Tunnel unmanned aerial vehicle cluster inspection link architecture

Abstract

The invention discloses a tunnel unmanned aerial vehicle cluster routing inspection link architecture, and relates to the technical field of tunnel intelligent routing inspection, the tunnel unmanned aerial vehicle cluster routing inspection link architecture comprises a water surface unmanned ship, a base station unmanned aerial vehicle group, a routing inspection unmanned aerial vehicle group and a mooring power supply network, and by adopting a clustered and intelligent unmanned aerial vehicle routing inspection mode, the tunnel routing inspection efficiency and accuracy are remarkably improved. Comprehensive and continuous detection of the long-distance tunnel is realized, and the problems of insufficient endurance, inaccurate positioning, unstable communication and the like in a traditional inspection mode are effectively solved. Meanwhile, in combination with a water surface unmanned ship and a laser positioning technology, leakage point scanning of a water area around the tunnel and construction of a space coordinate network are also realized, and the inspection precision and reliability are further improved. Besides, by dynamically adjusting the number of the unmanned aerial vehicles and communication links, flexible configuration can be carried out according to actual conditions of different tunnels, inspection blind areas are avoided, and safe operation of the tunnels is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent tunnel inspection, and in particular to a tunnel drone cluster inspection link architecture. Background Art

[0002] Current tunnel drone inspections use a single unit or small formation mode. They utilize infrared cameras and lidar to achieve 10km-level tunnel inspections, using wireless mesh networks for data transmission. Key technical limitations include a single-unit flight time of ≤45 minutes (approximately 10km range), inability to cover a 30km-long tunnel, interrupted detection continuity, GPS failure in underground environments, reliance on UWB positioning, a signal coverage radius of <500m, magnetic wave attenuation requiring the deployment of 50+ repeaters for a 30km tunnel, the need for manual recovery in the event of a failure, the inability to autonomously navigate around obstacles, and the lack of surface platform support for deepwater tunnels. Therefore, we propose a tunnel drone cluster inspection link architecture. Summary of the Invention

[0003] The purpose of the present invention is to solve the problems of single-machine endurance ≤ 45 minutes (about 10km range), inability to cover 30km long tunnels, GPS failure in underground environments, reliance on UWB positioning, signal coverage radius < 500m, manual recovery required in case of failure, inability to autonomously bypass obstacle areas, and no surface platform support for deep-water tunnels. The present invention provides a tunnel drone cluster inspection link architecture.

[0004] In order to achieve the above-mentioned purpose, the present invention specifically adopts the following technical solutions:

[0005] A tunnel drone cluster inspection link architecture includes the following functional modules:

[0006] Unmanned Surface Vessel (USV): Carries a 120kW generator set and docking station for deploying base station drones. It charges the drones via high-voltage cables and uses underwater sonar to scan for leaks.

[0007] Base station drone swarm: Fixedly deployed on the top of the tunnel at a spacing of 10km±1m, it receives a 48V±5% operating voltage via a tethered cable, emits a 532nm laser through a laser positioning unit to construct a spatial coordinate network, and dynamically extends the cable to reconfigure the communication link;

[0008] Inspection drone swarms (AUVs): Grouped at 1km intervals (30 per group), they use millimeter-wave radar to perform continuous rough scans of the structure to diagnose tunnel deformation, use multi-camera vision units to accurately locate cracks, and navigate around obstacles under the guidance of a laser net.

[0009] Tethered power supply network: transmits 5kV DC power to the base station, and after voltage reduction, it charges the docking drone through the magnetic suction interface.

[0010] Furthermore, the charging scheduling algorithm for the base station drone group on the surface unmanned ship is:

[0011]

[0012] Among them, T c is the single-machine charging time, T c =8min, N dock The number of idle charging docks, when the power SOC min When the charge level is less than 30%, priority will be given to charging in queue.

[0013] Furthermore, the fault recovery process when the communication link of the base station drone group is reconstructed is:

[0014] Step R1, monitoring the base station signal strength S<-90dBm;

[0015] Step R2: Calculate the extension distance between adjacent base stations.

[0016] Step R3: Extend the cables to reconstruct the ladder topology, and the delay compensation amount Δt = D e / 3×10 8 s.

