An indoor inspection signal enhancement system based on multi-unmanned aerial vehicle dynamic relay network

Abstract

The application discloses an indoor inspection signal enhancement system based on a multi-unmanned aerial vehicle dynamic relay network, comprising: an inspection path planning module, which is used for determining a planned inspection path of an unmanned aerial vehicle and calculating an equivalent communication path length; a relay demand prediction module, which is used for determining the number of relay unmanned aerial vehicles required by an inspection task; a multi-hop relay link construction module, which is used for generating relay deployment control instruction parameters according to the number of relay unmanned aerial vehicles and building structure data; a relay position adjustment module, which is used for periodically collecting link quality indicators and adjusting the hovering position of the relay unmanned aerial vehicle; a data forwarding path selection module, which is used for selecting a neighbor node for next-hop forwarding of the relay unmanned aerial vehicle; and a relay stable regulation and control module, which is used for route reconstruction, standby relay intervention and relay deployment correction. The application effectively improves the reliability, network stability and system robustness of indoor inspection communication, reduces the demand for manual intervention, and enhances the engineering practicability of the multi-unmanned aerial vehicle indoor inspection system.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) communication and intelligent inspection technology, and in particular to an indoor inspection signal enhancement system based on a multi-UAV dynamic relay network. Background Technology

[0002] With the continuous development of drone technology and intelligent inspection technology, drones have been widely used in inspection operations in industrial plants, underground spaces, large public buildings, and complex indoor environments. Compared with manual inspection, drone inspection has advantages such as high mobility, wide coverage, and high operational efficiency, effectively reducing labor costs and improving inspection accuracy. However, existing indoor drone inspection systems mostly adopt single-unit direct connection or static relay communication methods. Inspection path planning usually focuses on covering the target area or the shortest flight distance, failing to fully consider the continuity of the communication link and the need for multi-hop forwarding. In practical applications, when drones travel along preset paths deep into building interiors or across multi-story areas, communication quality can drop sharply or even be interrupted due to reasons such as exceeding communication distance limits, wall obstruction, or channel attenuation. This affects the real-time transmission of inspection data and the security of task execution. To improve communication coverage, some solutions introduce multiple relay drones working collaboratively. However, in existing technologies, the number of relay drones is often fixed, or deployed based solely on simple distance rules, lacking comprehensive consideration of inspection path length, building structure differences, and communication load variations. This easily leads to insufficient or redundant relay resources. Furthermore, the spatial deployment of relay nodes is often pre-set or manually intervened, making it difficult to adapt to dynamically changing communication conditions in complex indoor environments, resulting in large fluctuations in link quality and poor network stability. On the other hand, in multi-hop relay networks, link quality and node load change continuously over time. Existing technologies have limited ability to perceive and handle issues such as abnormal link quality and unbalanced node loads, typically lacking effective dynamic evaluation mechanisms and adaptive adjustment methods. When a relay node experiences a decline in communication quality or excessive load, it can easily trigger cascading communication problems, thereby affecting the routing reachability and task continuity of the entire inspection network, increasing manual maintenance costs and system operational risks. Therefore, there is an urgent need for an indoor inspection signal enhancement method that can combine indoor environmental characteristics, inspection task requirements, and communication status changes to collaboratively optimize inspection path planning, relay deployment, and multi-hop communication, so as to improve the communication reliability, deployment flexibility, and overall operational stability of multi-UAV inspection systems in complex indoor environments. Summary of the Invention

[0003] To address the problems existing in the prior art, this invention provides an indoor inspection signal enhancement system based on a multi-UAV dynamic relay network, mainly comprising:

[0004] The inspection path planning module is used to determine the planned inspection path of the UAV based on indoor environmental data and inspection task configuration files, and to calculate the equivalent communication path length of the planned inspection path.

[0005] The relay demand prediction module is used to determine the minimum number of working relay drones and the number of backup relay drones required for the inspection task based on the upper limit of single-hop communication distance and the equivalent communication path length of the inspection path.

[0006] The multi-hop relay link construction module is used to obtain the location of relay nodes and the corresponding three-dimensional relay deployment location data based on building structure data and the upper limit of single-hop communication distance, and generate relay deployment control command parameters.

[0007] The relay position adjustment module is used to periodically collect link quality indicators, monitor the link quality scores between the relay drone and upstream and downstream relay nodes, and adjust the hovering position of the relay drone.

[0008] The data forwarding path selection module is used to eliminate neighboring nodes with abnormal quality based on the link quality score between each relay drone and its neighboring nodes, and select the neighboring node of the next hop forwarding object to forward the inspection task data to the control station.

[0009] The relay stability control module is used to periodically collect link quality scores and load information between each relay drone and its neighboring nodes to determine the routing reachability status and relay deployment stability deviation of the multi-hop relay network, and to trigger route reconstruction, backup relay intervention, and relay deployment correction processing.

[0010] Furthermore, the inspection path planning module is used to determine the planned inspection path of the UAV based on indoor environmental data and inspection task configuration files, and to calculate the equivalent communication path length of the planned inspection path, including:

[0011] The UAV inspection system acquires indoor environmental data, an indoor 3D spatial map, and an inspection task configuration file, and determines the planned inspection path for the UAV. The indoor environmental data includes building structure information, floor distribution information, wall position relationships, floor connection relationships, control station location, and the UAV's task area information. The inspection task configuration file includes the inspection start position, inspection end position, and inspection area constraints. Based on the planned inspection path, the system discretizes the path at a preset step size to obtain several path nodes. The system acquires the path node sequence and its position information within the building structure, and marks the structural unit types traversed by each inspection path segment. Structural unit types include walls, floors, metal shielding surfaces, and cavity structures. The equivalent communication path length of the planned inspection path is calculated based on the additional equivalent length of the structural type traversed by each inspection path segment. This additional equivalent length is pre-calibrated based on empirical experiments or a simple path loss model.

[0012] Furthermore, the relay demand prediction module is used to determine the minimum number of working relay drones and the number of backup relay drones required for the inspection task based on the upper limit of single-hop communication distance and the equivalent communication path length of the inspection path, including:

[0013] By using the inspection communication detection database, historical data on the transmit power, operating frequency band, and receive sensitivity of the wireless communication module are obtained, and the upper limit of the single-hop communication distance is marked. A recurrent neural network is used to train the model and construct a prediction model for the upper limit of the single-hop communication distance. Based on the transmit power, operating frequency band, and receive sensitivity of the currently used wireless communication module, the upper limit of the single-hop communication distance is predicted using the prediction model. Based on the upper limit of the single-hop communication distance of the current inspection task and the equivalent communication path length of the planned inspection path, the minimum number of relay drones required for the current inspection task is calculated. Based on the importance of the current inspection task and the network reliability level, the number of backup relay drones matching the importance of the task and the network reliability level is extracted from the preset redundancy ratio configuration table.

