the accessibility and danger of the environment to be encoded into the three-dimensional grid model in real time, and the pheromone field itself becomes a dynamically updated "living map". When a mine collapses and blocks the original passage, the blocked path will not produce new path pheromones because no robots pass through it. p The value remains low; the newly formed navigable path will accumulate path pheromones due to the successful passage of the robot, C p The value gradually increases. The subsequent sub-robots calculate the path pheromone gradient ▽C p Get pointed towards high C p The navigation direction of the value area, natural selection has been verified to be passable path, and at the same time through the danger pheromone gradient ▽C d This pheromone gradient-based distributed navigation mechanism completely relies on no pre-existing maps and can adapt to changes in environmental structure caused by disasters in real time, achieving true map-free autonomous navigation. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The present application will be further described in the form of exemplary embodiments, which will be described in detail with reference to the accompanying drawings. These embodiments are not limiting, and in these embodiments, the same numbers represent the same structures, wherein:

[0034] Figure 1 is an exemplary flow chart of a method for collaborative disaster detection by a mother-and-child robot group according to some embodiments of the present application;

[0035] Figure 2 is an exemplary flow chart of generating path pheromones according to some embodiments of the present application;

[0036] Figure 3 is an exemplary flow chart of obtaining an increment of danger pheromone according to some embodiments of the present application;

[0037] Figure 4 is an exemplary flow chart of updating node pheromone concentration according to some embodiments of the present application. DETAILED DESCRIPTION

[0038] The method and system provided in the embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0039] This application is used for large-scale coal mine rescue. The mine emergency rescue command center platform embeds the parent-child robot group disaster collaborative detection method of this application, deploying one mother robot and multiple child robots with different functions to perform disaster environment detection tasks.

[0040] Among them, the command level is: rescue command center (manual decision-making) → combat command platform (human-machine collaboration) → mother robot (autonomous coordination) → sub-robot group (distributed execution); sub-robot grouping: 4 heavy-duty pathfinder robots (strong obstacle crossing capability), 4 environmental detection robots (multi-sensors), and 4 fast reconnaissance robots (high maneuverability).

[0041] like Figure 1 As shown in S1, a three-dimensional grid model of the mine space is established, and the space is discretized into grid nodes. Each node stores the path pheromone concentration C of the corresponding node position. p , danger pheromone concentration C d and exploration pheromone concentration C e ,include:

[0042] S1-1, obtain the spatial dimension parameters of the mine, including length L, width W and height H, and divide the mine space into cubic grid cells;

[0043] S1-2, establish a node data structure for each grid cell. Each node contains spatial coordinates (x, y, z) and three pheromone concentration storage variables: path pheromone concentration C p , danger pheromone concentration C d and exploration pheromone concentration C e Path pheromone concentration C p Indicates the cumulative degree of the sub-robot successfully passing the corresponding node position, C p The higher the value, the more reliable the corresponding path; the concentration of danger pheromone C d Indicates the environmental danger level of the corresponding node location, C d The higher the value, the more serious the disaster threat in the corresponding node area; explore the pheromone concentration C e represents the exploration coverage of the corresponding node, C e A higher value indicates that the sub-robot visits the corresponding node more frequently;

[0044] S1-3, the initial path pheromone concentration C of all nodes p and danger pheromone concentration C d Set to 0 to explore the pheromone concentration C e Set to the preset initial value C e0 , indicating that all areas are initially unexplored;

[0045] S1-4, establish adjacency relationships between nodes, and establish connection relationships between each node and its 26 adjacent nodes for the execution of the pheromone diffusion algorithm in step S3;

[0046] S1-5, create a node index mapping table, map the spatial coordinates (x, y, z) to the corresponding node storage address, so that the sub-robot in step S2 can quickly locate the corresponding grid node according to the position coordinates; initialize the global pheromone database to store the pheromone concentration values ​​of all nodes, and establish a data update interface for writing pheromone incremental data in step S3 and for the mother robot to update the global pheromone in step S6.