[0017] Furthermore, the laser positioning unit workflow of the base station drone group is as follows:

[0018] Step S1: emit laser to form a 20m×20m grid

[0019] Step S2: CMOS sensor captures the light spot distortion image

[0020] Step S3: Execute positioning compensation calculation:

[0021] ΔX=k·∫∫ A (dx d -dx0)dA

[0022] Among them, k is the tunnel curvature correction coefficient, ranging from 0.7 to 1.3, dx d is the pixel distortion detected in real time, dx0 is the standard spatial coordinate reference value without distortion, dA is the 20m×20m laser grid area, and double integration is used to eliminate cumulative errors and output three-dimensional offsets (ΔX, ΔY, ΔZ) to achieve ±5cm positioning accuracy, correct positioning deviations in a GPS-free environment, and eliminate positioning errors caused by tunnel air turbulence and equipment vibration.

[0023] Furthermore, the triggering conditions for the tunnel structure deformation alarm of the inspection drone group are:

[0024] ||M t -M ref || F>2.0mm

[0025] Among them, M t is the current millimeter-wave radar real-time reflection intensity matrix, M Tef The benchmark reflection intensity matrix in the historical safety benchmark database performs actions. When the structural deformation is greater than 2mm, the precision inspection mode of the multi-eye vision unit is automatically triggered.

[0026] Furthermore, the multi-vision unit uses three AUVs to form an equilateral triangle array with a side length of 2m±0.05m, synchronously captures 200 high-definition images of the crack area, and reconstructs a 3D point cloud model through triangulation:

[0027]

[0028] Where f is the camera focal length, B is the baseline distance, and d is the disparity pixel.

[0029] Furthermore, the path planning logic of the inspection drone group to bypass the obstacle area under the guidance of the laser network is as follows: the obstacle distance D detected by the millimeter wave radar obs , RRT algorithm generates a new path point set {P k}, formation reorganization parameters:

[0030]

[0031] Furthermore, the tethered power supply network uses a high-voltage transformer to reduce 5kV to a safe voltage, dynamically compensates for line voltage drop, and ensures that the output voltage is stable at 48V±5%. The line loss compensation formula is:

[0032]

[0033] Among them, V in is the input voltage (5kV), is the turn ratio of the step-down transformer (100:1), I is the load current (measured value 8~15A), L is the cable length (per 10km segment), R line ≤0.2Ω / km, which is the cable resistance, compensating for line loss to stabilize the output voltage.

[0034] Furthermore, the detection data of the surface unmanned ship and the inspection drone group are unified into the water, land and air data coordinate system to construct a full-section digital twin model. The coordinate unified transformation model is:

[0035]

[0036] Among them, R 3×3 is a 3×3 rotation matrix obtained by laser-sonar calibration, T is the translation vector (error ≤ 0.5 mm), (X s ,Y s ,Zs ) is the sonar underwater coordinate, (X w ,Y w ,Z w ) is the laser positioning coordinate system.

[0037] Furthermore, the configuration formula for the number of drone devices in the base station drone group and the inspection drone group is:

[0038]

[0039] Among them, N base is the number of base station drones, N auv is the number of inspection drones, L t : Total length of tunnel (unit km), k T =1.2 is the redundancy coefficient, which accurately configures drone resources to avoid inspection blind spots.

[0040] The beneficial effects of the present invention are as follows:

[0041] 1. The present invention significantly improves the efficiency and accuracy of tunnel inspections by adopting a clustered and intelligent drone inspection method. It achieves comprehensive and continuous inspections of long-distance tunnels, effectively solving the problems of insufficient endurance, inaccurate positioning, unstable communications, etc. in traditional inspection methods. At the same time, combined with surface unmanned boats and laser positioning technology, the present invention also realizes the scanning of leakage points in the waters around the tunnel and the construction of a spatial coordinate network, further improving the accuracy and reliability of inspections. In addition, by dynamically adjusting the number of drones and communication links, the present invention can be flexibly configured according to the actual conditions of different tunnels, avoiding inspection blind spots and ensuring the safe operation of the tunnel.

[0042] 2. The charging scheduling algorithm in this invention ensures that the drone with the lowest battery level receives priority charging, thus avoiding inspection mission interruptions due to battery depletion. In practice, the surface unmanned vessel intelligently schedules idle charging docks based on the current battery level of the drone swarm, making the entire inspection process more efficient and reliable. Furthermore, the algorithm is adaptive, dynamically adjusting the charging strategy based on the real-time battery level of the drone swarm to ensure the smooth progress of the inspection mission.