[0014] Furthermore, the multi-hop relay link construction module is used to obtain the relay node location and corresponding three-dimensional relay deployment location data based on building structure data and the upper limit of single-hop communication distance, and generate relay deployment control command parameters, including:

[0015] By extracting entrance space nodes, staircase space nodes, passageway connection nodes, and inspection path-related space nodes from building structure data, a set of candidate relay node locations is obtained. Based on the upper limit of single-hop communication distance, the candidate relay node locations are sequentially filtered along the inspection path to determine the relay node location. If the distance between adjacent candidate relay node locations exceeds the upper limit of single-hop communication distance, at least one supplementary relay location is generated on the corresponding inspection path segment according to an equidistant distribution rule, based on the path length between adjacent candidate relay node locations. Finally, based on floor height description information and relay hovering height description information... The parameters are described above to determine the three-dimensional relay deployment location data corresponding to the relay node location and the supplementary relay location, and to generate relay deployment control command parameters, including the three-dimensional position coordinates of the UAV deployment point, hovering altitude, identification, activation status, backup identification, and allocation order; the relay deployment control command parameters are sent to each relay UAV through the control station to perform take-off control, path tracking control, and hovering control on the relay UAV, and each relay UAV is sequentially flown to the corresponding target location, enters a stable hovering state, establishes a multi-hop relay link, and forms a communication channel from the control station, relay UAVs to inspection UAVs.

[0016] Furthermore, the relay position adjustment module is used to monitor the link quality scores between the relay UAV and upstream and downstream relay nodes by periodically collecting link quality indicators, and to adjust the hovering position of the relay UAV, including:

[0017] Based on the local temperature difference changes, local air disturbance velocity, material type, and span characteristics at the relay node and supplementary relay locations, it is determined whether the hovering position of the relay UAV is within the structural disturbance resonance sensitive zone, and the hovering point of the relay UAV is adjusted accordingly. Span characteristics include span size, span construction method, and span connection structure. Span construction methods include single-span and multi-span continuous beams, and span connection structures include fixed connections, semi-rigid connections, and hinged connections. Relay UAVs using multi-hop relay links send test data packets between their upstream and downstream relay nodes. During data forwarding by the relay UAVs, link quality indicators between each relay UAV and its upstream and downstream relay nodes are periodically collected according to a preset time window. Link quality indicators include received signal strength. The system calculates the link quality index, link quality index, signal-to-noise ratio, and packet loss rate. A weighted average of these indicators is used to obtain the link quality score for each relay UAV and its upstream and downstream relay nodes. If the link quality score is lower than a preset threshold for more than N consecutive times, the current hovering position of the relay UAV is deemed not to meet communication quality requirements, and a position adjustment operation is performed according to preset movement rules based on the UAV's current position. The system obtains the link quality score after each movement of the relay UAV. If the link quality score recovers to above the preset threshold within a limited number of movements, the UAV stops moving and maintains the new hovering position. If the threshold cannot be reached within a preset maximum number of movements, a backup relay UAV is controlled by the control station to take off and fill the gap.

[0018] This also includes determining whether the hovering position of the relay UAV is within the structural disturbance resonance sensitive zone based on the local temperature difference change, local air disturbance velocity, material type, and span characteristics of the relay node location and the supplementary relay location, and adjusting the hovering point of the relay UAV accordingly. Specifically, this includes:

[0019] Based on the relay node location and supplementary relay location, the material type and span characteristics of the surrounding building components are obtained and stored in the inspection communication detection database. Historical data on the material type and span characteristics of the building components are obtained from the inspection communication detection database, and the equivalent disturbance energy level of the building components is labeled. A recurrent neural network is used for model training to construct a prediction model for the equivalent disturbance energy level of the building components, predicting the equivalent disturbance energy level of the building components for the current inspection task. Temperature and airflow sensors from the relay UAV are used to obtain the local temperature difference change and local air disturbance velocity near the UAV deployment point, respectively. Combined with the equivalent disturbance energy level of the building components, the structural disturbance coupling index formula is used. Calculate the structural disturbance coupling index of the current relay UAV deployment site. ,in This represents the local temperature change of the building structure near the deployment point per unit time. The local air disturbance velocity detected at the deployment point location. The equivalent disturbance energy level of the building structure components in this area; if the structural disturbance coupling index of the current relay UAV deployment point is greater than the preset index threshold, it is determined that the location is in the structural disturbance resonance sensitive area and is not suitable as a relay hovering point. Then, an alternative hovering point that meets the condition of having a structural disturbance coupling index lower than the threshold is searched in the neighborhood of the deployment point, and the corresponding relay deployment control command parameters are updated.

[0020] Furthermore, the data forwarding path selection module is used to eliminate neighboring nodes with abnormal link quality based on the link quality score between each relay UAV and its neighboring nodes, and select the neighboring node of the next-hop forwarding object to forward the inspection task data to the control station, including:

[0021] The inspection drones, controlled by the control station, take off from the building entrance and fly along the planned inspection path to perform inspection tasks, continuously collecting inspection data. This data, including image and environmental data, is then transmitted to the nearest relay drone. The relay drones initiate an ad hoc network protocol to obtain a list of communicable neighboring nodes, sending the collected inspection data to these nodes and collecting link quality indicators during communication. A weighted average of these link quality indicators is used to obtain a link quality score between each relay drone and its neighboring nodes. If the link quality score is lower than a preset threshold, the neighboring node is marked as an abnormal node and removed from the next-hop forwarding process. Based on the remaining neighboring nodes' identifiers, relative positions, link quality scores, and current forwarding load indicators, a path cost evaluation formula is used. Calculate the cost of the current relay drone i choosing neighbor node j as the next hop. The node with the lowest cost is selected as the next-hop forwarding target, and the inspection task data is forwarded to the control station. The forwarding load metric is the ratio of the amount of data currently cached by a node to its maximum cache capacity. Let be the estimated number of hops from node j to control station 1. To score the link quality, Let α be the current forwarding load metric for node j, and let β and γ be the weighting coefficients obtained by fitting historical data.

[0022] Furthermore, the relay stability control module is used to determine the route reachability status and relay deployment stability deviation of the multi-hop relay network by periodically collecting link quality scores and load information between each relay drone and its neighboring nodes, and to trigger route reconstruction, backup relay intervention, and relay deployment correction processing, including:

[0023] According to a preset time window, during the data forwarding process of relay drones, the link quality scores between each relay drone and its neighboring nodes are periodically collected. If a relay drone detects that the link quality scores between itself and all its neighboring nodes are below the score threshold or that more than M consecutive data forwarding failures occur, it broadcasts a route reconstruction request to surrounding nodes. Based on the reachability and link quality indicators returned to the control station by the neighboring nodes that receive the route reconstruction request, an available path is reconstructed. If the path cannot be reconstructed within the preset time threshold, an anomaly is reported to the control station, and a backup relay drone is triggered to join the network to reconstruct an available path. The load information of each relay node per unit time is statistically analyzed through communication logs, and the relay deployment stability deviation evaluation formula is used. Calculate the relay deployment stability deviation value Where K is the total number of relay drones currently in operation. Let be the forwarding load metric of the i-th relay node per unit time. The value is a natural logarithmic function used to enhance sensitivity to distribution sparsity. If the relay deployment stability deviation is less than the preset stability threshold, the relay deployment is considered relatively balanced and no adjustment is needed. If the relay deployment stability deviation is greater than the preset stability threshold, the relay deployment is considered unstable, and a relay stability correction mechanism is initiated through the control station. This includes analyzing the set of candidate relay locations in the area surrounding high-load nodes, searching for potential alternative nodes, activating backup relay drones, inserting buffer nodes downstream of high-load nodes, reallocating forwarding paths between nodes, and reducing the communication density of specific nodes.

