A method for tunnel emergency rescue structure detection based on a UAV group

Abstract

The application discloses a kind of tunnel emergency rescue structure detection methods based on unmanned aerial vehicle group, comprising the following steps: control detection unmanned aerial vehicle enters danger tunnel and flies along planned path;Control communication unmanned aerial vehicle enters danger tunnel, dynamically adjusts the position of each communication unmanned aerial vehicle, to build and maintain a dynamic mobile ad hoc network communication link;Through the micro data processor carried by detection unmanned aerial vehicle, the data collected are processed, to identify and mark abnormal area data as effective data;Detection unmanned aerial vehicle will effective data be transmitted in real time to ground control station through mobile ad hoc network communication link;Ground control station analyzes the data received, constructs tunnel structure model, assesses danger, and generates rescue plan;Recycle unmanned aerial vehicle group, and synchronize the full amount of raw data stored in detection unmanned aerial vehicle local to ground control station.The application has the advantages that: communication reliability and adaptability are improved;Detection accuracy and efficiency are improved;Decision response speed is improved.

Description

Technical Field

[0001] This invention relates to the technical field of tunnel emergency rescue detection, and in particular to a method for tunnel emergency rescue structure detection based on a drone swarm. Background Technology

[0002] With the rapid development of infrastructure construction in my country, as of 2024, the total length of tunnels nationwide had exceeded 30,000 kilometers, with over 7,000 kilometers under construction, playing a crucial role in transportation, water conservancy, mining, and other fields. However, due to complex geological conditions and the enclosed and narrow construction environment, tunnel projects are prone to sudden emergencies such as collapses, water inrushes, and mudslides. Delayed detection often leads to secondary disasters, causing significant economic losses and personnel safety risks.

[0003] Current traditional tunnel emergency rescue detection methods mainly rely on manual inspection, traditional drones, or inspection robots in conjunction with manual detection. While existing air-space-ground joint communication solutions attempt to build aerial channels through drone swarms, they still depend on a fixed number of relay points combined with wired ground equipment. This approach is unsuitable for scenarios with dynamic signal attenuation and randomly distributed obstacles within tunnels, making communication prone to interruptions. Furthermore, detection relies on high-altitude scanning and manual on-site data collection, resulting in a limited scope and requiring personnel to enter hazardous areas, leading to low efficiency and high safety risks. Currently, for tunnels with signal gaps or complex conditions, problems such as unstable communication, low detection efficiency, poor accuracy, and high safety risks exist. Therefore, there is an urgent need for a tunnel emergency rescue structural detection method that combines high-precision positioning, multi-source data acquisition, highly reliable communication, and rapid decision support to address the shortcomings of traditional methods. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by providing a method for tunnel emergency rescue structure detection based on a drone swarm. This method utilizes multiple communication drones paired with detection drones to establish communication links within the tunnel through a self-organizing network, enabling data transmission with the ground. No prior hardware deployment is required; the communication drones can establish a network immediately upon startup, meeting the "every second counts" requirement of emergency rescue. The signal coverage can be dynamically adjusted according to the flight path of the detection drones, ensuring high-speed data transmission even deep within the tunnel. An integrated "network-detection-analysis-decision" system is constructed, including communication drones, detection drones, a ground control station, a positioning system, data processing and analysis software, a security system, and a rescue decision support system. The detection drones utilize "high-definition cameras + infrared imaging..." This method uses "sound and ultrasonic ranging" to collect multi-dimensional data; a communication drone relays data to the ground control station in real time; software automatically constructs a tunnel structure model, assists in identifying hidden dangers and risks; a safety assurance system monitors the drone's flight status in real time, automatically triggering warnings or returning to base when abnormalities occur to avoid equipment damage; and returns to base for charging and battery swapping when the battery is low. It features autonomous networking, wide coverage, high reliability, and strong flexibility. Through a standardized and automated process: base station deployment → release of detection drone → release of communication drone and automatic networking → line exploration → automatic data collection → real-time transmission → intelligent analysis → automatic warning → drone recovery, it achieves "unmanned" detection: no personnel need to enter the tunnel throughout the process, reducing the risks of collapse and toxic gases; data collection is done without human intervention, reducing subjective errors and ensuring data objectivity. This method solves the problems of existing communication solutions requiring pre-set relay points, being unable to cope with dynamic signal attenuation scenarios in tunnels, and requiring ground equipment supplementation, resulting in poor adaptability; and existing detection solutions having a single dimension, requiring full data transmission to the backend for processing, leading to high latency, and relying on manual identification of hazards, resulting in low efficiency and high safety risks.