[0047] S2, when a successful passage is detected, a path pheromone increment ΔC is generated p When disaster characteristics are detected, the hazard level pheromone increment ΔC is generated according to the hazard level. d ,include:

[0048] S2-1, obtain the current position coordinates (x, y, z), velocity v and obstacle collision signal of the sub-robot, and locate it to the corresponding grid node according to the three-dimensional grid model established in step S1;

[0049] like Figure 2 As shown in S2-2, when v>v within the preset time interval Δt minIf there is no collision signal, it is judged as a successful passing state and the path pheromone increment ΔC is generated. p =α p ×f(v), where: velocity-related function β is the speed influence coefficient, and its value range is [0.5, 2]. max is the maximum speed of the robot; in particular, after a mine collapse, the original passage may be partially blocked, and it is impossible to determine whether the path is truly passable based on position information alone. By setting the speed threshold v min The collision-free condition ensures that path pheromones are only released along paths that the robot can successfully traverse at normal speeds. Furthermore, in this application, the speed-dependent function f(v) reflects that wide passages traversed at high speeds gain more path pheromone increments, while narrow passages traversed with difficulty gain less pheromone increments. This allows the pheromone field to distinguish between traversable paths of varying quality.

[0050] Path pheromone release intensity coefficient Among them, C p,base is the basic release intensity, the value range is [10, 50], η is the group coordination coefficient, N is the number of currently active sub-robots, N max is the total number of sub-robots. Specifically, when multiple sub-robots explore simultaneously, the intensity of pheromone release from each robot increases accordingly, accelerating the accumulation of pheromones along reliable paths and enabling newly discovered traversable paths to be more quickly identified and utilized by the group. This mechanism is particularly important in the early stages of a disaster—when a large number of robots explore simultaneously, they can quickly establish a pheromone field network that reflects the current state of the environment.

[0051] S2-3, the sub-robot collects disaster characteristic parameters through sensors, including temperature T, toxic gas concentration G and smoke concentration S. When T>T safe or G>G safe or S>S safe When the disaster detection is triggered;

[0052] like Figure 3 As shown in S2-4, the hazard level is calculated based on the disaster characteristic parameters Among them, w1, w2, w3 are normalized weight coefficients and w1+w2+w3=1, T max ,G max ,S max is the upper limit of danger for the corresponding parameter;

[0053] S2-5, generating a danger pheromone increment ΔC according to the danger level D d =α d ×D, where the danger pheromone release intensity coefficient α d =C d,base ×exp(λ×D), Cd,base is the basic danger pheromone intensity, ranging from [20, 100], and λ is the danger amplification coefficient, ranging from [1, 3]. Among them, the exponential amplification mechanism α of the danger pheromone release intensity d =C d,base ×exp(λ×D), which makes the pheromone concentration in high-risk areas increase exponentially, ensuring that the dangerous information can generate a strong enough "repulsion field" to effectively prevent subsequent robots from entering the high-risk area.

[0054] S2-6, grid node address, path pheromone increment ΔC p and the danger pheromone increment ΔC d The data is packaged into pheromone data for use in step S3 to update the pheromone concentration of the corresponding node.

[0055] like Figure 4 As shown, S3, ΔC p and ΔC d Write the grid node at the corresponding position, update the pheromone concentration of the corresponding node, and update the pheromone concentration of the adjacent nodes according to the diffusion algorithm, including:

[0056] S3-1, update the path pheromone concentration of the grid node where the child robot is currently located to C p (t+1)=C p (t)×(1-ρ p )+ΔC p , the concentration of danger pheromone is updated to C d (t+1)=C d (t)×(1-ρ d )+ΔC d , where ρ p and ρ d are the volatility coefficients of path pheromone and danger pheromone, respectively, with a value range of [0.01, 0.1];

[0057] S3-2, determine the diffusion range of pheromone, for path pheromone C p The 6-neighborhood diffusion mode is used, which only diffuses to the 6 adjacent nodes up, down, left, right, front, and back. d A 26-neighborhood diffusion model is used to rapidly propagate hazard information to all adjacent nodes. Path pheromones represent the robot's actual trajectory, and their diffusion should remain in the primary direction of movement to avoid misleading "false paths." Therefore, limiting diffusion to a 6-neighborhood model ensures accurate path guidance. Danger pheromones simulate the diffusion of hazardous substances like gas and smoke, which diffuse in all directions in real environments. A 26-neighborhood diffusion model better reflects the actual propagation characteristics of catastrophic hazards.