[0024] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0025] This invention provides an indoor inspection signal enhancement system based on a multi-UAV dynamic relay network. By combining indoor environmental data with inspection task configuration files, this invention unifies the planning of UAV inspection paths and introduces equivalent communication path length as a basis for communication demand assessment. This ensures that inspection path planning and communication link assurance are coordinated, effectively avoiding communication interruptions caused by mismatches between path design and communication capabilities. The number of relay UAVs is dynamically determined based on the upper limit of single-hop communication distance, and three-dimensional relay deployment locations are generated in conjunction with building structures. This allows the relay deployment scale to adapt to different building structures and inspection task complexity, reducing redundancy or insufficiency of relay resources. This invention periodically collects link quality indicators to monitor the communication status between relay UAVs and upstream and downstream nodes in real time, and dynamically adjusts the hovering position of relay UAVs based on link quality changes, improving the continuity and stability of multi-hop communication links in complex indoor environments. By evaluating and screening the link quality of neighboring nodes, this invention ensures that inspection data is reliably transmitted to the control station along high-quality links, reducing the probability of data loss and communication interruptions. Furthermore, by comprehensively analyzing link quality and node load information, the routing reachability and deployment stability of the multi-hop relay network can be determined in a timely manner. In the event of communication anomalies or deployment imbalances, automatic route reconstruction, backup relay intervention, and relay deployment correction processing can be triggered, enabling the multi-hop relay network to possess adaptive recovery and optimization capabilities. This invention can significantly improve the reliability, network stability, and system robustness of indoor inspection communication, reduce the need for manual intervention, and enhance the engineering practicality of multi-UAV indoor inspection systems. Attached Figure Description

[0026] Figure 1 This is a flowchart of an indoor inspection signal enhancement system based on a multi-UAV dynamic relay network according to the present invention;

[0027] Figure 2 This is a schematic diagram of an indoor inspection signal enhancement system based on a multi-UAV dynamic relay network according to the present invention.

[0028] Figure 3 This is another schematic diagram of an indoor inspection signal enhancement system based on a multi-UAV dynamic relay network according to the present invention. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0030] like Figures 1-3 This embodiment of an indoor inspection signal enhancement system based on a multi-UAV dynamic relay network may specifically include:

[0031] Step S101, the inspection path planning module is used to determine the planned inspection path of the UAV based on indoor environmental data and inspection task configuration file, and calculate the equivalent communication path length of the planned inspection path.

[0032] The UAV inspection system acquires indoor environmental data, an indoor 3D spatial map, and an inspection task configuration file, and determines the planned inspection path for the UAV. The indoor environmental data includes building structure information, floor distribution information, wall position relationships, floor connection relationships, control station location, and the inspection UAV's task area information. The inspection task configuration file includes the inspection start position, inspection end position, and inspection area constraints. Based on the planned inspection path, the system discretizes the path at a preset step size to obtain several path nodes. The system acquires the path node sequence and their position information within the building structure, and marks the structural unit types traversed by each inspection path segment. Structural unit types include walls, floors, metal shielding surfaces, and cavity structures. The equivalent communication path length of the planned inspection path is calculated based on the additional equivalent length of the structural type traversed by each inspection path segment. This additional equivalent length is pre-calibrated based on empirical experiments or a simple path loss model.

[0033] For example, in a specific implementation scenario, a drone inspection system is deployed in a three-story office building with a total length of approximately 60 meters and a floor height of 3 meters. The control station is located at the entrance on the first floor. The inspection drone's starting point is at the first-floor entrance, and its ending point is in an equipment room on the innermost part of the third floor. The drone inspection system acquires an indoor 3D spatial map and an inspection task configuration file for the building. The indoor environmental data includes floor distribution information from the first to the third floor, the positional relationships of load-bearing walls and partition walls on each floor, the vertical connections of floor slabs, and the location of any existing metal shielding surfaces, such as the metal structure surrounding the elevator shaft. Based on the inspection task configuration file, the system determines the planned inspection path for the drone. This path starts from the first-floor entrance, ascends via stairs to the second floor, continues via stairs to the third floor, and finally enters the target equipment room. The total geometric length of the planned path is approximately 45 meters. Subsequently, the planned inspection path is discretized according to a preset step size of 3 meters, generating 15 path nodes along the entire path and forming 14 inspection path segments connected by adjacent path nodes. For each inspection path segment, the system identifies the type of structural unit traversed by the path segment based on building structure information. For example, 6 paths are located in open spaces such as corridors and stairwells and do not traverse walls; 4 paths traverse ordinary concrete walls; and 2 paths traverse floor slabs. Two additional paths are located near elevator shafts and are adjacent to metal shielding surfaces. The additional equivalent lengths corresponding to different structural units are pre-determined through empirical experiments: 2 meters for each ordinary wall traversed, 4 meters for each floor slab traversed, and 3 meters for each path segment adjacent to a metal shielding surface. Based on the above calibration results, the equivalent communication path length of the planned inspection path is calculated. That is, on the basis of the original geometric path length of 45 meters, an additional equivalent length of 8 meters is added by the wall, an additional equivalent length of 8 meters is added by the floor, and an additional equivalent length of 6 meters is added by the metal shielding surface. Finally, the equivalent communication path length of the planned inspection path is 67 meters. This equivalent communication path length is used for the subsequent calculation of the number of relays deployed and the planning of communication links.

[0034] Step S102, the relay demand prediction module is used to determine the minimum number of working relay drones and the number of backup relay drones required for the inspection task based on the upper limit of single-hop communication distance and the equivalent communication path length of the inspection path.

[0035] Historical data on the transmit power, operating frequency band, and receive sensitivity of wireless communication modules are obtained from the inspection communication detection database, and the upper limit of single-hop communication distance is marked. A recurrent neural network is used to train the model and construct a prediction model for the upper limit of single-hop communication distance. Based on the transmit power, operating frequency band, and receive sensitivity of the currently used wireless communication modules, the upper limit of single-hop communication distance for the current inspection task is predicted using the prediction model. Based on the upper limit of single-hop communication distance for the current inspection task and the equivalent communication path length of the planned inspection path, the minimum number of relay UAVs required for the current inspection task is calculated. Based on the importance of the current inspection task and the network reliability level, the number of backup relay UAVs matching the task importance and network reliability level is extracted from a preset redundancy ratio configuration table.