[0005] The objective of this invention is achieved through the following technical solutions: A method for tunnel emergency rescue structure detection based on unmanned aerial vehicle (UAV) swarms, the method comprising the following steps: S1: Deploy a signal base station outside the entrance of the tunnel shaft of the dangerous tunnel, and deploy an unmanned aerial vehicle (UAV) field and a ground control station on the ground; wherein, the UAV field is equipped with multiple communication UAVs and at least one detection UAV, forming a UAV swarm; S2: Control the detection drone to enter the dangerous tunnel and fly along the planned path; The communication drone is controlled to enter the dangerous tunnel, and the position of each communication drone is dynamically adjusted based on the signal strength returned by the detection drone, so as to build and maintain a dynamic mobile ad hoc network communication link; wherein, the position adjustment of the communication drone is controlled by a dynamic relay node optimization algorithm. S3: The tunnel structure is scanned using a laser scanner mounted on the detection drone, and scan data is collected; The collected data is processed by the micro data processor carried by the detection drone to identify and mark abnormal area data as valid data, while the data in normal areas is de-densified. S4: The detection drone transmits the effective data to the ground control station in real time through the mobile ad hoc network communication link; when a tunnel hazard location is detected, hazard characteristic data to characterize the hazard is transmitted simultaneously; S5: The ground control station analyzes the received data, constructs a tunnel structure model, assesses the potential dangers, and generates an emergency response plan; S6: After the detection mission is completed, the drone swarm is recovered, and all the raw data stored locally on the detection drones is synchronized to the ground control station.

[0006] In step S2, the execution process of the dynamic relay node optimization algorithm includes: a. Signal prediction: Based on the signal strength measurement value at the previous moment P prev The signal strength prediction value at the next moment is calculated using the Kalman filter algorithm. P pred This enables short-term prediction of signal strength; among which, P pred The calculation formula is: P pred = AP prev + Bu + w , A Here is the state transition matrix. B To control the input matrix, u For motion control variables, w This is process noise; b. Movement decision: When real-time signal strength P Signal strength below the first threshold or the predicted value at the next time step P pred When the signal level drops below the second threshold, the communication drone prepares to move; based on the signal attenuation gradient Δ P / Δ t Determine the priority of the moves, where Δ P Δ is the change in signal strength. t The distance is a time-varying quantity; the optimal spacing between the communication UAV and the reconnaissance UAV is calculated based on the logarithmic distance path loss model. d opt ,in,d opt The calculation formula is: d opt =10^(( P tx - P req - L 0 ) / (10 n )), P tx For the transmission power of communication drones, P req For the target signal strength, L 0 For reference distance loss, n The path loss index within the tunnel; based on the optimal spacing d opt Current actual distance d current The difference Δ d To adjust the movement distance and speed; c. Path planning and obstacle avoidance: An improved A* algorithm is used to generate the movement path, where the improved A* algorithm employs a signal strength cost function. f ( n ), f ( n The formula for calculating ) is: f ( n )= g ( n )+ h ( n )+ k * p ( n ), g ( n ) represents the distance cost from the current node to the starting node. h ( n ) represents the heuristic cost from the current node to the target node. p ( n ) represents the signal strength cost of the current node. k The weighting coefficients are used; when obstacles exist on the planned path, multiple alternative detour paths are generated, and the signal strength cost function is selected. f ( n The shortest path; if detouring would cause the signal strength to fall below a tolerance threshold, then a cluster coordination mechanism is activated: the adjacent communication drone moves to the other side of the obstacle to build a temporary cross-obstacle relay link.

[0007] The dynamic relay node optimization algorithm is implemented collaboratively by a signal prediction module, an environment modeling module, a path planning module, and a dynamic adjustment module. The signal prediction module executes the signal prediction step. The environment modeling module collects surrounding environmental data in real time using a laser scanner mounted on the communication UAV to construct a local 3D grid map to identify obstacles. The path planning module executes the path planning and obstacle avoidance steps. The dynamic adjustment module determines and outputs the movement direction of the communication UAV based on the outputs of the signal prediction module and the path planning module. θ Distance of movement d Movement speed v and posture adjustment angle α .

[0008] The first threshold is -80dBm, and the second threshold is -75dBm; when Δ P / Δ t When the speed is <-5dBm / s, it is judged as medium-strong attenuation, and the moving speed is... v =1.5m / s, attitude adjustment angle α ≤15° / s; when -5dBm / s≤Δ P / Δ t When the speed is less than -1 dBm / s, it is considered weak attenuation, and the moving speed is... v =0.8m / s, attitude adjustment angle α ≤10° / s; when Δ d When >5m, v =1.5m / s; when 2m≤Δ d When ≤5m, v =0.8m / s; when Δ d When <2m, v =0.3m / s.

[0009] The dynamic relay node optimization algorithm also implements cluster coordination and fault tolerance mechanisms, including: 1) Load balancing: Based on the signal coverage overlap and remaining battery power of each of the communication drones, relay tasks are dynamically allocated; when the data transmission load of a certain communication drone exceeds the load threshold or its remaining battery power is lower than the battery power threshold, some of its relay tasks are automatically transferred to the adjacent communication drones. 2) Fault tolerance: When a fault is detected in one of the communication drones in the cluster, the adjacent communication drones are controlled to quickly move to the faulty node location to fill the gap, according to the preset replacement priority rules.