[0043] 3. The multi-eye vision unit of the present invention uses 3 AUVs to form an equilateral triangle array with a side length of about 2 meters (error range ±0.05 meters). These AUVs synchronously capture high-definition images of the crack area, taking a total of 200 images. Using the principle of triangulation and combining these image information, a 3D point cloud model of the crack area can be reconstructed. During the reconstruction process, by calculating the camera focal length f, baseline distance B, and parallax pixel d, a 3D point cloud model can be constructed with an accuracy of ±50 microns. Based on the reconstructed 3D point cloud model, the morphology, size and other information of the crack are analyzed in detail to provide data support for subsequent repair work. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below.

[0045] The present invention provides a tunnel drone cluster inspection link architecture, including the following functional modules:

[0046] Unmanned Surface Vessel (USV): Carries a 120kW generator set and docking station for deploying base station drones. It charges the drones via high-voltage cables and uses underwater sonar to scan for leaks.

[0047] Base station drone swarm: Fixedly deployed on the top of the tunnel at a spacing of 10km±1m, it receives a 48V±5% operating voltage via a tethered cable, emits a 532nm laser through a laser positioning unit to construct a spatial coordinate network, and dynamically extends the cable to reconfigure the communication link;

[0048] Inspection drone swarms (AUVs): Grouped at 1km intervals (30 per group), they use millimeter-wave radar to perform continuous rough scans of the structure to diagnose tunnel deformation, use multi-camera vision units to accurately locate cracks, and navigate around obstacles under the guidance of a laser net.

[0049] Tethered power supply network: transmits 5kV DC power to the base station, and after voltage reduction, it charges the docking drone through the magnetic suction interface.

[0050] An unmanned surface vessel (USV) is deployed in the waters near the tunnel entrance, carrying a 120kW generator set and docking station. When the base station drone is deployed, the USV transfers power to the drone via a high-voltage cable for charging, ensuring sufficient energy for inspections. The USV is also equipped with an underwater sonar scanning system, which scans the waters surrounding the tunnel for leaks, promptly identifying and reporting potential leaks.

[0051] The base station drone swarm is deployed at a preset 10km ±1m spacing on the tunnel roof. These drones receive a 48V ±5% operating voltage from the tethered power supply network via tethered cables, ensuring stable operation. They are also equipped with laser positioning units that emit 532nm laser light to construct a spatial coordinate grid, providing precise navigation information for the inspection drone swarm. When communication links need to be reestablished, the base station drone swarm can dynamically extend the cables to ensure a stable communication connection for the inspection drone swarm.

[0052] The inspection drone swarm, grouped 30 drones at 1km intervals, conducts tunnel inspections. Traveling in formation at 5m / s, they use millimeter-wave radar for continuous, coarse structural scanning, enabling rapid diagnosis of tunnel deformation. If an anomaly is detected, the drones use multi-camera vision units for precise positioning, capturing details such as cracks. Guided by a laser network, they navigate around obstructions, ensuring smooth inspections.

[0053] The tethered power supply network transmits 5kV DC power to the base station, where it is stepped down and then charged via a magnetic interface. This ensures the continuous and stable operation of the entire tunnel drone cluster inspection link architecture, providing strong support for tunnel inspection work.

[0054] By adopting a clustered and intelligent drone inspection method, the efficiency and accuracy of tunnel inspections have been significantly improved. Comprehensive and continuous inspections of long-distance tunnels have been achieved, effectively solving the problems of insufficient endurance, inaccurate positioning, unstable communications, and other issues existing in traditional inspection methods. At the same time, combined with surface unmanned boats and laser positioning technology, the present invention also realizes the scanning of leakage points in the waters around the tunnel and the construction of a spatial coordinate network, further improving the accuracy and reliability of inspections. In addition, by dynamically adjusting the number of drones and communication links, the present invention can be flexibly configured according to the actual conditions of different tunnels, avoiding inspection blind spots and ensuring the safe operation of the tunnel.