[0058] S3-3, calculate pheromone diffusion, path pheromone diffusion Danger pheromone diffusion Among them, κ p ,κ d is the diffusion coefficient, d is the Euclidean distance between nodes, σ p ,σ d is the diffusion attenuation factor; specifically, the path pheromone diffusion coefficient κ p The value range is [0.1, 0.3], the diffusion coefficient of danger pheromone κ d The value range is [0.4, 0.8], the path pheromone diffusion attenuation factor σ p The value range is [1.0, 2.0], the danger pheromone diffusion attenuation factor σ d The value range is [3.0,6.0].

[0059] S3-4, taking into account the characteristics of the mine environment, the vertical diffusion of the danger pheromone is corrected: when the diffusion direction is upward, the diffusion amount is multiplied by the amplification factor μ up =1.5, simulating the rising characteristics of toxic gases and smoke; when the diffusion direction is downward, the diffusion amount is multiplied by the attenuation coefficient μ down =0.7; The density of gas in the mine is less than that of air, and it has the characteristic of gathering upward; the high-temperature smoke generated by the fire will also diffuse upward first. By setting the upward diffusion amplification coefficient μ up =1.5 and downward attenuation coefficient μ down =0.7, which makes the spatial distribution of danger pheromones consistent with the distribution pattern of actual catastrophic substances.

[0060] S3-5, update the pheromone concentration of adjacent nodes:

[0061] And set the upper limit of pheromone concentration C p,max and C d,max , to prevent excessive accumulation of pheromones;

[0062] Preferably, in step S3-6, when the danger pheromone increment ΔC is detected d >C d,threshold When , the emergency diffusion mode is triggered, and the diffusion range is expanded to the 2-hop neighborhood, and the diffusion coefficient κ d Increased to 2 times, ensuring that dangerous information is quickly transmitted to more sub-robots and achieving rapid response to disaster warnings.

[0063] S4-1, the sub-robot reads the path pheromone concentration C of the current node and all adjacent nodes in its 26 neighborhoods from the 3D grid model updated in step S3. p and danger pheromone concentration C dThis application avoids the dependency of traditional path planning on a complete map, and each sub-robot can still make independent navigation decisions even in the event of communication interruption.

[0064] S4-2, use the finite difference method to calculate the local pheromone gradient of the current node and the path pheromone gradient Danger pheromone gradient The partial derivatives in each direction are calculated by the difference of pheromone concentrations of adjacent nodes; in particular, the path pheromone gradient Pointing to the direction of the fastest increase in pheromone concentration, naturally pointing to a verified passable path; danger pheromone gradient It points to directions of increasing danger, providing navigational information for risk avoidance. As the mine environment changes, the pheromone gradient field dynamically adjusts accordingly. New collapsed areas, impassable and lacking path pheromones, naturally guide the robot around them. Newly discovered passages accumulate pheromones as the robot passes through them, gradually pointing the gradient toward a new path.

[0065] S4-3, constructing composite navigation vector Among them, δ1 is the path attraction coefficient, which ranges from [0.6 to 1.0]; δ2 is the danger repulsion coefficient, which is dynamically adjusted according to the danger pheromone concentration of the current node: in, is the basic rejection coefficient, the value range is [0.8, 1.2], is the reference hazardous concentration, The value range is [0.3C d,max ,0.5C d,max By following the path pheromone gradient and avoiding the danger pheromone gradient, the robot achieves a navigation strategy that seeks benefits and avoids risks. Specifically, in low-risk areas, path attraction dominates, and the robot prioritizes finding the optimal path. In high-risk areas, the danger repulsion coefficient increases, and risk avoidance becomes the primary goal. This dynamic balance mechanism enables the robot to find the optimal trade-off between exploration efficiency and safety.

[0066] Preferably, S4-4, considering the structural constraints of the mine tunnel, the navigation vector V nav Conduct feasibility test: When there are obstacles in the navigation direction, select nav The passable direction with the smallest angle is used as the corrected navigation direction;

[0067] S4-5, according to the navigation vector V nav Determine the next target node: select the node with the navigation vector V in the reachable neighborhood of the current node nav The grid node closest in direction is used as the next target node, and a navigation path is generated from the current node to the target node.

