Unmanned aerial vehicle cooperative multi-strategy dynamic monitoring method for multiple monitoring points of a construction site

Abstract

The application discloses a kind of unmanned aerial vehicle coordination multi-strategy dynamic monitoring methods for construction site multiple monitoring points, first, build computable digital twin scene;Then according to the sampling frequency of monitoring point, spatial distribution and priority, calculate the best monitoring area for each monitoring point, and on this basis, respectively decision to adopt differentiated strategy, generate monitoring scheme.During task execution, continuously monitor the power of unmanned aerial vehicle, task progress, charging platform usage, through task waiting queue and redundancy capability evaluation mechanism, the active intervention task triggered by power warning, new or released task, charging completion and fatigue alarm is dynamically rescheduled, and task relay and path re-planning are realized between working unmanned aerial vehicle and standby unmanned aerial vehicle.The application can realize efficient cooperative monitoring of unmanned aerial vehicle cluster under the condition of multiple monitoring points in complex dynamic construction site environment.

Description

Technical Field

[0001] This invention relates to the field of drone monitoring and scheduling optimization technology, specifically to a collaborative multi-strategy dynamic monitoring method for drones with multiple monitoring points at construction sites. Background Technology

[0002] With the increasing demands for digitalization and safety supervision in the construction industry, construction sites are facing comprehensive monitoring needs in terms of high-frequency, long-term, and near-real-time inspections, monitoring, and early warning systems, covering aspects such as construction quality, safety risks, environmental conditions, and progress control. Drones, due to their convenient deployment, high mobility, and wide coverage, have gradually become an important means of construction site monitoring. However, their practical application in complex construction site environments still faces several challenges.

[0003] On the one hand, construction site spatial structures are highly dynamic and exhibit significant three-dimensional characteristics. Scaffolding, tower cranes, temporary structures, material storage areas, and other facilities are constantly added to and removed from the site, and their locations frequently change. The height and outline of the buildings themselves change continuously during construction, creating temporally varying occlusion relationships and height restrictions. Safe flight zones and visible monitoring areas change with the construction phase. Many existing drone inspection solutions still plan flight routes based on two-dimensional or near-static scenes, making it difficult to accurately depict three-dimensional visual field occlusion and dynamic no-fly zones. This results in limited observation lines of sight, numerous blind spots, and path planning and observation location selection that are easily disturbed, making organization quite challenging.

[0004] On the other hand, at the level of monitoring task strategy and scheduling, existing solutions mostly adopt fixed-route patrols or dispatching tasks on an average basis according to monitoring points. They usually assume "one aircraft corresponds to several points" or adopt a simple periodic inspection mode, lacking refined strategy selection based on the sampling frequency, spatial distribution, and priority differences of different monitoring points. Task allocation, monitoring sequence arrangement, flight path planning, and accessibility of charging platforms and queuing factors are often handled in isolation, making it difficult to comprehensively optimize under a unified model. This easily leads to duplicate flights, high empty-range ratios, and high energy consumption, as well as monitoring gaps or delays at important monitoring points, resulting in poor overall scheduling efficiency and operational stability.

[0005] On the other hand, short flight time and low recharging efficiency are the core bottlenecks restricting the large-scale, long-term application of drones in continuous operations. Existing solutions generally rely on a few fixed charging stations for periodic inspections. When the work area is far from the charging station, the drones will consume a lot of time and energy to go back and forth to charge, significantly reducing the effective working time. Once the layout of fixed charging stations is determined, it is difficult to flexibly adjust with the shift of the construction focus. Moreover, there is a lack of unified modeling and incorporation of factors such as charging platform capacity, queuing time, charging efficiency, and energy consumption of drone return paths into scheduling decisions. Energy configuration is rigid and cannot achieve dynamic optimization at the drone swarm level.

[0006] Furthermore, existing monitoring systems typically treat "monitoring quality" and "personnel safety" as independent subsystems, lacking a closed-loop mechanism that links worker physiological states with drone dispatch. This prevents them from quickly dispatching drones to the site to proactively intervene, such as issuing verbal warnings, providing focused lighting, or conducting video checks, once fatigue risks are detected. Regarding personnel safety, traditional construction sites primarily rely on manual inspections by safety officers and worker self-reporting. They lack real-time proactive monitoring and early warning mechanisms based on physiological parameters for potential physiological limits such as fatigue and heatstroke that may occur under high-temperature, high-humidity, and high-intensity working conditions in summer. In extreme cases like heatstroke, traditional monitoring models relying on video playback and post-incident verification are insufficient for timely detection and intervention, failing to effectively provide real-time, individualized safety care for frontline workers. Summary of the Invention

[0007] The present invention aims to at least partially solve one of the technical problems existing in the related art.