[0010] In step S3, when the laser scanner detects that the distance change of the tunnel segment exceeds the structural deformation warning threshold, it automatically triggers other sensors carried by the detection drone to collect data in the corresponding area in a coordinated manner.

[0011] The other sensors are one or more of the following: high-definition camera, infrared imager, and ultrasonic rangefinder.

[0012] The hazard characteristic data includes one or more of the following: spatial coordinates of the hazard location, geometric shape data, visual image data, and temperature distribution data.

[0013] The ground control station integrates a positioning system, data processing and analysis software, a safety assurance system, and a disaster relief decision support system. The positioning system is used to accurately locate the tunnel hazard. The data processing and analysis software is used to process the received data, construct a tunnel structure model, and analyze the hazard. The safety assurance system is used to monitor the UAV's attitude, battery level, and distance to obstacles in real time, and issue early warning signals to ensure equipment safety. The disaster relief decision support system is used to generate a disaster relief plan based on data analysis results and a built-in expert opinion database and hazard handling case database.

[0014] The advantages of this invention are: (1) Improved communication reliability and adaptability: Through the UAV swarm network, the dynamic relay node optimization algorithm and frequency hopping communication technology, compared with the traditional static relay with fixed relay and ground supplementation in the existing technology, the signal transmission stability in the deep tunnel is improved, which effectively solves the problem of signal loss detection in the long tunnel under construction, ensures the smooth flow of data link and high communication reliability; the whole process is unmanned, avoiding personnel from entering the dangerous environment, and the multi-machine collaboration improves the exploration efficiency; (2) Improved detection accuracy and efficiency: Based on the detection drone equipped with multi-source sensor acquisition equipment and front-end data preprocessing, compared with the single high-altitude scanning and manual acquisition in the existing technology, the detection dimension is expanded from "image recognition" to "image + temperature + size", which can perceive potential tunnel structure deformation and hidden dangers, accurately locate the location of tunnel hazards, make the tunnel exploration perception results richer, and at the same time, the data transmission efficiency is greatly improved, effectively avoiding full data delay; (3) Improved decision response speed: By using the UAV network to link the ground control station system in real time, end-to-end automation from data collection and processing to risk assessment is realized, which greatly shortens the time for risk assessment and provides an immediate and intuitive scientific basis for emergency decision-making, making the decision response faster. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the steps of the method for tunnel emergency rescue structure detection based on UAV swarms according to the present invention; Figure 2 This is a schematic diagram of the working scenario of the method for tunnel emergency rescue structure detection based on UAV swarm according to the present invention; like Figures 1-2 As shown in the figure, the labels represent: 1. Signal base station; 2. Communication drone; 3. Detection drone; 4. Unmanned aerial vehicle (UAV) airport; 5. Tunnel in danger; 51. Tunnel shaft; 52. Tunnel segment; 53. Location of tunnel hazard; 6. Ground control station; 61. Positioning system; 62. Data processing and analysis software; 63. Safety assurance system; 64. Emergency rescue decision support system. Detailed Implementation

[0016] The features and other related features of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments, so as to facilitate understanding by those skilled in the art: Example: Figures 1-2 As shown, this embodiment relates to a method for tunnel emergency rescue structure detection based on a drone swarm. The method mainly includes the following steps: S1: A signal base station 1 is deployed outside the entrance of the tunnel shaft 51 of the dangerous tunnel 5, and an unmanned aerial vehicle (UAV) field 4 and a ground control station 6 are deployed on the ground. The UAV field 4 is equipped with multiple communication UAVs 2 (3 to 4 communication UAVs 2 are configured in this embodiment) and at least one detection UAV 3 (one detection UAV 3 is configured in this embodiment), forming a UAV swarm.

[0017] In this embodiment, the surrounding geological conditions of the tunnel where the hazard occurred are complex, the construction environment is enclosed and narrow, and there is a lack of signal in the internal space. Therefore, a signal base station 1 needs to be deployed at a suitable location outside the tunnel to establish a core communication link between the ground control station 6 and the UAV cluster. Specifically, the signal base station 1 is deployed in an open area within a certain distance of the tunnel shaft 51 entrance, connecting power and network. The internal environment of the tunnel where the hazard occurred is complex, and manual detection faces communication difficulties and the risk of secondary disasters. UAVs can autonomously network, have wide coverage, high reliability, and strong flexibility, compensating for the shortcomings of traditional detection methods. Therefore, an UAV airport 4 is deployed, equipped with communication UAVs 2 and detection UAVs 3.