[0036] For example, the inspection communication detection database pre-stores historical operating data for various wireless communication modules, including transmit power, operating frequency band, receiver sensitivity, and records of stable single-hop communication distances under corresponding environmental conditions. Taking one type of wireless communication module as an example, its historical data shows that when the transmit power is 20dBm, the operating frequency band is 2.4GHz, and the receiver sensitivity is -90dBm, the stable single-hop communication distance measured in a typical indoor environment is concentrated between 18 meters and 22 meters. Using the above historical data as training samples, a recurrent neural network is used to learn and train the temporal correlation between wireless communication parameters and single-hop communication distance, thus constructing a single-hop communication distance upper limit prediction model. Subsequently, in a new inspection task, the transmit power of the currently used wireless communication module is detected to be 18dBm, the operating frequency band is still 2.4GHz, and the receiver sensitivity is -88dBm. This set of parameters is input into the trained single-hop communication distance upper limit prediction model for inference calculation. The model predicts that under the current communication conditions and indoor environmental characteristics, the upper limit of the single-hop communication distance for this inspection task is approximately 16 meters. Meanwhile, the system has calculated the equivalent communication path length of the planned inspection route to be 64 meters based on the building structure and path crossing conditions. This means the path is equivalent to a 64-meter unobstructed straight path in a communication sense. Based on this, the system compares the equivalent communication path length with the upper limit of single-hop communication distance, determining that, without considering redundancy, at least four working relay drones are needed to cover the entire inspection route, ensuring that the equivalent distance between adjacent communication nodes does not exceed 16 meters. Furthermore, based on the high importance of the task and the strong network reliability level set in the inspection task configuration file, the system reads the corresponding backup relay configuration rules from the preset redundancy ratio configuration table. For example, under high importance and strong reliability conditions, the number of backup relays is set to 50% of the number of working relays. Based on this rule, the system configures two additional backup relay drones on top of the four working relay drones, ultimately determining a total of six relay drones required for this inspection task: four working relay drones and two backup relay drones, used to handle link fluctuations or node failures during the inspection process.

[0037] Step S103, the multi-hop relay link construction module is used to obtain the location of the relay node and the corresponding three-dimensional relay deployment location data based on the building structure data and the upper limit of the single-hop communication distance, and generate relay deployment control command parameters.

[0038] By extracting entrance space nodes, staircase space nodes, passageway connection nodes, and inspection path-related space nodes from building structure data, a set of candidate relay node locations is obtained. Based on the upper limit of single-hop communication distance, the candidate relay node locations are sequentially filtered along the inspection path to determine the relay node positions. If the distance between adjacent candidate relay node locations exceeds the upper limit of single-hop communication distance, at least one supplementary relay location is generated on the corresponding inspection path segment according to an equidistant distribution rule, based on the path length between adjacent candidate relay node locations. Based on floor height description information and relay hovering height description parameters, the corresponding 3D relay deployment location data for the relay node locations and supplementary relay locations is determined, and relay deployment control command parameters are generated, including the 3D coordinates of the UAV deployment point, hovering height, identification, activation status, backup identification, and allocation order. The control station sends relay deployment control command parameters to each relay UAV, performs takeoff control, path tracking control, and hovering control on the relay UAVs, and sequentially flies each relay UAV to its corresponding target position, enters a stable hovering state, establishes a multi-hop relay link, and forms a communication channel from the control station, relay UAVs to the inspection UAVs.

[0039] For example, in a specific indoor inspection scenario, the inspection target is a four-story research and office building, with each floor approximately 3.2 meters high. The building's interior consists of an entrance hall, stairwells, a circular corridor, and several laboratory rooms. The UAV inspection system first extracts indoor space nodes based on building structure data. These include two entrance space nodes located in the first-floor lobby, four stairwell space nodes running from the first to the fourth floor, eight corridor connection nodes at the corners of each floor's corridors, and six space nodes directly related to the planned inspection path. This generates a set of 20 candidate relay node locations. Based on the communication module's performance parameters, the system sets the maximum single-hop communication distance to 15 meters and, starting from the first-floor entrance where the control station is located, sequentially filters the aforementioned set of candidate relay node locations along the planned inspection path. During the screening process, it was found that the path distance between the first-floor entrance node and the second-floor staircase node was approximately 12 meters, meeting the single-hop communication distance requirement. Therefore, this node was retained as a relay node location. However, the path distance between the second-floor staircase node and the third-floor corridor corner node was approximately 24 meters, significantly exceeding the 15-meter single-hop communication distance limit. To address this, the system, based on the inspected path length between these two candidate locations, inserted a supplementary relay location in the corresponding staircase-corridor connection path segment according to an equidistant distribution rule. This divided the path segment into two sub-path segments of approximately 12 meters in length, ensuring that the communication distance between any adjacent relay nodes does not exceed the single-hop communication distance limit. After checking and supplementing all path segments, five relay node locations and one supplementary relay location were ultimately determined. Subsequently, combining floor height description information and relay hovering height description parameters, corresponding three-dimensional relay deployment location data is generated for each relay node. For example, the hovering height of the first and second floor relay nodes is set to 2.5 meters above the ground, and the hovering height of the third and fourth floor relay nodes is set to 2.8 meters above the ground. Each relay node is assigned a unique identifier, an active status marker, and a backup identifier. Simultaneously, allocation order information is generated according to the order from nearest to farthest along the inspection path. Based on the above three-dimensional relay deployment location data, the system generates relay deployment control command parameters and issues corresponding commands to each relay drone through the control station. This controls each relay drone to take off sequentially, reach the designated spatial position along the predetermined flight path, and enter a stable hovering state. Thus, a multi-hop relay link is formed within the building, connecting the control station, the first-floor relay drone, the second-floor relay drone, the supplementary relay drone, and the third and fourth-floor relay drones in a hierarchical manner. This constructs a stable communication channel covering the entire inspection path, enabling the inspection drones to maintain communication with the control station through the relay drones throughout the inspection mission.

[0040] Step S104, the relay position adjustment module is used to monitor the link quality score between the relay UAV and the upstream and downstream relay nodes by periodically collecting link quality indicators, and adjust the hovering position of the relay UAV.

[0041] Based on the local temperature difference changes, local air disturbance velocity, material type, and span characteristics of the relay node and supplementary relay locations, it is determined whether the hovering position of the relay UAV is within the structural disturbance resonance sensitive zone, and the hovering point of the relay UAV is adjusted accordingly. Span characteristics include span size, span construction method, and span connection structure. Span construction methods include single-span and multi-span continuous beams, and span connection structures include fixed connections, semi-rigid connections, and hinged connections. Relay UAVs using multi-hop relay links send test data packets between their upstream and downstream relay nodes. During data forwarding by the relay UAVs, link quality indicators between each relay UAV and its upstream and downstream relay nodes are periodically collected according to a preset time window. Link quality indicators include received signal strength indication, link quality indication, signal-to-noise ratio, and packet loss rate. A weighted average of the link quality indicators is used to obtain the link quality score for each relay UAV and its upstream and downstream relay nodes. If the link quality score is lower than the preset score threshold for more than N consecutive times, the current hovering position of the relay drone is determined to not meet the communication quality requirements, and a position adjustment operation is performed according to the preset movement rules based on the current position of the relay drone. The link quality score is obtained after each movement of the relay drone. If the link quality score recovers to above the preset score threshold within a limited number of movements, the movement stops and the new hovering position is maintained. If the threshold cannot be reached within a preset maximum number of movements, a backup relay drone is controlled by the control station to take off and fill the gap.