[0008] The purpose of this invention is to provide a collaborative multi-strategy dynamic monitoring method for UAVs with multiple monitoring points on construction sites. In complex and dynamic construction site environments, driven by the sampling frequency, priority, and spatial distribution of monitoring points, and under a unified three-dimensional digital twin model of the construction site, the method dynamically selects differentiated monitoring strategies, jointly optimizes monitoring task matching, optimal monitoring area and observation position, time-division multiplexing of monitoring sequences, and three-dimensional path planning, so as to achieve efficient collaborative monitoring of UAV swarms in multi-monitoring environment.

[0009] To achieve the above objectives, the present invention provides a method for collaborative multi-strategy dynamic monitoring of drones at multiple monitoring points on a construction site. This method includes the following steps:

[0010] Step 1: Obtain a 3D map, obstacle information, and no-fly zone information of the construction site to be monitored. Combine this with the monitoring requirements of all monitoring points, as well as the drone swarm status data and drone energy network status data, to construct a computable digital twin scenario.

[0011] Step 2: Based on the monitoring requirements of each monitoring point, calculate the three-dimensional feasible observation space of each monitoring point that satisfies the observation distance constraint, the line of sight unobstructed constraint, and the no-fly zone avoidance constraint, and use it as the optimal monitoring area for that monitoring point.

[0012] Step 3: Based on the monitoring requirements and optimal monitoring areas of each monitoring point, divide all monitoring points into independent monitoring points consisting of a single monitoring point and monitoring point groups consisting of multiple monitoring points; and formulate monitoring strategies for each independent monitoring point and monitoring point group.

[0013] Step 4: Based on the current drone cluster status data, determine the set of available drones that can perform monitoring tasks, and generate a matching scheme between drones and each independent monitoring point or monitoring point group according to the established monitoring strategy. Plan the monitoring sequence and flight path of each drone, and assign the corresponding drone to perform the monitoring task.

[0014] Step 5: During the monitoring mission, continuously acquire drone cluster status data and drone energy network status data; when a drone's battery level is detected to be below a preset threshold, release the drone's unfinished monitoring mission and add it to the mission waiting queue. Through redundancy assessment, redistribute the mission between the flying drone and the standby drone, and replan the monitoring sequence and flight path of the relevant drones to achieve mission relay; at the same time, based on the current location, battery status, and mission urgency level of the low-battery drone, as well as the drone's energy network status data, designate the target refueling platform for the drone, and plan a three-dimensional obstacle avoidance return path for the drone to the target refueling platform.

[0015] A further preferred technical solution of the present invention is that, in step two, based on the monitoring requirements of each monitoring point, a three-dimensional feasible observation space is calculated for each monitoring point to satisfy the constraints of observation distance, unobstructed line of sight, and no-fly zone avoidance, which is then used as the optimal monitoring area for that monitoring point; specifically:

[0016] Best monitoring area Defined as a set of three-dimensional points satisfying multiple constraints, expressed as:

[0017] ;

[0018] in, Possible observation locations for the drone. The coordinates of the monitoring point; To constrain the observation distance, the observation distance must be within the minimum safe distance and based on the ground sampling resolution. Between the calculated maximum effective observation distances; To ensure unobstructed line of sight, a ray tracing algorithm is used to determine the set of observation lines and obstacles. No intersection; To constrain the observation perspective, the angle between the observation vector and the vertical direction and the normal vector of the measured surface is limited; To circumvent restrictions in no-fly zones and ensure that observation locations are not situated within static no-fly zones. or dynamic tower crane no-fly zone internal.

[0019] As a preferred option, step three specifically involves:

[0020] When the sampling frequency or priority of a monitoring point is high, and the best monitoring area of ​​the monitoring point has no usable intersection with the best monitoring areas of other monitoring points, the monitoring point is treated as an independent monitoring point, and a one-to-one dedicated monitoring strategy is selected. Based on the sampling frequency of the monitoring point, an access time series is generated, and a dedicated drone is designated to monitor the monitoring point according to the access time series.

[0021] When multiple monitoring points have low sampling frequencies and their optimal monitoring areas have available intersections, the multiple monitoring points are divided into the same monitoring point group, a one-to-many static monitoring strategy is selected, and a single hovering observation position is determined in the available intersection. The UAV is then deployed to that position, and time-division multiplexing observations are performed on multiple monitoring points through gimbal attitude adjustment.