[0018] Signal base station 1, serving as the ground communication hub and control signal transmitter, undertakes the core function of two-way data interaction between the ground control station and the UAV. Communication UAV 2, as the core relay node of the entire network, undertakes data relay and network coordination functions between the signal base station, the detection UAV, and the ground control station. It can execute dynamic relay node optimization algorithms to achieve the dual goals of signal stability and obstacle avoidance, possessing strong data transmission and communication relay capabilities, stable flight performance, long endurance, and strong anti-interference capabilities. Detection UAV 3, as the core data acquisition unit, is equipped with a laser scanner, high-definition camera, infrared imager, ultrasonic rangefinder, and micro data processor. It is responsible for deep penetration into the hazardous tunnel 5 to collect structural data and detect hazards. Combined with a multi-source collaborative detection mechanism and tunnel environment adaptive algorithm, it balances flexibility and multi-source perception capabilities. The UAV airport 4 is mainly responsible for the storage, charging, and scheduling of communication UAV 2 and detection UAV 3. Hazardous Tunnel 5 represents the application scenario, including situations involving collapses, large-scale water leakage, and extensive structural deformation in enclosed and narrow underground projects such as long tunnels under construction and underground utility tunnels. The main environmental characteristics are the lack of communication signals, insufficient lighting inside the tunnel, and potential risks such as falling rocks, toxic gases, and water accumulation. Traditional detection equipment cannot function properly in these environments. Hazardous Tunnel 5 is connected to Tunnel Shaft 51, which serves as the access point for the communication drone 2 and the detection drone 3 to enter Hazardous Tunnel 5. Ground Control Station 6 is the command center and data processing core of the entire system, integrating full functions of "control-analysis-decision-early warning." Ground Control Station 6 integrates a positioning system 61, data processing and analysis software 62, a safety assurance system 63, and a rescue decision support system 64. The positioning system 61 is an important component of the networking technology between the ground control station 6 and the UAV. It adopts a multi-mode positioning technology of "inertial navigation + vision + laser scanning" to achieve accurate positioning of the detection UAV 3 and the location of the tunnel hazard 53. The data processing and analysis software 62 is a core component of the ground control station 6. It uses a "point cloud reconstruction + AI recognition" algorithm to automatically process and analyze the received multi-source data. The safety assurance system 63 provides safety warnings by receiving real-time data such as UAV attitude, battery level, and distance to obstacles to ensure equipment safety. The emergency response decision support system 64 has powerful data mining and analysis capabilities. It has a built-in expert opinion database and hazard handling case database. Based on the data analysis results and hazard images, combined with expert opinions and hazard handling case database, it provides risk assessment, automatically generates emergency response plans, and helps the team quickly formulate efficient and feasible emergency response plans.

[0019] S2: Control the detection drone 3 to enter the dangerous tunnel 5 and fly along the planned path; control the communication drone 2 to enter the dangerous tunnel 5, and dynamically adjust the position of each communication drone 2 based on the signal strength returned by the detection drone 3, so as to build and maintain a dynamic mobile ad hoc network communication link; wherein, the position adjustment of the communication drone 2 is controlled by the dynamic relay node optimization algorithm.

[0020] In this embodiment, by acquiring terrain data and basic data of the hazardous tunnel 5, including tunnel design information such as axis coordinates, cross-sectional dimensions, and preliminary hazardous reports, a BIM model is generated, and a detection path is generated based on the BIM model.

[0021] Before the drone swarm enters the hazardous tunnel 5, the communication drone 2 is debugged to test the frequency hopping communication switching efficiency, signal relay delay ≤100ms, and the stability of the dynamic relay node optimization algorithm. If the signal is abnormal or the algorithm execution fails, the position of the signal base station 1 is adjusted. If the signal is normal, the communication drone 2 and the detection drone 3 enter the hazardous tunnel 5 in a network to conduct detection. The detection drone 3 enters the hazardous tunnel 5 first along the BIM planned path, and 3-4 communication drones 2 follow in the "spacing dynamically adjusted according to signal and obstacles" mode driven by the "dynamic relay node optimization algorithm". The detection drone 3 provides real-time feedback on its position and signal status, and the communication drone 2 calculates the optimal movement path through the algorithm.

[0022] The dynamic relay node optimization algorithm maintains the communication link signal strength ≥ -80dBm in real time while avoiding obstacles such as tunnel segments and falling rocks. The algorithm consists of four core units: a signal prediction module, an environment modeling module, a path planning module, and a dynamic adjustment module. The input parameter is the real-time position of the detection drone 3. x , y , z ), Current location of communication drone 2 ( x’ , y’ , z’ Real-time signal strength P Laser scanning obstacle coordinates ( x i , y i , z i ), communication drone 2 remaining battery power E The output parameter is: direction of movement. θ Distance of movement d Movement speed v Posture adjustment angle α .

[0023] The execution process of the dynamic relay node optimization algorithm includes: a. Signal prediction: Based on the signal strength measurement value at the previous moment P prev The signal strength prediction value at the next moment is calculated using the Kalman filter algorithm. P pred This enables short-term prediction of signal strength; among which, P pred The calculation formula is: P pred = AP prev + Bu + w , A Here is the state transition matrix. B To control the input matrix, u For motion control variables, w This represents process noise. Specifically, the prediction window is 500ms.