[0055] In this embodiment, preferably, the charging scheduling algorithm for the base station drone group on the surface unmanned ship is:

[0056]

[0057] Among them, T c is the single-machine charging time, T c =8min, N dock is the number of idle charging docks, SOC min The remaining power percentage of the drone with the lowest power in the drone group. min When the charge level is less than 30%, priority will be given to charging in queue.

[0058] Calculate the time slot T for charging scheduling of the base station drone group on the surface unmanned ship slot According to the current minimum power state SOC of the drone group min and available charging resources N dock , dynamically calculate a time slot T slot , which is used to schedule the charging order of the drone swarm. In this way, the drone with the lowest battery level can be charged first, thereby improving the continuous operation capability of the entire drone swarm.

[0059] The charging scheduling algorithm ensures that drones with the lowest battery levels receive priority charging, thus preventing inspection missions from being interrupted due to depleted batteries. In practice, the surface drone intelligently schedules unused charging docks based on the current battery level of the drone swarm, making the entire inspection process more efficient and reliable. Furthermore, the algorithm is adaptive, dynamically adjusting charging strategies based on the real-time battery level of the drone swarm to ensure smooth inspection missions.

[0060] In this embodiment, preferably, the fault recovery process when the communication link of the base station drone group is reconstructed is:

[0061] Step R1: Monitor the base station signal strength S < -90dBm. Use professional signal monitoring equipment or software to monitor the signal strength of the base station in real time. If the signal strength is lower than -90dBm, the subsequent fault recovery process is triggered. Signal strength is an important indicator for measuring the coverage and quality of base station signals. When the signal strength is too low, it may lead to degraded communication quality or even communication interruption. Therefore, set the signal strength threshold (-90dBm) as the basis for determining whether the base station is faulty.

[0062] Step R2: Calculate the extension distance between adjacent base stations. Get the power difference P between adjacent base stations f -P adj , where P f is the current base station transmit power, P adj is the transmission power of the adjacent base station. Then, substitute the power difference into the formula to calculate the extended distance D e , calculated based on the relationship between power and distance. In free space, signal power is inversely proportional to the square of the propagation distance. Therefore, by comparing the transmit power of adjacent base stations, the relative distance between them can be estimated. The 5% error range here takes into account the impact of various factors in the actual environment, such as obstacles and air attenuation, on signal propagation;

[0063] Step R3: Extend the cables to reconstruct the ladder topology, and the delay compensation amount Δt = D e / 3×10 8 s, after calculating the extended distance D of the adjacent base statione Finally, use cables to connect adjacent base stations to form a trapezoidal topology. At the same time, the delay compensation amount Δt is calculated according to the formula, and corresponding settings are made in the system to ensure the synchronization and stability of communication. The trapezoidal topology has good stability and scalability, and is suitable for the reconstruction of communication links of base station drone swarms. The calculation of the delay compensation amount Δt takes into account the time delay of the signal propagating in the cable. Since the speed of light in a vacuum is about 3×10^8m / s, the delay time can be calculated by dividing the extended distance De by the speed of light. The division by 3 here is to simplify the calculation and take into account that the transmission speed of the cable in the actual environment may be slightly lower than the speed of light. By setting the delay compensation amount Δt, the synchronization and stability of communication between adjacent base stations can be ensured.

[0064] In this embodiment, preferably, the laser positioning unit of the base station drone group has the following working process:

[0065] Step S1: Laser emission forms a 20m x 20m grid. The laser emitter emits laser light to form a 20m x 20m grid area. This step provides a basic reference frame for positioning.

[0066] Step S2: The CMOS sensor captures a light spot distortion image. The CMOS sensor captures a light spot distortion image formed by the laser irradiating the target area. The light spot image may be distorted due to environmental factors such as tunnel curvature, air turbulence, or equipment vibration.

[0067] Step S3: Execute positioning compensation calculation:

[0068] ΔX=k·∫∫ A (dx d -dx0)dA

[0069] Among them, k is the tunnel curvature correction coefficient, ranging from 0.7 to 1.3, dx d is the pixel distortion detected in real time, dx0 is the standard spatial coordinate reference value without distortion, dA is the 20m×20m laser grid area, and double integration is used to eliminate cumulative errors and output three-dimensional offsets (ΔX, ΔY, ΔZ) to achieve ±5cm positioning accuracy, correct positioning deviations in a GPS-free environment, and eliminate positioning errors caused by tunnel air turbulence and equipment vibration.