[0068] S5-1, the sub-robot determines the target moving direction according to the navigation path generated in step S4, and converts the navigation vector V nav Convert it into actual movement control instructions to control the sub-robot to move towards the target grid node;

[0069] S5-2, record the movement trajectory of the sub-robot, and obtain the grid node sequence {n1,n2,.....,n k};

[0070] S5-3, for each grid node n in the moving trajectory i , read the current exploration pheromone concentration C of the node e (n i ,t);

[0071] S5-4, according to the exploration pheromone concentration C e Indicates the exploration coverage of the corresponding node and updates the exploration pheromone concentration of the node: C e (n i ,t+1)=C e (n i ,t)+ΔC e , where ΔC e To explore the increment of pheromone,

[0072] α e To explore the pheromone release intensity coefficient, the value range is 10 to 20; is the saturation concentration value, generally taken as 100%;

[0073] S5-5, through an exponential decay function Realize the nonlinear accumulation of exploration pheromone, when the node C e When the value is low, ΔC e Larger, it encourages coverage of under-explored areas; when C e Value close to When ΔC e Approaching 0, avoiding repeated visits to fully explored areas;

[0074] S5-6, the updated exploration pheromone concentration C e (n i ,t+1) writes the corresponding grid node to complete the exploration coverage record during the movement of the sub-robot.

[0075] In particular, through C e The spatial distribution of values ​​realizes implicit coordination among multiple robots. The exploration behavior of each child robot will improve the C eThis information is perceived by other robots through the pheromone field. In step S7, the mother robot identifies the low C e The pheromone-based coordination mechanism ensures complete coverage of the disaster area without requiring complex task negotiation, thus avoiding missing critical areas.

[0076] S6, the mother robot uses the pheromone update data uploaded by all the child robots to update the C of the corresponding node in the three-dimensional grid model p ,C d ,C e Value, generate global pheromone, including:

[0077] S6-1, the mother robot receives the pheromone update data set {U1, U2, ..., U m}, where each updated data U i Contains grid node coordinates (x, y, z), update timestamp t i , and the path pheromone concentration C of the node p , danger pheromone concentration C p and exploration pheromone concentration C e In a mine disaster environment, a single child robot has a limited field of view and can only obtain local information. However, multiple child robots can explore in a decentralized manner and cover different areas in parallel. The mother robot aggregates all local information to construct a global view. This mechanism maintains the exploration efficiency of the child robots while effectively integrating global information.

[0078] S6-2, update the received pheromone data set according to the timestamp t i Sort in ascending order to get the time series data sequence {U1',U2',......,U m A mine disaster is a dynamic process—fires can spread, water can rise, and gas can diffuse. Temporal sorting ensures causal consistency in information updates, allowing the global pheromone graph to reflect the latest state of the environment.

[0079] S6-3, cluster the time series data according to the grid node coordinates, group the update data with the same coordinates (x, y, z) into a group, and form a node update data group set {G1, G2,....., G n}, where each G j Contains multiple update data points to the same grid node;

[0080] S6-4, update data group G for each node j , extract the pheromone concentration values ​​of all paths respectively

[0081] {C p1 ,C p2 ,.....,C pk Danger pheromone concentration value {C d1 ,C d2 ,.....,C dk} and explore pheromone concentration values

[0082] {C e1 ,C e2 ,.....,C ek};

[0083] S6-5, update data group G for each node j , using the maximum selection strategy for data fusion: fusion path pheromone concentration Fusion danger pheromone concentration Fusion Exploration Pheromone Concentration Among them, for danger pheromone C d , choosing the maximum value means "better to believe it than not" - as long as one sub-robot detects danger, the information will be retained, avoiding the underestimate of danger caused by individual sensor failure or detection omission. p and Exploration Pheromone C e ,The maximum strategy also ensures that verified traversable paths and completed ,explorations will not be lost due to data conflicts.

[0084] S6-6, locate the corresponding node in the three-dimensional grid model according to the node coordinates (x, y, z), and update the fused pheromone concentration value to the node:

[0085] S6-7, after traversing all node update data groups, a global pheromone graph is generated that includes the path pheromone distribution, danger pheromone distribution and exploration pheromone distribution of the entire mine space. The global pheromone graph reflects the path reliability, danger level and exploration coverage of the mine space, providing a global decision-making basis for the task allocation in step S7.