[0024] b. Movement decision: When real-time signal strength P Signal strength below the first threshold or the predicted value at the next time step P pred When the signal strength drops below the second threshold, the communication drone 2 prepares to move to prevent the signal from falling below the threshold. This is based on the signal attenuation gradient Δ. P / Δ t Determine the priority of the moves, where Δ P Δ is the change in signal strength. t This represents the change over time. Specifically, the first threshold is -80 dBm, and the second threshold is -75 dBm; when Δ P / Δ t When the speed is less than -5dBm / s, it is considered medium to strong attenuation. Prioritize signal strength and movement speed. v =1.5m / s, attitude adjustment angle α ≤15° / s; when -5dBm / s≤Δ P / Δ t When the speed is less than -1 dBm / s, it is considered as weak attenuation. The balance between signal strength and power consumption should be considered, along with the moving speed. v =0.8m / s, attitude adjustment angle α ≤10° / s. The optimal distance between the communication UAV 2 and the reconnaissance UAV 3 is calculated based on the logarithmic distance path loss model. d opt ,in, d opt The calculation formula is: d opt =10^(( P tx - P req -L 0 ) / (10 n )), P tx For the transmission power of communication drones, P req For the target signal strength, L 0 For reference distance loss, n The path loss index within the tunnel is dynamically adjusted based on the tunnel cross-sectional dimensions; based on the optimal spacing. d opt Current actual distance d current The difference Δ d To adjust the movement distance and speed. Specifically, when Δ d When the distance is greater than 5m, move quickly. v =1.5m / s; when 2m≤Δ d When the distance is ≤5m, move at a constant speed. v =0.8m / s; when Δ d When the distance is less than 2m, make minor adjustments. v =0.3m / s.

[0025] c. Path planning and obstacle avoidance: The environmental modeling module uses a laser scanner mounted on the communication drone 2 to collect surrounding environmental data in real time, constructing a local 3D grid map (grid size 0.5m × 0.5m × 0.5m) and marking obstacle grids (grids with an occupancy rate ≥ 70% are considered obstacles). The path planning module uses an improved A* algorithm (a direct search method for finding the shortest path in a static road network) to generate the movement path. The improved A* algorithm uses a signal strength cost function. f ( n ), f ( n The formula for calculating ) is: f ( n )= g ( n )+ h ( n )+ k * p ( n ), g ( n ) represents the distance cost from the current node to the starting node. h ( n (Optimal distance) from the current node to the target node d opt The heuristic cost of the corresponding position. p ( n ) represents the signal strength cost of the current node, whenP When <-80dBm, p ( n )=5, when -80dBm≤ P When <-75dBm, p ( n )=1, when P When ≥-75dBm, p ( n )=0, k This is a weighting coefficient, ranging from 1.2 to 1.5. When obstacles exist on the planned path, multiple alternative detour paths are generated, and the signal strength cost function is selected. f ( n The algorithm selects the path with the shortest distance; if detouring would cause the signal strength to fall below a tolerance threshold, a cluster coordination mechanism is activated: the adjacent communication drone 2 moves to the other side of the obstacle to build a temporary cross-obstacle relay link. Specifically, when there is an obstacle on the optimal path, the algorithm automatically generates three alternative paths (detouring to the left, detouring to the right, and detouring above), and calculates the total cost of each path. f ( n Choose the path with the lowest cost; if taking a detour causes further signal attenuation ( P <-85dBm), initiate cluster collaboration mechanism: send collaboration request to neighboring communication drone 2, neighboring drone moves to the other side of the obstacle, establishes a "cross-obstacle relay link", and restores the original relay link structure after the current drone has circled to the target position; if the obstacle range is large (≥5m), the algorithm automatically adjusts the optimal spacing. d opt (Within ±1m) Find the second-best location with "signal support and no obstacles", and simultaneously report back to ground control station 6, marking the obstacle area for subsequent emergency response reference.

[0026] d. Cluster collaboration and fault tolerance mechanisms: 1) Load balancing calculation: Based on the signal coverage overlap and remaining battery power of each communication drone 2, relay tasks are dynamically allocated. When the data transmission load of a communication drone 2 exceeds the load threshold or its remaining battery power falls below the battery power threshold, some of its relay tasks are automatically transferred to adjacent communication drone 2. Specifically, relay tasks are allocated based on "signal coverage overlap + remaining battery power". When the load of a communication drone 2 exceeds 80% or its data transmission volume exceeds 30%, the algorithm automatically transfers some of its relay tasks to adjacent communication drone 2 to avoid overloading a single device.

[0027] 2) Fault tolerance: When a communication drone 2 in the cluster is detected to have malfunctioned, the algorithm controls adjacent communication drones 2 to quickly move to the location of the malfunctioning node to fill the gap, based on a preset replacement priority rule. Specifically, when a communication drone 2 malfunctions, the algorithm initiates "rapid replacement of relay nodes." Adjacent communication drones 2 quickly move to the location of the malfunctioning node according to a preset replacement priority (closest distance > strongest signal > most powerful battery), maintaining the continuity of the relay chain. During the replacement process, the signal strength fluctuation is ≤3dBm. The communication drone 2 receives the position and signal status feedback from the detection drone 3 in real time, dynamically matching a collaborative mode of "1 detection drone + 3~4 communication drones" to avoid communication interruptions caused by the malfunction of a single communication drone 2.