[0042] For example, during an indoor inspection task, after a relay drone located in a third-floor corridor reaches its planned hovering point, the system first determines whether the hovering point is within a structural disturbance resonance sensitive zone based on the building structure data of that area. This relay node is located beneath a reinforced concrete floor slab with a span of 5.5 meters, a single-span structure. One end is fixedly connected to a load-bearing beam, and the other end is semi-rigidly connected. The system detects, through the drone's onboard temperature sensor, that the average change in local temperature per second near this location within 5 seconds is approximately 0.7℃, and the local air disturbance velocity is approximately 0.32 meters per second. Ultimately, the system determines that the relay drone's current hovering point is within a structural disturbance resonance sensitive zone, which may cause continuous attitude deviation and link jitter in the relay drone at this location. Therefore, according to preset hovering point adjustment rules, the system performs a 0.95-meter translation adjustment along the corridor's lateral direction without deviating from the relay node's functional coverage area. The system sets a preset time window of 10 seconds, meaning that every 10 seconds, test data packets are sent between the relay drone and its upstream and downstream relay nodes, and a statistical evaluation of the communication status between the relay drone and its upstream and downstream relay nodes is performed. When a relay drone located in the second-floor corridor is performing a data forwarding task, link quality indicators between the relay drone and its upstream and downstream relay nodes are periodically collected within multiple consecutive time windows. These indicators include received signal strength (RSS), link quality indicator (LSI), signal-to-noise ratio (SNR), and packet loss rate. For example, within a certain time window, the received SRS between the relay drone and its upstream node is detected to be approximately -78 dBm, with a LSI of 70, an SNR of approximately 18 dB, and a packet loss rate of approximately 6%. The received SRS between the relay drone and its downstream node is approximately -82 dBm, with a LSI of 65, an SNR of approximately 15 dB, and a packet loss rate of approximately 9%. The weighted average of the aforementioned link quality indicators yielded a comprehensive link quality score of 0.58 for the relay drone within the given time window, with a preset score threshold set at 0.65. In the subsequent three consecutive time windows, the link quality scores were 0.57, 0.55, and 0.54, respectively, all below the preset threshold. Furthermore, the number of consecutive times the score fell below the threshold reached a preset value N (e.g., N is set to 3). Based on this, it was determined that the relay drone's current hovering position no longer met the communication quality requirements. Subsequently, based on the relay drone's current position, a position adjustment operation was performed according to preset movement rules. These rules specifically referred to moving laterally along the corridor at 1-meter intervals without deviating from the original inspection path coverage area. After the relay drone completed its first position movement, the system re-collected the link quality indicators and calculated a new link quality score of 0.61, still below the threshold of 0.65. After the second position movement, the link quality score improved to 0.66, exceeding the preset score threshold.Based on this, the system determines that the current hovering position has been restored to a state that meets the communication quality requirements, stops further movement, and maintains the new hovering position to continue performing data forwarding tasks. If, in this scenario, the relay drone completes the preset maximum number of movements, such as 5 position adjustments, and the link quality score remains below 0.65, the control station will send takeoff and replacement commands to the backup relay drone, controlling the backup relay drone to fly to the area corresponding to the relay node to perform communication replacement, thereby maintaining the continuity of the multi-hop relay link.

[0043] Specifically, based on the local temperature difference, local air disturbance velocity, material type, and span characteristics of the relay node and supplementary relay locations, it is determined whether the hovering position of the relay UAV is in the structural disturbance resonance sensitive zone, and the hovering point of the relay UAV is adjusted accordingly.

[0044] Based on the relay node location and supplementary relay location, the material type and span characteristics of surrounding building components are obtained and stored in the inspection communication detection database. Historical data on the material type and span characteristics of building components are obtained from the inspection communication detection database, and the equivalent disturbance energy level of the building components is labeled. A recurrent neural network is used for model training to construct a prediction model of the equivalent disturbance energy level of building components, predicting the equivalent disturbance energy level of building components for the current inspection task. Temperature and airflow sensors from the relay UAV are used to obtain the local temperature difference change and local air disturbance velocity near the UAV deployment point, respectively. Combined with the equivalent disturbance energy level of the building components, the structural disturbance coupling index formula is used. Calculate the structural disturbance coupling index of the current relay UAV deployment site. ,in This represents the local temperature change of the building structure near the deployment point per unit time. The local air disturbance velocity detected at the deployment point location. This represents the equivalent disturbance energy level of the building structure components in this region. If the structural disturbance coupling index of the current relay UAV deployment point is greater than the preset index threshold, it is determined that the location is in a structural disturbance resonance sensitive area and is not suitable as a relay hovering point. An alternative hovering point that meets the condition of having a structural disturbance coupling index lower than the threshold is searched in the neighborhood of the deployment point, and the corresponding relay deployment control command parameters are updated.

[0045] For example, in high-precision hovering operations of indoor drones, a hidden challenge originates from the building itself. Walls, beams, columns, and metal frames undergo continuous micron-level deformation and vibration due to temperature differences, vibrations, and equipment operation. These subtle disturbances accumulate and couple within the confined space, causing imperceptible periodic "micro-fluctuations" in the local airflow field and electromagnetic environment. When the drone hovers, its sensitive flight control system constantly counteracts the resulting air turbulence, leading to small, continuous oscillations in the drone's attitude. These oscillations not only affect positioning accuracy but also alter antenna orientation and signal multipath propagation paths, ultimately causing periodic jitter in communication link strength and latency, severely threatening the real-time stability of control commands and data transmission. This deep-seated physical coupling effect is often simply attributed to poor signal, but it is actually a fundamental environmental challenge that must be overcome to achieve reliable indoor autonomous operation, and therefore, this situation needs to be eliminated. In an indoor inspection mission, the system first detects the material type and span characteristics of the surrounding building components based on the planned relay node location. The relay node is located beneath a reinforced concrete floor slab with a span of 5.5 meters. It is a single-span structure, with one end fixedly connected to a load-bearing beam and the other end semi-rigidly connected. The system stores the corresponding material type (reinforced concrete) and span characteristics (including span size of 5.5 meters, single-span construction, and connection structures including fixed and semi-rigid connections) in the inspection communication database. Span characteristics include span size, construction method, and connection structure. Construction methods include single-span and multi-span continuous beams, and connection structures include fixed, semi-rigid, and hinged connections. Before the task begins, historical data on similar floor slab structures is retrieved from the inspection communication database. This historical data records the response of building components under temperature disturbances and equipment operating loads with different materials and span characteristics, and assigns equivalent disturbance energy levels to each type of component. Subsequently, the system uses a recurrent neural network to train on historical data, constructing a predictive model for the equivalent disturbance energy level of building components. Based on the material and span characteristics of the current floor slab, it predicts that the equivalent disturbance energy level of the floor slab under the current inspection environment is 2.1. After the relay drone arrives at the deployment point, the system activates the onboard temperature and airflow sensors to monitor the microenvironment near the deployment point in real time. During a continuous 5-second measurement, the temperature sensor recorded an average change in local temperature difference of approximately 0.7°C per second, and the airflow sensor measured a local air disturbance velocity of approximately 0.32 meters per second at the deployment point. These two measured parameters and the predicted equivalent disturbance energy level are substituted into the structural disturbance coupling index formula. Calculations are performed, including the local temperature difference change. The air turbulence velocity is 0.7℃. The equivalent perturbation energy level is 0.32 m / s. The calculated value is 2.1, resulting in a structural disturbance coupling index of approximately 0.127 for this deployment point. The system's preset structural disturbance threshold is 0.10, used to distinguish whether a deployment point is located in a structural disturbance resonance sensitive area. Since the calculated value of 0.127 is greater than the threshold of 0.10, the system determines that the floor area at the current location of the relay UAV exhibits a significant structural disturbance coupling effect, potentially leading to continuous attitude drift or local link fluctuations during UAV hovering. Therefore, the system considers this deployment point unsuitable as a relay hovering point. Based on this determination, the system, following a preset alternative hovering point search strategy, searches for alternative points with a structural disturbance coupling index lower than the threshold within a 1.2-meter radius of the current deployment point. After three fine-tuning adjustments, the system finds a new hovering point approximately 0.95 meters from the original location. At the alternative hovering point, the relay UAV re-collected environmental parameters and measured a new structural disturbance coupling index of approximately 0.063. Since 0.063 is lower than the preset threshold of 0.10, the system confirmed that the alternative point was not located in the structural disturbance resonance sensitive area and could be used as a stable relay hovering position. The system then updated the alternative position to the relay deployment control command parameters, enabling the relay UAV to enter a stable hovering state at this point. After completing the structural disturbance detection process, the UAV continued to be used for the construction of multi-hop relay links and data forwarding.