[0086] S7, the mother robot according to the C in the global pheromone e The value identifies the unexplored area, according to C d The value identifies the dangerous area, generates task assignment instructions, and guides the sub-robot to explore C first. e The value is lower than the threshold, and C d Areas with values ​​below the threshold, including:

[0087] S7-1, the mother robot extracts the exploration pheromone concentration C of all grid nodes from the global pheromone generated in step S6 e and danger pheromone concentration C d;

[0088] S7-2, set the exploration threshold C e,threshold =β e ×C e,avg , where β e is the exploration coefficient, the value range is [0.3, 0.5], C e,avg The average value of the pheromone concentration is explored for all nodes; all nodes in the three-dimensional grid model are traversed and the pheromone concentration C is explored. e Indicates the exploration coverage of the corresponding node, identifying C e <C e,threshold The node set of is regarded as the unexplored area;

[0089] S7-3, according to the concentration of danger pheromone Cd, the environmental danger level of the corresponding node position is represented, and the danger threshold C is set. d,threshold =β d ×(C d,min +σ d ), where β d is the safety factor, the value range is [1.5, 2.0], C d,min is the minimum value of the danger pheromone concentration of all nodes, σ d is the standard deviation of the concentration of danger pheromone; identify C d <C d,threshold The node set is used as a safe area;

[0090] S7-4, perform security screening on the nodes in the unexplored area and eliminate C d ≥C d,threshold Nodes, generate a set of target nodes to be explored, and the target nodes to be explored also meet C e <C e,threshold And C d <C d,threshold ;

[0091] S7-5, calculate the priority index for each node in the exploration target node set Among them, w e is the exploration weight, w s is the safety weight, C e,max and C d,max are the maximum values ​​of exploration pheromone and danger pheromone, respectively;

[0092] S7-6, sort the target nodes to be explored according to the priority index P, and allocate the nodes with higher priority first;

[0093] S7-7, according to the current number of available sub-robots N and the position of each sub-robot, a distance-based greedy allocation algorithm is used: for each target node to be assigned, the Euclidean distance from all sub-robots with unassigned tasks to the node is calculated, and the sub-robot with the closest distance is selected for assignment; a maximum of K target nodes are assigned to each sub-robot, where M is the total number of target nodes to be explored; a task assignment instruction containing a sequence of target node coordinates is generated;

[0094] S7-8, send the task assignment instructions to the corresponding sub-robot, each instruction contains the target node coordinate sequence {(x1,y1,z1),(x2,y2,z2),.....,(x k ,y k ,z k )}, guiding the child robots to visit the assigned target nodes in sequence, achieving preferential coverage of unexplored safe areas. In particular, the task instructions only contain a sequence of target coordinates, rather than traditional path planning results. This design fully utilizes the child robot's autonomous navigation capability based on pheromone gradients - the mother robot only needs to specify "where to go", and the child robot will solve "how to get there" by itself through local pheromone gradients. This decoupling of task allocation and path planning enables the system to achieve effective global coordination even in the absence of a map. Even if the environment changes during the execution of the task, the child robot can adjust its path through real-time pheromone gradients, demonstrating its strong environmental adaptability.

[0095] S7-9, the mother robot uploads the generated task allocation plan to the combat command platform, forming a three-level scheduling system: autonomous layer: the child robots navigate autonomously according to the pheromone gradient and adjust their paths when encountering sudden obstacles; coordination layer: the mother robot monitors the progress of task execution, and automatically deploys nearby idle robots for support when it detects that the task completion rate of a child robot is lower than expected; decision-making layer: the combat command platform displays the overall situation and task allocation plan, and the commander can intervene in real time, such as directly issuing priority coverage instructions when a new key search area is discovered.

[0096] The invention of the present application and its implementation methods are described schematically above. This description is not restrictive. Without departing from the spirit or basic features of the present application, the present application can be implemented in other specific forms. What is shown in the accompanying drawings is only one of the implementation methods of the invention of the present application, and the actual structure is not limited to this. Therefore, if a person of ordinary skill in the art is inspired by it, without departing from the purpose of the invention, a structural method and embodiment similar to the technical solution are designed without creativity, which should all fall within the scope of protection of the present application. In addition, the word "including" does not exclude other elements or steps, and the word "one" before an element does not exclude the inclusion of "multiple" elements. Words such as first and second are used to indicate names and do not indicate any specific order.