[0028] In addition, when there are interferences such as residual electromagnetic signals from construction or abnormal geological magnetic fields in the dangerous tunnel 5, the communication drone 2 automatically switches between 2.4GHz and 5.8GHz dual frequency bands through "frequency hopping communication technology" to avoid interference sources. It can achieve anti-interference communication independently without relying on existing ground wired transmission channels or portable communication vehicle-mounted supplements.

[0029] S3: The probe drone 3 scans the tunnel structure using a laser scanner and collects the scan data; the probe drone 3 processes the collected data using a micro data processor to identify and mark abnormal area data as valid data, while reducing the density of data in normal areas.

[0030] In this embodiment, when the laser scanner detects that the distance change of the tunnel segment 52 (the difference between the actual distance to the inner wall of the tunnel segment 52 measured in real time by the laser scanner and the preset reference distance) exceeds the structural deformation warning threshold, it automatically triggers other sensors carried by the detection drone 3 to collect data in the corresponding area. These other sensors include a high-definition camera, an infrared imager, and an ultrasonic rangefinder. The laser scanner, high-definition camera, infrared imager, and ultrasonic rangefinder form a multi-source sensor linkage triggering and cross-verification system. Specifically, a "master sensor leading + slave sensor linkage" mechanism is adopted. The laser scanner acts as the master sensor. When it detects that the distance change of the tunnel segment 52 exceeds the structural deformation warning threshold, it automatically triggers slave sensor linkage: the high-definition camera activates "local magnification shooting mode" to focus on the deformed area, while the infrared imager activates "temperature difference precise scanning mode," and the ultrasonic rangefinder simultaneously collects distance data from three points around the deformed area, forming a four-dimensional data set of "deformation + image + temperature + multi-point distance." This avoids misjudgment by a single sensor. When the optical scan is interfered with by dust, the ultrasonic rangefinder... The system employs cross-validation of data, including a high-definition camera with a resolution of no less than 16 million pixels to capture high-definition images of the tunnel's internal lining structure and the location of potential hazards; an infrared imager with a temperature measurement range of -20℃ to 150℃ to identify conditions such as water leakage, abnormal high temperatures, and fires within the tunnel; and a laser scanner with an accuracy of no less than ±2mm and an effective detection distance of no less than 10m. This allows for the safe flight positioning and adjustment of the detection drone 3 through distance measurement, preventing the drone from colliding with walls. The distance between the communication drone 2 and the detection drone 3 is determined based on signal strength, while simultaneously enabling drone obstacle avoidance and tunnel segment scanning measurements.

[0031] The micro data processor can automatically reduce the data density of the raw data collected by the detection equipment through a preset "abnormal feature recognition algorithm". For data in normal areas such as flat lining and no temperature anomalies, the density is reduced (sampling density is reduced); for data in abnormal areas such as cracks, water leakage, and structural deformation, all sampling points are retained and marked as "valid data" and transmitted to the communication drone 2 first, thereby obtaining "valid data" with a significantly compressed volume. This prevents all data from being transmitted to the ground control station 6 for processing. If the collection of high-definition video, point cloud data, etc., is delayed due to insufficient transmission bandwidth, it will affect the efficiency of emergency decision-making. This avoids the delay of more than 30 minutes caused by the full data transmission to the backend in the existing technology. This process is completed at the detection drone 3.

[0032] In addition, the detection drone 3 flies along the detection path, setting up a monitoring node at regular intervals (e.g., 5m) along the tunnel axis. It hovers at the monitoring node, and the laser scanner first performs a full-range scan. If a distance change of ≥±3mm is detected, it automatically triggers high-definition camera local magnification shooting, infrared imaging precise temperature measurement, and ultrasonic multi-point ranging to form a four-dimensional data set. At the same time, the laser scanning identifies the distance between the drone swarm and the tunnel wall. The ground control station 6 has a built-in safety protection system 63 that combines the environmental data of the detection drone 3 to provide real-time warnings of collision risks.

[0033] S4: The detection drone 3 transmits effective data to the ground control station 6 in real time through the mobile ad hoc network communication link; when the location of the tunnel hazard 5 is detected, the hazard characteristic data used to characterize the hazard is transmitted simultaneously.

[0034] In this embodiment, the hazard characteristic data includes the spatial coordinates of the hazard location, geometric shape data, visual image data, and temperature distribution data.

[0035] S5: Ground control station 6 analyzes the received data, constructs a tunnel structure model, assesses potential hazards, and generates an emergency response plan.