[0046] Step S105, the data forwarding path selection module is used to eliminate neighboring nodes with abnormal quality based on the link quality score between each relay UAV and neighboring nodes, and select the neighboring node of the next hop forwarding object to forward the inspection task data to the control station.

[0047] The inspection drones, controlled by the control station, take off from the building entrance and fly along the planned inspection path to perform inspection tasks, continuously collecting inspection data. This data, including image and environmental data, is then transmitted to the nearest relay drone. The relay drones initiate an ad hoc network protocol to obtain a list of communicable neighboring nodes, send the collected inspection data to these nodes, and collect link quality indicators during communication. A weighted average of these link quality indicators is used to obtain a link quality score between each relay drone and its neighboring nodes. If the link quality score is lower than a preset threshold, the neighboring node is marked as an abnormal node and removed from the next-hop forwarding process. Based on the remaining neighboring nodes' identifiers, relative positions, link quality scores, and current forwarding load indicators, a path cost evaluation formula is used. Calculate the cost of the current relay drone i choosing neighbor node j as the next hop. The node with the lowest cost is selected as the next-hop forwarding target, and the inspection task data is forwarded to the control station. The forwarding load metric is the ratio of the amount of data currently cached by a node to its maximum cache capacity. Let be the estimated number of hops from node j to control station 1. To score the link quality, Let α be the current forwarding load metric for node j, and let β and γ be the weighting coefficients obtained by fitting historical data.

[0048] For example, an inspection drone takes off from the building's first-floor entrance and enters the building along a pre-planned inspection path to perform its inspection task. During flight, the inspection drone continuously collects inspection task data, including image data and environmental data. The image data consists of high-resolution images of corridors and equipment areas, while the environmental data includes temperature, humidity, and gas concentration information. The collected inspection task data is first sent to the nearest relay drone, which then forwards the data hop-by-hop towards the control station. If a relay drone i is currently forwarding data, its neighboring nodes available for the next hop include two relay drones, j1 and j2. The system collects link quality metrics between the relay UAV i and its neighboring nodes j1 and j2 within a preset time window. For example, the received signal strength with node j1 is -75dBm, the link quality indicator is 72, the signal-to-noise ratio is 20dB, and the packet loss rate is 5%; the received signal strength with node j2 is -82dBm, the link quality indicator is 65, the signal-to-noise ratio is 15dB, and the packet loss rate is 10%. The system performs a weighted average of the above link quality metrics to obtain a link quality score, where the link quality score Q corresponding to node j1 is... ij1 The link quality score Q for node j2 is 0.80. ij2 The score is 0.60. Since the system's preset scoring threshold is 0.65, node j2's link quality score is below this threshold, therefore it is marked as an abnormal quality node and removed from the candidate next-hop nodes. For the remaining node j1, the system further obtains its estimated hop count H to the control station. j1 For example, based on the current topology, it is determined that two hops are still needed from node j1 to the control station. Simultaneously, the current forwarding load metric L of node j1 is obtained. j1 For example, if node j1 currently has 30% of its maximum cached data to be forwarded, the corresponding load metric is 0.30. The path cost evaluation formula is used. Calculate the current relay drone i Select a residential node j As the price of the next jump Where α is 0.4, reflecting the impact of hop count on path cost; β is 0.4, reflecting the impact of link quality on path cost; and γ is 0.2, reflecting the impact of node load on path cost. Substituting the parameters of node j1 into the above formula, the path cost of the current relay UAV i selecting node j1 as the next hop is 1.32. Since only node j1 meets the conditions after removing nodes with abnormal quality, node j1 is ultimately selected as the next hop forwarding object. The image data and environmental data collected by the inspection UAV are then forwarded to the control station via node j1, ultimately reaching the control station, thus completing the path selection and data transmission within this forwarding cycle.

[0049] Step S106, the relay stability control module is used to determine the routing reachability status and relay deployment stability deviation of the multi-hop relay network by periodically collecting the link quality score and load information between each relay drone and its neighboring nodes, and to trigger route reconstruction, backup relay intervention and relay deployment correction processing.

[0050] According to a preset time window, link quality scores between each relay drone and its neighboring nodes are periodically collected during data forwarding. If a relay drone detects that the link quality score between itself and all its neighboring nodes is below the score threshold or that more than M consecutive data forwarding failures have occurred, it broadcasts a route reconstruction request to surrounding nodes. Based on the reachability and link quality indicators returned to the control station by the neighboring nodes that receive the route reconstruction request, an available path is reconstructed. If the path cannot be reconstructed within the preset time threshold, an anomaly is reported to the control station, and a backup relay drone is triggered to join the network to reconstruct an available path. The load information of each relay node per unit time is statistically analyzed through communication logs, and the relay deployment stability deviation assessment formula is used. Calculate the relay deployment stability deviation value Where K is the total number of relay drones currently in operation. Let be the forwarding load metric of the i-th relay node per unit time. The value is the natural logarithm function, used to enhance sensitivity to distribution sparsity. If the relay deployment stability deviation is less than the preset stability threshold, the relay deployment is considered relatively balanced and no adjustment is needed. If the relay deployment stability deviation is greater than the preset stability threshold, the relay deployment is considered unstable, and a relay stability correction mechanism is initiated through the control station. This includes analyzing the set of candidate relay locations in the area surrounding high-load nodes, searching for potential alternative nodes, activating backup relay drones, inserting buffer nodes downstream of high-load nodes, reallocating forwarding paths between nodes, and reducing the communication density of specific nodes.