[0070] Based on the calculations, the system corrects the swarm's position, ensuring it can achieve a positioning accuracy of ±5 cm in GPS-denied environments. This step helps eliminate positioning errors caused by factors like tunnel air turbulence and equipment vibration, improving positioning accuracy and stability.

[0071] In this embodiment, preferably, the triggering condition for the inspection drone group to diagnose tunnel structure deformation alarm is:

[0072] ||M t -M ref || F >2.0mm

[0073] Among them, M t is the current millimeter-wave radar real-time reflection intensity matrix, M Tef The benchmark reflection intensity matrix in the historical safety benchmark database performs actions. When the structural deformation is greater than 2mm, the precision inspection mode of the multi-eye vision unit is automatically triggered.

[0074] Start the inspection drone swarm and let the drones fly into the tunnel. The millimeter-wave radar in the drone swarm begins to monitor the tunnel structure surface in real time and obtain the current reflection intensity matrix M t The drone swarm will obtain the reflection intensity matrix M in real time t Compared with the benchmark reflection intensity matrix M in the historical security benchmark database Tef The system automatically calculates the difference between the two and determines whether deformation exceeds a threshold. If deformation exceeds 2.0 mm, the system automatically triggers the multi-camera vision unit's fine-inspection mode. The multi-camera vision unit then begins a more detailed inspection and analysis of the tunnel structure to determine the specific nature and severity of the deformation. Based on the results of the fine-inspection mode, appropriate maintenance or repair measures can be implemented.

[0075] In this embodiment, preferably, the multi-eye vision unit uses three AUVs to form an equilateral triangle array with a side length of 2m±0.05m, synchronously captures 200 high-definition images of the crack area, and reconstructs a 3D point cloud model through triangulation:

[0076]

[0077] Where f is the camera focal length, B is the baseline distance, and d is the disparity pixel.

[0078] The multi-camera vision unit uses three AUVs to form an equilateral triangle array with a side length of approximately 2 meters (error range of ±0.05 meters). These AUVs simultaneously capture high-definition images of the crack area, taking a total of 200 images. Using the principle of triangulation and combining this image information, a 3D point cloud model of the crack area can be reconstructed. During the reconstruction process, by calculating the camera focal length f, baseline distance B, and parallax pixel d, a 3D point cloud model can be constructed with an accuracy of ±50 microns. Based on the reconstructed 3D point cloud model, detailed analysis of the crack's morphology, size, and other information is carried out to provide data support for subsequent repair work.

[0079] In this embodiment, preferably, the path planning logic of the inspection drone group to bypass the obstacle area under the guidance of the laser network is: the millimeter wave radar detects the obstacle distance Dobs , RRT algorithm generates a new path point set {P k}, formation reorganization parameters:

[0080]

[0081] A swarm of inspection drones, guided by a laser network, plans a path around obstacles. Millimeter-wave radar is used to detect the distance Dobs to the obstacle ahead. This step is crucial because it provides the foundational data for subsequent path planning. The RRT algorithm is introduced to generate a new set of pathpoints Pk. The RRT algorithm is a fast random search tree algorithm for path planning that efficiently searches and finds feasible paths in high-dimensional space. In this embodiment, the RRT algorithm uses the obstacle distance Dobs detected by the millimeter-wave radar as input. Through random sampling and tree expansion, it gradually generates a series of new pathpoints, which ultimately form the new path for the drone swarm to circumvent the obstacle. To ensure that the drone swarm maintains a proper formation when circumventing obstacles, a formation reorganization parameter θ is set. This parameter is calculated using the arctan function, where Dobs is the obstacle distance and v is the drone's speed. The arctan function ensures that θ is between -90 and 90 degrees. To limit the swarm's tilt angle, θ is further constrained to no more than 30 degrees.

[0082] This allows the drone swarm to quickly adjust its formation, even when faced with complex obstacle layouts, ensuring safe and efficient navigation around obstructions. Furthermore, a fault detection and recovery mechanism has been designed to enhance system robustness. If a drone fails and is unable to continue its mission, the system quickly identifies and triggers a backup drone to take over, while simultaneously replanning the route to ensure the entire inspection mission remains unaffected. This highly intelligent link architecture not only improves the efficiency and accuracy of tunnel inspections but also lays a solid foundation for the future widespread application of drone swarms in complex environments.