[0036] In this embodiment, the ground control station 6 receives multi-source "valid data" of the tunnel interior collected by the detection drone 3, and marks the coordinates of the tunnel hazard location 53 according to the positioning system 61. The data processing and analysis software 62 automatically performs in-depth processing and analysis on the received multi-source data, constructing a detailed model of the tunnel structure using tunnel imagery and point cloud data. Simultaneously, it analyzes the tunnel structure deformation at other locations besides the hazard location, sensing potential tunnel structural deformation and hidden dangers. The safety assurance system 63 monitors the drone swarm's flight status in real time, issuing timely warning signals to ensure drone flight safety. The data processed by the data processing and analysis software 62 and the tunnel imagery of the hazard location determined by the positioning system 61 are simultaneously transmitted to the emergency response decision support system 64. Based on the severity of the hazard, historical emergency response cases, and expert opinions, the system provides the emergency response team with an efficient and feasible emergency response plan.

[0037] S6: After the detection mission is completed, recover the drone swarm and synchronize all the raw data stored locally on the detection drone 3 to the ground control station 6.

[0038] In this embodiment, after the exploration is completed, multiple communication drones 2 and exploration drones 3 are recovered (exploration drones 3 are recovered first because they cache the full amount of raw data that was not transmitted in real time; multiple communication drones 2 are recovered simultaneously afterward). During the recovery process, the drone attitude is monitored in real time to avoid collisions. When exploration drones 3 return to the ground, they can transmit the full amount of data to the ground control station 6 for complete data analysis. The data is stored on a local server and in a cloud database, labeled with the tunnel name, exploration time, and equipment number for easy traceability later. Exploration drones 3 have built-in storage cards for local storage, caching the full amount of raw data in real time. After flying out of the tunnel, the data is synchronized to the ground control station 6 via a high-speed wireless link, ensuring the dual requirements of "rapid decision-making based on effective data + subsequent analysis of full data".

[0039] The beneficial technical effects of this embodiment are as follows: (1) Improved communication reliability and adaptability: Through the UAV swarm network, the dynamic relay node optimization algorithm and frequency hopping communication technology, compared with the traditional static relay with fixed relay and ground supplementation in the existing technology, the signal transmission stability in the deep tunnel is improved, which effectively solves the problem of signal loss detection in the long tunnel under construction, ensures the smooth flow of data link and high communication reliability; the whole process is unmanned, avoiding personnel from entering the dangerous environment, and the multi-machine collaboration improves the exploration efficiency; (2) Improved detection accuracy and efficiency: Based on the detection drone equipped with multi-source sensor acquisition equipment and front-end data preprocessing, compared with the single high-altitude scanning and manual acquisition in the existing technology, the detection dimension is expanded from "image recognition" to "image + temperature + size", which can perceive potential tunnel structure deformation and hidden dangers, accurately locate the location of tunnel hazards, make the tunnel exploration perception results richer, and at the same time, the data transmission efficiency is greatly improved, effectively avoiding full data delay; (3) Improved decision response speed: By using the UAV network to link the ground control station system in real time, end-to-end automation from data collection and processing to risk assessment is realized, which greatly shortens the time for risk assessment and provides an immediate and intuitive scientific basis for emergency decision-making, making the decision response faster.

Claims

1. A method for structural detection in tunnel emergency rescue based on unmanned aerial vehicle (UAV) swarms, characterized in that... The method includes the following steps: S1: Deploy a signal base station outside the entrance of the tunnel shaft of the dangerous tunnel, and deploy an unmanned aerial vehicle (UAV) field and a ground control station on the ground; wherein, the UAV field is equipped with multiple communication UAVs and at least one detection UAV, forming a UAV swarm; S2: Control the detection drone to enter the dangerous tunnel and fly along the planned path; The communication drone is controlled to enter the dangerous tunnel, and the position of each communication drone is dynamically adjusted based on the signal strength returned by the detection drone, so as to build and maintain a dynamic mobile ad hoc network communication link; wherein, the position adjustment of the communication drone is controlled by a dynamic relay node optimization algorithm. The dynamic relay node optimization algorithm includes a movement decision process: When real-time signal strength P Signal strength below the first threshold or the predicted value at the next time step P pred When the signal level drops below the second threshold, the communication drone prepares to move; based on the signal attenuation gradient Δ P / Δ t Determine the priority of the moves, where Δ P Δ is the change in signal strength. t It is a quantity that changes over time; The optimal spacing between the communication UAV and the reconnaissance UAV was calculated based on the logarithmic distance path loss model. d opt ,in, d opt The calculation formula is: d opt =10^(( P tx - P req - L 0 ) / (10 n )), P tx For the transmission power of communication drones, P req For the target signal strength, L 0 For reference distance loss, n The path loss index within the tunnel; based on the optimal spacing d opt Current actual distance d current The difference Δ d To adjust the movement distance and speed; S3: The tunnel structure is scanned using a laser scanner mounted on the detection drone, and scan data is collected; The collected data is processed by the micro data processor carried by the detection drone to identify and mark abnormal area data as valid data, while the data in normal areas is de-densified. S4: The detection drone transmits the effective data to the ground control station in real time through the mobile ad hoc network communication link; when a tunnel hazard location is detected, hazard characteristic data to characterize the hazard is transmitted simultaneously; S5: The ground control station analyzes the received data, constructs a tunnel structure model, assesses the potential dangers, and generates an emergency response plan; S6: After the detection mission is completed, the drone swarm is recovered, and all the raw data stored locally on the detection drones is synchronized to the ground control station.