[0051] For example, in an indoor inspection task, there are currently four relay drones operating in the network, denoted as Relay A, Relay B, Relay C, and Relay D. Therefore, the total number of relay drones currently operating is K=4. The system's preset time window is 60 seconds. Within this time window, each relay drone, while performing inspection data forwarding, periodically collects the link quality score between itself and its neighboring nodes. The link quality score ranges from 0 to 1, with a higher value indicating a more stable communication link. The system's preset link quality score threshold is 0.35, and the threshold for consecutive data forwarding failures is M=3. Within a certain time window, the link quality scores between Relay B and its three neighboring relay nodes are 0.31, 0.29, and 0.33, respectively, all below the preset score threshold of 0.35. Simultaneously, Relay B experienced three consecutive data forwarding failures during its most recent data forwarding process. Therefore, Relay B determines that the current path communication conditions have deteriorated and broadcasts a route reconstruction request to its surrounding neighboring nodes. Upon receiving the request, neighboring nodes report their reachability information and corresponding link quality metrics to the control station. For example, neighbor node E reports a usable multi-hop path to the control station with a comprehensive link quality score of 0.58, while neighbor node F reports a link quality score of 0.44. Based on the returned information, the control station prioritizes node E with the higher link quality to attempt to rebuild the usable path. However, due to severe wall obstruction in the area, a stable communication path is not established within the preset 5-second time threshold. Therefore, relay B reports an anomaly to the control station, which then dispatches a backup relay drone to join the network, supplementing the connection between relay B and the control station with a new forwarding node, thus completing the path reconstruction. Simultaneously, the system uses communication logs to statistically analyze the forwarding load metrics of each relay node within this 60-second time window. These forwarding load metrics characterize the current data forwarding pressure on the relay nodes, specifically quantified by the proportion of data to be forwarded in the node's cache to its maximum cache capacity, with a value ranging from 0 to 1. Statistical results show that the current cache utilization rate of relay A is 0.20%, relay B is 0.75%, relay C is 0.25%, and relay D is 0.30%, indicating that the forwarding pressure of relay B is significantly higher than that of other nodes. This is based on the relay deployment stability deviation assessment formula. The stability deviation value of relay deployment was calculated. The value is 0.29, where K is the total number of relay drones currently in operation. Let be the forwarding load metric of the i-th relay node per unit time. The natural logarithm function is used to enhance sensitivity to sparse distribution. The +1 is to avoid the problem of an undefined logarithm when the number of forwardings is 0. The natural logarithm function amplifies the impact of differences in the number of forwardings when the distribution is sparse. If the system's preset stability threshold is 0.25, and the current stability deviation value is greater than this threshold, it indicates an uneven distribution of forwarding load among multiple relay nodes, suggesting unstable relay deployment. Based on this judgment, the control station initiates a relay stability correction mechanism, analyzes the area surrounding relay B, generates multiple feasible candidate relay locations, and schedules a backup relay UAV to insert a buffer node downstream of relay B. Simultaneously, some forwarding paths are reallocated, distributing some inspection data flow from relay B to the newly added relay node, and appropriately reducing the communication density of relay B. After these adjustments, the forwarding load indicators of each relay node gradually converge, the relay deployment stability deviation value drops below the stability threshold, and the system returns to a stable operating state.

[0052] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to the technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the concept of this application. The exemplary features described above are examples of technical solutions formed by mutually substituting with (but not limited to) technical features with similar functions disclosed in this application.

Claims

1. An indoor inspection signal enhancement system based on a multi-UAV dynamic relay network, characterized in that, The system includes: The inspection path planning module is used to determine the planned inspection path of the UAV based on indoor environmental data and inspection task configuration files, and to calculate the equivalent communication path length of the planned inspection path. The relay demand prediction module is used to determine the minimum number of working relay drones and the number of backup relay drones required for the inspection task based on the upper limit of single-hop communication distance and the equivalent communication path length of the inspection path. The multi-hop relay link construction module is used to obtain the location of relay nodes and the corresponding three-dimensional relay deployment location data based on building structure data and the upper limit of single-hop communication distance, and generate relay deployment control command parameters. The relay position adjustment module is used to periodically collect link quality indicators, monitor the link quality scores between the relay drone and upstream and downstream relay nodes, and adjust the hovering position of the relay drone. The data forwarding path selection module is used to eliminate neighboring nodes with abnormal quality based on the link quality score between each relay drone and its neighboring nodes, and select the neighboring node of the next hop forwarding object to forward the inspection task data to the control station. The relay stability control module is used to periodically collect link quality scores and load information between each relay drone and its neighboring nodes to determine the routing reachability status and relay deployment stability deviation of the multi-hop relay network, and to trigger route reconstruction, backup relay intervention, and relay deployment correction processing.

2. The system according to claim 1, wherein, The inspection path planning module is used to determine the planned inspection path of the UAV based on indoor environmental data and inspection task configuration files, and to calculate the equivalent communication path length of the planned inspection path, including: The UAV inspection system acquires indoor environmental data, an indoor 3D spatial map, and an inspection task configuration file, and determines the planned inspection path for the UAV. The indoor environmental data includes building structure information, floor distribution information, wall position relationships, floor connection relationships, control station location, and the UAV's task area information. The inspection task configuration file includes the inspection start position, inspection end position, and inspection area constraints. Based on the planned inspection path, the system discretizes the path at a preset step size to obtain several path nodes. The system acquires the path node sequence and their position information within the building structure, and marks the structural unit types traversed by each inspection path segment. Structural unit types include walls, floors, metal shielding surfaces, and cavity structures. The equivalent communication path length of the planned inspection path is calculated based on the additional equivalent length of the structural type traversed by each inspection path segment. This additional equivalent length is pre-calibrated based on empirical experiments or a simple path loss model.

3. The system according to claim 1, wherein, The relay demand prediction module is used to determine the minimum number of working relay drones and the number of backup relay drones required for the inspection task based on the upper limit of single-hop communication distance and the equivalent communication path length of the inspection path, including: By using the inspection communication detection database, historical data on the transmit power, operating frequency band, and receiver sensitivity of wireless communication modules are obtained, and the upper limit of single-hop communication distance is marked. A recurrent neural network is used to train the model and construct a prediction model for the upper limit of single-hop communication distance. Based on the transmit power, operating frequency band, and receiver sensitivity of the currently used wireless communication module, the upper limit of single-hop communication distance is predicted using the prediction model for the upper limit of single-hop communication distance for the current inspection task. Based on the upper limit of the single-hop communication distance of the current inspection task and the equivalent communication path length of the planned inspection path, calculate the minimum number of relay drones required for the current inspection task; according to the importance of the current inspection task and the network reliability level, extract the number of backup relay drones that match the importance of the task and the network reliability level from the preset redundancy ratio configuration table.