[0083] In this embodiment, preferably, the tethered power supply network uses a high-voltage transformer to reduce 5 kV to a safe voltage, dynamically compensates for line voltage drop, and ensures that the output voltage is stable at 48 V ± 5%. The line loss compensation formula is:

[0084]

[0085] Among them, V in is the input voltage (5kV), is the turn ratio of the step-down transformer (100:1), I is the load current (measured value 8~15A), L is the cable length (per 10km segment), R line ≤0.2Ω / km, which is the cable resistance, compensating for line loss to stabilize the output voltage.

[0086] By reducing the voltage of the high-voltage transformer and dynamically compensating for line voltage drops, the output voltage stability is effectively ensured and the power supply quality is improved. Using a precise line loss compensation formula, compensation is performed based on real-time load current and cable parameters, reducing voltage losses caused by line resistance and further improving voltage stability. The overall solution is safe and reliable and can be applied to various occasions requiring stable power supply, especially in long-distance transmission or when the load varies greatly, where its advantages are more prominent.

[0087] In this embodiment, preferably, the detection data of the surface unmanned ship and the inspection drone group are unified in the water, land and air data coordinate system to construct a full-section digital twin model, and the coordinate unified transformation model is:

[0088]

[0089] Among them, R 3×3 is a 3×3 rotation matrix obtained by laser-sonar calibration, T is the translation vector (error ≤ 0.5 mm), (X s ,Y s ,Z s ) is the sonar underwater coordinate, (X w ,Y w ,Z w ) is the laser positioning coordinate system.

[0090] This unified coordinate transformation model ensures the precise integration of data collected by surface unmanned vessels and patrol drones in different waters and airspaces, forming a complete integrated data system for land, sea, and air. Through the precise calculation of the 3×3 rotation matrix R3×3 and the translation vector T, seamless integration between the sonar underwater coordinate system and the laser positioning coordinate system is achieved, with an error control within 0.5mm, significantly improving the accuracy and reliability of the data.

[0091] The construction of a full-section digital twin model, based on this precisely converted data, can truly reflect the actual conditions inside and outside the tunnel, including its structure, equipment, environment, and other aspects. This not only provides strong data support for the tunnel's daily operations and maintenance, but also provides a scientific basis for rapid response and decision-making in emergency situations.

[0092] In this embodiment, preferably, the configuration formula for the number of drone devices in the base station drone group and the inspection drone group is:

[0093]

[0094] Among them, N base is the number of base station drones, N auv is the number of inspection drones, L t : Total length of tunnel (unit km), kT =1.2 is the redundancy coefficient, which accurately configures drone resources to avoid inspection blind spots.

[0095] The equipment configuration formula comprehensively considers tunnel length and inspection redundancy requirements, ensuring comprehensive inspection coverage in all circumstances. By adjusting the number of base station drone swarms and inspection drone swarms, it can flexibly adapt to tunnels of varying lengths, ensuring inspection efficiency and accuracy while avoiding resource waste. Furthermore, the introduction of redundancy factors further enhances inspection reliability and safety, ensuring continuity and integrity even in the event of partial drone failure or anomalies. This precise resource allocation strategy is crucial for improving the intelligence level and operational efficiency of tunnel inspections.

[0096] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A tunnel drone cluster inspection link architecture, characterized by: Includes the following functional modules: Unmanned Surface Vessel (USV): Carries a 120kW generator set and docking station for deploying base station drones. It charges the drones via high-voltage cables and uses underwater sonar to scan for leaks. Base station drone swarm: Fixedly deployed on the top of the tunnel at a spacing of 10km±1m, it receives a 48V±5% operating voltage via a tethered cable, emits a 532nm laser through a laser positioning unit to construct a spatial coordinate network, and dynamically extends the cable to reconfigure the communication link; Inspection drone swarms (AUVs): Grouped at 1km intervals (30 drones / group), they use millimeter-wave radar to perform continuous structural rough scans to diagnose tunnel structural deformation, use multi-camera vision units to accurately locate cracks, and navigate around obstructions under the guidance of a laser net. Tethered power supply network: transmits 5kV DC power to the base station, and after voltage reduction, it charges the docking drone through the magnetic suction interface.