Claims

1. A method for collaborative disaster detection by a mother-and-child robot group, characterized in that: include: S1, establish a three-dimensional grid model of the mine space, discretize the space into grid nodes, and each node stores the path pheromone concentration C of the corresponding node position p , danger pheromone concentration C d and exploration pheromone concentration C e ; S2, using the sub-robot to collect environmental data, which includes position coordinates, passing status and disaster characteristic parameters, and converting the collected environmental data into corresponding pheromone data: when a successful passage is detected, a path pheromone increment ΔC is generated. p When disaster characteristics are detected, the hazard level pheromone increment ΔC is generated according to the hazard level. d ; S3, increase the path pheromone by ΔC p and the hazard level pheromone increment ΔC d Write the grid node at the corresponding position, update the pheromone concentration of the corresponding node, and update the pheromone concentration of the adjacent nodes according to the diffusion algorithm; S4, use the sub-robot to obtain the path pheromone concentration C of the current node and adjacent nodes p and danger pheromone concentration C d , calculate the local path pheromone gradient ▽C p and danger pheromone gradient ▽C d , and according to ▽C p and ▽C d Generate navigation paths; S5, the sub-robot moves according to the navigation path and updates the exploration pheromone concentration C of the node it passes during the movement. e ; S6, the mother robot uses the pheromone update data uploaded by all the child robots to update the C of the corresponding node in the three-dimensional grid model p 、C d and C e value, generate global pheromone; S7, the mother robot according to the C in the global pheromone e The value identifies the unexplored area, according to C d The value identifies the dangerous area, generates task assignment instructions, and guides the sub-robot to explore C first. e The value is lower than the threshold, and C d Areas with values ​​below the threshold.

2. The method for coordinated disaster detection using a mother-and-child robot group according to claim 1, characterized in that: Path pheromone concentration C p Indicates the cumulative degree of the sub-robot successfully passing the corresponding node position, C p A higher value indicates a more reliable corresponding path; Danger pheromone concentration C d Indicates the environmental danger level of the corresponding node location, C d The higher the value, the more serious the disaster threat in the corresponding node area; Exploring pheromone concentration C e represents the exploration coverage of the corresponding node, C e A higher value means that the sub-robot visits the corresponding node more frequently.

3. The method for coordinated disaster detection using a mother-and-child robot group according to claim 2, wherein: S2, when a successful passage is detected, a path pheromone increment ΔC is generated p When disaster characteristics are detected, the hazard level pheromone increment ΔC is generated according to the hazard level. d ,include: Get the sub-robot's current position coordinates (x, y, z), speed v, and obstacle collision signal; When v>v min , and there is no collision signal, it is judged as a successful passing state, and the path pheromone increment ΔC is generated p =α p ×f(v), where α p is the path pheromone release intensity coefficient, f(v) is the speed-related function; The sub-robot collects disaster characteristic parameters through sensors, including temperature T, toxic gas concentration G, and smoke concentration S. When any parameter exceeds the safety threshold, disaster detection is triggered; Calculate the hazard level based on disaster characteristic parameters Among them, w1, w2, w3 are weight coefficients, T max ,G max ,S max is the upper limit of danger for the corresponding parameter; Generate danger pheromone increment ΔC according to danger level D d =α d ×D, where α d is the intensity coefficient of danger pheromone release.

4. The method for coordinated disaster detection using a mother-and-child robot swarm according to claim 2, wherein: S3, increase the path pheromone by ΔC p and the hazard level pheromone increment ΔC d Write the grid node at the corresponding position, update the pheromone concentration of the corresponding node, and update the pheromone concentration of the adjacent nodes according to the diffusion algorithm, including: According to the path pheromone increment ΔC p and the hazard level pheromone increment ΔC d , the pheromone concentration of the corresponding node is updated using a time-series update method; Set the path pheromone concentration C respectively p and danger pheromone concentration C d The spatial diffusion range of C p Adopt 6-neighborhood local diffusion, which only diffuses to the 6 adjacent nodes: up, down, left, right, front, and back; C d Use 26-neighborhood omnidirectional diffusion to spread to all adjacent nodes; For nodes within the diffusion range, the path pheromone diffusion amount is calculated based on the spatial distance d between the current node and the adjacent node. and the diffusion of danger pheromones Using path pheromone diffusion and the diffusion of danger pheromones Update the path pheromone concentration C of adjacent nodes respectively p and danger pheromone concentration C d .

5. The method for coordinated disaster detection using a mother-and-child robot group according to claim 2, wherein: S4, according to and Generate navigation paths, including: Get the path pheromone concentration C of the sub-robot's current node and adjacent nodes p and danger pheromone concentration C d ; The finite difference method is used to calculate the local pheromone concentration of the current node: Path pheromone gradient Danger pheromone gradient Among them, the partial derivatives in each direction are calculated by the difference of pheromone concentrations of adjacent nodes; according to and Generate navigation vector V nav ; According to the navigation vector V nav Determine the next target node and generate a navigation path from the current node to the target node.