[0022] When multiple monitoring points have low sampling frequencies and their optimal monitoring areas have no usable overlap but are spatially close to each other, the multiple monitoring points are divided into the same monitoring point group, a one-to-many dynamic monitoring strategy is selected, and a round-trip flight path is generated between multiple monitoring areas, and the drone is designated to observe in sequence according to the monitoring sequence.

[0023] When the sampling frequency of a monitoring point is low, and its location is remote or its optimal monitoring area is small and has no usable overlap with the optimal monitoring areas of other monitoring points, the monitoring point is treated as an independent monitoring point. A many-to-one collaborative monitoring strategy is selected, and the observation task of the monitoring point is inserted into the task gap of multiple assigned task drones to complete collaboratively.

[0024] As a preferred option, step four specifically involves:

[0025] First, based on the monitoring strategies formulated by each independent monitoring point or monitoring point group, the minimum number of drones required to perform all monitoring tasks is estimated, and based on the current drone swarm status data, the set of available drones is determined.

[0026] For monitoring points employing a one-to-one dedicated monitoring strategy, their corresponding optimal monitoring area Internally, numerical optimization methods are used to solve for the optimal monitoring position for the assigned dedicated UAV under multi-target conditions. , is represented as:

[0027] ;

[0028] in, Indicates the drone's takeoff point The flight energy required to reach the candidate observation point v is obtained by time integration based on the dynamics model and power model of the UAV; Indicates at candidate observation points The predicted path loss of the communication link between the UAV and the monitoring center is calculated using the free space propagation model and the shadow fading model. It is a penalty function that measures the stability of the wind field at that location, calculated based on historical wind field data or CFD simulation results; These are the weights for energy consumption, communication link loss, and wind farm stability penalty, respectively.

[0029] For a monitoring point group employing a one-to-many static monitoring strategy, select an optimal static hovering point within the public area. To balance the observation quality of multiple targets with link performance; let the comprehensive observation quality function be:

[0030] ;

[0031] in and These are normalized quality scores based on viewpoint and distance, respectively. To be based on monitoring weight The obtained normalized coefficients, These are, respectively, the view quality weight, the distance quality weight, and the communication loss penalty weight;

[0032] An improved particle swarm optimization or multi-starting-point gradient descent method is employed to search for the maximization within a common region. The point is used to obtain the optimal static monitoring position for the drone. In subsequent missions, the drone only needs to remain in the optimal static monitoring position. The system can hover nearby and perform time-sharing patrols of each monitoring point in the monitoring point group by controlling the attitude of the gimbal.

[0033] For a monitoring point group that adopts a one-to-many dynamic monitoring strategy, a monitoring sequence that reflects the time-division multiplexing principle is formulated based on the sampling frequency, single monitoring duration and UAV flight speed parameters. A round-trip flight path is generated between multiple monitoring areas, and the UAV is designated to observe sequentially according to the monitoring sequence.

[0034] For monitoring points that select a many-to-one collaborative monitoring strategy, a monitoring sequence that reflects the time-division multiplexing principle is formulated based on the assignment of multiple drone tasks and the sampling requirements of the monitoring points, and path planning is performed for the assigned drones based on the formulated monitoring sequence.

[0035] Preferably, the step involves estimating the minimum number of drones required to perform all monitoring tasks based on the monitoring strategies established by each independent monitoring point or group of monitoring points, and determining the available drone set based on the current drone swarm status data; specifically:

[0036] The problem of estimating the minimum number of drones is modeled as a set-coverage integer programming problem with time windows and energy constraints. Let the entire monitoring task cycle be... For monitoring points The sampling frequency is The duration of a single monitoring session is If the time limit is seconds, then the minimum number of accesses required within a task cycle is:

[0037] ;

[0038] Each access request is recorded as a task unit. ;

[0039] Let the set of candidate drones be For each candidate drone Define binary variables Indicates enabling the first Deploy drones; for each task unit and drones Define variables Indicates by the first A drone was used for this visit;

[0040] The objective function is to minimize the number of drones activated, expressed as:

[0041] ;

[0042] The constraints include task coverage constraints, single-machine time and energy constraints, and time window constraints;

[0043] The task coverage constraint is expressed as follows:

[0044] ;

[0045] That is, each task unit must be carried out by one and only one drone;

[0046] In the single-machine time and energy constraints, let the unmanned aerial vehicle (UAV) be considered. The total battery capacity is The charge / discharge efficiency is The flight and hovering power model is The round trip time to the charging station is Then, the total effective working time available within the task cycle is... Constrained by these factors; For drones Following the planned path, fly from the previous task point to the monitoring point. Execute the The required flight time for this visit should meet the following requirements:

[0047] ;

[0048] The theoretical lower bound for the above integer programming problem is obtained by solving it.