2. The method for tunnel emergency rescue structure detection based on UAV swarm as described in claim 1, characterized in that... In step S2, the execution process of the dynamic relay node optimization algorithm includes: a. Signal prediction: Based on the signal strength measurement value at the previous moment P prev The signal strength prediction value at the next moment is calculated using the Kalman filter algorithm. P pred This enables short-term prediction of signal strength; among which, P pred The calculation formula is: P pred = AP prev + Bu + w , A Here is the state transition matrix. B To control the input matrix, u For motion control variables, w This is process noise; b. Moving decision; c. Path planning and obstacle avoidance: Adopting improved A The algorithm generates movement paths, where the improved A The algorithm uses a signal strength cost function. f ( n ), f ( n The formula for calculating ) is: f ( n )= g ( n )+ h ( n )+ k p ( n ), g ( n ) represents the distance cost from the current node to the starting node. h ( n ) represents the heuristic cost from the current node to the target node. p ( n ) represents the signal strength cost of the current node. k The weighting coefficients are used; when obstacles exist on the planned path, multiple alternative detour paths are generated, and the signal strength cost function is selected. f ( n The shortest path; if detouring would cause the signal strength to fall below a tolerance threshold, then a cluster coordination mechanism is activated: the adjacent communication drone moves to the other side of the obstacle to build a temporary cross-obstacle relay link.

3. The method for tunnel emergency rescue structure detection based on UAV swarm as described in claim 2, characterized in that... The dynamic relay node optimization algorithm is implemented collaboratively by a signal prediction module, an environment modeling module, a path planning module, and a dynamic adjustment module. The signal prediction module performs the signal prediction step. The environment modeling module collects surrounding environmental data in real time using a laser scanner mounted on the communication UAV to construct a local 3D grid map to identify obstacles. The path planning module performs the path planning and obstacle avoidance steps. The dynamic adjustment module determines and outputs the movement direction of the communication UAV based on the outputs of the signal prediction module and the path planning module. θ Distance of movement d Movement speed v and posture adjustment angle α .

4. The method for tunnel emergency rescue structure detection based on UAV swarm as described in claim 2, characterized in that... The first threshold is -80dBm, and the second threshold is -75dBm; when Δ P / Δ t When the speed is <-5dBm / s, it is judged as medium-strong attenuation, and the moving speed is... v =1.5m / s, attitude adjustment angle α ≤15° / s; when -5dBm / s≤Δ P / Δ t When the speed is less than -1 dBm / s, it is considered weak attenuation, and the moving speed is... v =0.8m / s, attitude adjustment angle α ≤10° / s; when Δ d When >5m, v =1.5m / s; when 2m≤Δ d When ≤5m, v =0.8m / s; when Δ d When <2m, v =0.3m / s.

5. The method for tunnel emergency rescue structure detection based on UAV swarm as described in claim 2, characterized in that... The dynamic relay node optimization algorithm also implements cluster coordination and fault tolerance mechanisms, including: 1) Load balancing: Based on the signal coverage overlap and remaining battery power of each of the communication drones, relay tasks are dynamically allocated; when the data transmission load of a certain communication drone exceeds the load threshold or its remaining battery power is lower than the battery power threshold, some of its relay tasks are automatically transferred to the adjacent communication drones. 2) Fault tolerance: When a fault is detected in one of the communication drones in the cluster, the adjacent communication drones are controlled to quickly move to the faulty node location to fill the gap, according to the preset replacement priority rules.

6. The method for tunnel emergency rescue structure detection based on UAV swarm as described in claim 1, characterized in that... In step S3, when the laser scanner detects that the distance change of the tunnel segment exceeds the structural deformation warning threshold, it automatically triggers other sensors carried by the detection drone to collect data in the corresponding area in a coordinated manner.

7. The method for tunnel emergency rescue structure detection based on UAV swarm as described in claim 6, characterized in that... The other sensors are one or more of the following: high-definition camera, infrared imager, and ultrasonic rangefinder.

8. The method for tunnel emergency rescue structure detection based on UAV swarm as described in claim 7, characterized in that... The hazard characteristic data includes one or more of the following: spatial coordinates of the hazard location, geometric shape data, visual image data, and temperature distribution data.

9. The method for tunnel emergency rescue structure detection based on unmanned aerial vehicle (UAV) swarms as described in claim 1, characterized in that... The ground control station integrates a positioning system, data processing and analysis software, a safety assurance system, and a disaster relief decision support system. The positioning system is used to accurately locate the tunnel hazard. The data processing and analysis software is used to process the received data, construct a tunnel structure model, and analyze the hazard. The safety assurance system is used to monitor the UAV's attitude, battery level, and distance to obstacles in real time, and issue early warning signals to ensure equipment safety. The disaster relief decision support system is used to generate a disaster relief plan based on data analysis results and a built-in expert opinion database and hazard handling case database.

Images