6. The method for coordinated disaster detection using a mother-and-child robot group according to claim 5, characterized in that: Navigation vector Among them, δ1 is the path attraction coefficient and δ2 is the hazard repulsion coefficient.

7. The method for coordinated disaster detection using a mother-and-child robot group according to claim 2, wherein: S5, the sub-robot moves according to the navigation path and updates the exploration pheromone concentration C of the node it passes during the movement. e ,include: The sub-robot moves along the navigation path, and its trajectory data is collected in real time to generate a sequence of grid nodes it passes through; Read the current exploration pheromone concentration C of each node in the grid node sequence e value; Calculation of exploration pheromone increment ΔC based on nonlinear cumulative function e ; The exploration pheromone increment ΔC e and the current exploration pheromone concentration C e The value is accumulated and the exploration pheromone concentration C of each grid node sequence is updated e .

8. The method for coordinated disaster detection by a mother-and-child robot swarm according to any one of claims 2 to 7, characterized in that: S6, generating global pheromones, including: The mother robot receives the pheromone update data uploaded by all child robots, and the update data includes the path pheromone concentration C in step S3. p and danger pheromone concentration C d , and the exploration pheromone concentration C updated in step S5 e ; Perform time series sorting and node clustering on the received pheromone update data; For multiple update data of the same grid node, select C p 、C e and C d The maximum value of is taken as the fusion pheromone concentration of the corresponding node; The fused pheromone concentration is used to update the pheromone concentration of the corresponding node in the three-dimensional grid model to obtain the global pheromone.

9. The method for coordinated disaster detection using a mother-and-child robot group according to claim 8, wherein: S7, the mother robot according to the C in the global pheromone e The value identifies the unexplored area, according to C d Identify hazardous areas and generate task allocation instructions, including: Extract the exploration pheromone concentration C of each grid node from the global pheromone e and danger pheromone concentration C d , build node attribute dataset; Select C e Less than the preset threshold and C d Nodes smaller than the preset threshold are used to build a target node set; For each node in the target node set, the priority index of each node is calculated according to the multi-objective priority evaluation function; The target nodes are sorted according to the priority index, and the task assignment instructions containing the target node coordinates are generated based on the current position and available number of each sub-robot, and then sent to the corresponding sub-robot; Multi-objective priority evaluation function: Among them, w e is the exploration weight, w s is the safety weight, C e,max and C d,max are the maximum values ​​of exploration pheromone and danger pheromone, respectively.

10. A coordinated disaster detection system for a mother-and-child robot group, characterized in that: include: The model building module builds a three-dimensional grid model of the mine space, discretizes the space into grid nodes, and stores the path pheromone concentration C of the corresponding node position at each node. p , danger pheromone concentration C d and exploration pheromone concentration C e ; The environmental perception module is set in the sub-robot to collect environmental data, including position coordinates, passing status and disaster characteristic parameters, and convert the collected environmental data into corresponding pheromone data: when a successful passage is detected, a path pheromone increment ΔC is generated. p When disaster characteristics are detected, the hazard level pheromone increment ΔC is generated according to the hazard level. d ; The pheromone update module increases the path pheromone increment ΔC p and the hazard level pheromone increment ΔC d Write the grid node at the corresponding position, update the pheromone concentration of the corresponding node, and update the pheromone concentration of the adjacent nodes according to the diffusion algorithm; Gradient navigation module, set in the sub-robot, obtains the path pheromone concentration C of the current node and adjacent nodes p and danger pheromone concentration C d , calculate the local path pheromone gradient and danger pheromone concentrations And according to and Generate navigation paths; The exploration recording module is set in the sub-robot to control the sub-robot to move according to the navigation path and update the exploration pheromone concentration C of the node passed during the movement. e ; The global fusion module is set in the mother robot, receives the pheromone update data uploaded by all the child robots, and updates the C of the corresponding nodes in the three-dimensional grid model. p 、C d and C e value, generate global pheromone; The task allocation module is set in the mother robot, according to the C in the global pheromone e The value identifies the unexplored area, according to C d The value identifies the dangerous area, generates task assignment instructions, and guides the sub-robot to explore C e The value is lower than the threshold, and C d Areas with values ​​below the threshold.

Images