2. The tunnel drone cluster inspection link architecture according to claim 1 is characterized by: The charging scheduling algorithm for the base station drone group on the surface unmanned ship is: Among them, T c is the single-machine charging time, T c =8min, N dock The number of idle charging docks, when the power SOC min When the charge level is less than 30%, priority will be given to charging in queue.

3. The tunnel drone cluster inspection link architecture according to claim 1 is characterized by: The fault recovery process when the communication link of the base station drone group is reconstructed is as follows: Step R1, monitoring the base station signal strength S<-90dBm; Step R2: Calculate the extension distance between adjacent base stations. Step R3: Extend the cables to reconstruct the ladder topology, and the delay compensation amount Δt = D e / 3×10 8 s.

4. The tunnel drone cluster inspection link architecture according to claim 1 is characterized by: The laser positioning unit workflow of the base station drone group is as follows: Step S1: emit laser to form a 20m×20m grid Step S2: CMOS sensor captures the light spot distortion image Step S3: Execute positioning compensation calculation: ΔX=k·∫∫ A (dx d -dx0)dA Among them, k is the tunnel curvature correction coefficient, ranging from 0.7 to 1.3, dx d is the pixel distortion detected in real time, dx0 is the standard spatial coordinate reference value without distortion, dA is the 20m×20m laser grid area, and double integration is used to eliminate cumulative errors and output three-dimensional offsets (ΔX, ΔY, ΔZ) to achieve ±5cm positioning accuracy, correct positioning deviations in a GPS-free environment, and eliminate positioning errors caused by tunnel air turbulence and equipment vibration.

5. The tunnel drone cluster inspection link architecture according to claim 1 is characterized by: The triggering conditions for the tunnel structure deformation alarm of the inspection drone group are: ||M t -M ref || F >2.0mm Among them, M t is the current millimeter-wave radar real-time reflection intensity matrix, M Tef The benchmark reflection intensity matrix in the historical safety benchmark database performs actions. When the structural deformation is greater than 2mm, the precision inspection mode of the multi-eye vision unit is automatically triggered.

6. The tunnel drone cluster inspection link architecture according to claim 1 is characterized by: The multi-eye vision unit uses three AUVs to form an equilateral triangle array with a side length of 2m±0.05m, synchronously captures 200 high-definition images of the crack area, and reconstructs a 3D point cloud model through triangulation: (Accuracy ±50μm) Where f is the camera focal length, B is the baseline distance, and d is the disparity pixel.

7. The tunnel drone cluster inspection link architecture according to claim 1 is characterized by: The path planning logic of the inspection drone group to bypass the obstacle area under the guidance of the laser network is as follows: the millimeter wave radar detects the obstacle distance D obs , RRT algorithm generates a new path point set {P k }, formation reorganization parameters: (Limit θ≤30°).

8. The tunnel drone cluster inspection link architecture according to claim 1 is characterized by: The tethered power supply network uses a high-voltage transformer to reduce 5kV to a safe voltage, dynamically compensates for line voltage drop, and ensures that the output voltage is stable at 48V±5%. The line loss compensation formula is: Among them, V in is the input voltage (5kV), is the turn ratio of the step-down transformer (100:1), I is the load current (measured value 8~15A), L is the cable length (per 10km segment), R line ≤0.2Ω / km, which is the cable resistance, compensating for line loss to stabilize the output voltage.

9. The tunnel drone cluster inspection link architecture according to claim 1 is characterized by: The detection data of the surface unmanned ship and the inspection drone group are unified in the water, land and air data coordinate system to build a full-section digital twin model. The coordinate unified conversion model is: Among them, R 3×3 is a 3×3 rotation matrix obtained by laser-sonar calibration, T is the translation vector (error ≤ 0.5 mm), (X s ,Y s ,Z s ) is the sonar underwater coordinate, (X w ,Y w ,Z w ) is the laser positioning coordinate system.

10. The tunnel drone cluster inspection link architecture according to claim 1, characterized in that: The configuration formula for the number of drone devices in the base station drone group and the inspection drone group is: Among them, N base is the number of base station drones, N auv is the number of inspection drones, L t : Total length of tunnel (unit km), k T =1.2 is the redundancy coefficient, which accurately configures drone resources to avoid inspection blind spots.

Images