4. The system according to claim 1, wherein, The multi-hop relay link construction module is used to obtain the relay node location and corresponding three-dimensional relay deployment location data based on building structure data and the upper limit of single-hop communication distance, and generate relay deployment control command parameters, including: By extracting entrance space nodes, staircase space nodes, passageway connection nodes, and inspection path-related space nodes from building structure data, a set of candidate relay node locations is obtained. Based on the upper limit of single-hop communication distance, the candidate relay node locations are sequentially filtered along the inspection path to determine the relay node location. If the distance between adjacent candidate relay node locations exceeds the upper limit of single-hop communication distance, at least one supplementary relay location is generated on the corresponding inspection path segment according to an equidistant distribution rule, based on the path length between adjacent candidate relay node locations. Finally, based on floor height description information and relay hovering height description information... The parameters are described above to determine the three-dimensional relay deployment location data corresponding to the relay node location and the supplementary relay location, and to generate relay deployment control command parameters, including the three-dimensional position coordinates of the UAV deployment point, hovering altitude, identification, activation status, backup identification, and allocation order; the relay deployment control command parameters are sent to each relay UAV through the control station to perform take-off control, path tracking control, and hovering control on the relay UAV, and each relay UAV is sequentially flown to the corresponding target location, enters a stable hovering state, establishes a multi-hop relay link, and forms a communication channel from the control station, relay UAVs to inspection UAVs.

5. The system according to claim 1, wherein, The relay position adjustment module is used to monitor the link quality scores between the relay UAV and upstream and downstream relay nodes by periodically collecting link quality indicators, and to adjust the hovering position of the relay UAV, including: Based on the local temperature difference changes, local air disturbance velocity, material type, and span characteristics at the relay node and supplementary relay locations, it is determined whether the hovering position of the relay UAV is within the structural disturbance resonance sensitive zone, and the hovering point of the relay UAV is adjusted accordingly. Span characteristics include span size, span construction method, and span connection structure. Span construction methods include single-span and multi-span continuous beams, and span connection structures include fixed connections, semi-rigid connections, and hinged connections. Relay UAVs using multi-hop relay links send test data packets between their upstream and downstream relay nodes. During data forwarding by the relay UAVs, link quality indicators between each relay UAV and its upstream and downstream relay nodes are periodically collected according to a preset time window. Link quality indicators include received signal strength. The system calculates the link quality index, link quality index, signal-to-noise ratio, and packet loss rate. A weighted average of these indicators is used to obtain the link quality score for each relay UAV and its upstream and downstream relay nodes. If the link quality score is lower than a preset threshold for more than N consecutive times, the current hovering position of the relay UAV is deemed not to meet communication quality requirements, and a position adjustment operation is performed according to preset movement rules based on the UAV's current position. The system obtains the link quality score after each movement of the relay UAV. If the link quality score recovers to above the preset threshold within a limited number of movements, the UAV stops moving and maintains the new hovering position. If the threshold cannot be reached within a preset maximum number of movements, a backup relay UAV is controlled by the control station to take off and fill the gap.

6. The system according to claim 5, wherein, The process of determining whether the hovering position of the relay UAV is within the structural disturbance resonance sensitive zone based on the local temperature difference change, local air disturbance velocity, material type, and span characteristics of the relay node location and the supplementary relay location, and adjusting the hovering point of the relay UAV, includes: Based on the relay node location and supplementary relay location, the material type and span characteristics of the building components surrounding these locations are obtained and stored in the inspection communication detection database. Historical data on the material type and span characteristics of the building components are obtained from this database, and the equivalent disturbance energy levels of the building components are labeled. A recurrent neural network is used for model training to construct a prediction model for the equivalent disturbance energy levels of the building components, predicting the equivalent disturbance energy levels of the building components for the current inspection task. Temperature and airflow sensors from the relay UAV are used to obtain the local temperature difference changes and local air disturbance velocity near the UAV deployment point. Combined with the equivalent disturbance energy levels of the building components, the structural disturbance coupling index formula is applied. Calculate the structural disturbance coupling index of the current relay UAV deployment site. ,in This represents the local temperature change of the building structure near the deployment point per unit time. The local air disturbance velocity detected at the deployment point location. The equivalent disturbance energy level of the building structure components in this area; if the structural disturbance coupling index of the current relay UAV deployment point is greater than the preset index threshold, it is determined that the location is in the structural disturbance resonance sensitive area and is not suitable as a relay hovering point. Then, an alternative hovering point that meets the condition of having a structural disturbance coupling index lower than the threshold is searched in the neighborhood of the deployment point, and the corresponding relay deployment control command parameters are updated.

7. The system according to claim 1, wherein, The data forwarding path selection module is used to eliminate neighboring nodes with abnormal link quality based on the link quality score between each relay UAV and its neighboring nodes, and select the neighboring node of the next-hop forwarding object to forward the inspection task data to the control station, including: The inspection drones, controlled by the control station, take off from the building entrance and fly along the planned inspection path to perform inspection tasks, continuously collecting inspection data. This data, including image and environmental data, is then transmitted to the nearest relay drone. The relay drones initiate an ad hoc network protocol to obtain a list of communicable neighboring nodes, sending the collected inspection data to these nodes and collecting link quality indicators during communication. A weighted average of these link quality indicators is used to obtain a link quality score between each relay drone and its neighboring nodes. If the link quality score is lower than a preset threshold, the neighboring node is marked as an abnormal node and removed from the next-hop forwarding process. Based on the remaining neighboring nodes' identifiers, relative positions, link quality scores, and current forwarding load indicators, a path cost evaluation formula is used. Calculate the cost of the current relay drone i choosing neighbor node j as the next hop. The node with the lowest cost is selected as the next-hop forwarding target, and the inspection task data is forwarded to the control station. The forwarding load metric is the ratio of the amount of data currently cached by a node to its maximum cache capacity. Let be the estimated number of hops from node j to control station 1. To score the link quality, Let α be the current forwarding load metric for node j, and let β and γ be the weighting coefficients obtained by fitting historical data.

8. The system according to claim 1, wherein, The relay stability control module is used to periodically collect link quality scores and load information between each relay drone and its neighboring nodes to determine the route reachability status and relay deployment stability deviation of the multi-hop relay network, and trigger route reconstruction, backup relay intervention, and relay deployment correction processing, including: According to the preset time window, the link quality score between each relay drone and its neighboring nodes is periodically collected during the data forwarding process. If the relay drone detects that the link quality score between itself and all its neighboring nodes is lower than the score threshold or that more than M consecutive data forwarding failures occur, it broadcasts a route reconstruction request to the surrounding nodes. Based on the reachability and link quality indicators returned to the control station by the neighboring nodes that received the route reconstruction request, the available path is reconstructed. If the path cannot be rebuilt within the preset time threshold, an anomaly is reported to the control station, and a backup relay drone is triggered to join the network to rebuild a usable path. By analyzing the load information of each relay node per unit time through communication logs, the relay deployment stability deviation assessment formula is used. Calculate the relay deployment stability deviation value Where K is the total number of relay drones currently in operation. Let be the forwarding load metric of the i-th relay node per unit time. It is the natural logarithm function, used to enhance the sensitivity to the sparsity of the distribution; If the relay deployment stability deviation is less than the preset stability threshold, it indicates that the relay deployment is relatively balanced and no adjustment is needed. If the relay deployment stability deviation is greater than the preset stability threshold, the relay deployment is considered unstable, and the relay stability correction mechanism is activated through the control station. This includes analyzing the set of candidate relay locations in the area surrounding high-load nodes, searching for potential alternative nodes, activating backup relay drones, inserting buffer nodes downstream of high-load nodes, reallocating forwarding paths between nodes, and reducing the communication density of specific nodes.

Images