[0049] After calculating the theoretical lower bound based on the above model, and then combining the site geometry and no-fly zone to determine the transfer time... An estimate was made, and an approximate solution was obtained using a heuristic algorithm. This yielded the minimum number of drones required to simultaneously operate within the agreed task period to complete all monitoring tasks at each monitoring point, taking into account energy and no-fly zones. ;

[0050] Subsequently, based on the drone status data, the drones were screened, and those with a battery level of at least 90% and a standby status constituted a set of usable drones. Meanwhile, the drones that are charging and the rest of the drones serve as backup or standby members of the group, which are used to take over the task during subsequent dynamic scheduling.

[0051] On the other hand, this invention constructs a multi-layered, three-dimensional drone energy security network consisting of ground-based fixed charging stations, a tower crane-integrated solar-powered drone charging platform, and mobile charging vehicles with autonomous navigation capabilities. Under the unified scheduling of the monitoring center, the charging platform with the lowest overall cost is selected and the return route is planned based on the drone's current location, power status, mission urgency, platform capacity, and queuing situation. In conjunction with the mobile charging vehicles, which autonomously cruise and deploy in hotspot areas based on flight heat maps, the return distance and waiting time are significantly shortened, thereby improving the drone swarm's endurance and overall resource utilization efficiency.

[0052] Therefore, the preferred technical solution of the present invention is that the UAV energy network in step one includes:

[0053] Fixed charging stations provide a reliable parking and charging platform for drones in fixed locations;

[0054] The integrated solar-powered drone landing pad for tower cranes is installed on the tower of a high-altitude tower crane structure, providing high-altitude parking and charging services for drones.

[0055] The autonomous mobile charging vehicle can autonomously plan its path and avoid obstacles in complex construction site environments. It has a photovoltaic power generation mechanism and can provide a flexible parking and charging platform for drones on the ground.

[0056] Preferably, step five involves continuously acquiring drone cluster status data and drone energy network status data during the drone's monitoring task; specifically:

[0057] From the moment the drone swarm takes off to perform monitoring tasks, the monitoring center enters the real-time monitoring and closed-loop control phase, continuously monitoring and estimating the drones in flight, fixed charging stations, tower crane integrated solar-powered drone landing pads, and automatic cruise mobile charging vehicles.

[0058] For each drone in flight, the monitoring center receives its position, attitude, speed, and battery level data at a high frequency. The status information includes the current task number, task completion progress, and link quality.

[0059] For fixed charging stations, tower crane integrated solar-powered drone landing pads, and automated mobile charging vehicles, the monitoring center collects the number of berths occupied, available capacity, queue length, and voltage and current parameters for each platform.

[0060] In addition, for the integrated solar-powered drone landing pad for tower cranes, the monitoring center obtains the tower boom angular velocity in real time. Hook load The platform tilt angle and local wind speed information are used to calculate the safety status indicator variables of the integrated solar-powered UAV landing pad for tower cranes.

[0061] ;

[0062] in, The maximum permissible safe angular velocity threshold; The maximum allowable load limit; when When =1, the integrated solar-powered drone landing pad of the tower crane is considered a usable charging platform; when When the value is 0, it cannot be used for drone take-off and landing or charging.

[0063] For autonomous cruise mobile charging vehicles, the monitoring center estimates their state vectors using the extended Kalman filter method, based on data from the Global Navigation Satellite System, inertial measurement unit, wheel speedometer, and lidar sensors.

[0064] ;

[0065] in, The position of the mobile charging vehicle in the ground coordinate system; This refers to the vehicle's heading angle; Linear velocity; ω represents the angular velocity; this state vector will be used for subsequent optimization of the mobile charging vehicle's cruising path and stopping position.

[0066] Preferably, in step five, when the drone's battery level is detected to be below a preset threshold, based on the drone's current location, battery status, mission urgency marker, and drone energy network status data, a target refueling platform for the drone is designated, and a three-dimensional obstacle avoidance and return path for the drone to the target refueling platform is planned; specifically:

[0067] First, determine the availability of each platform based on its capacity and current available capacity; for integrated tower crane solar-powered drone landing pads, further calculations are needed. To determine whether it is safe and usable;

[0068] Subsequently, for each pair of drones and platform combinations The monitoring center estimated that from the drone Fly to the platform from the current location Required energy and time And based on the queue theory model, the expected waiting time of the platform is estimated. Meanwhile, to more accurately characterize the energy consumption and risks caused by the height difference, a height matching degree function is introduced. This is used to measure the impact of the difference between the drone's current operating altitude and the platform altitude on return-to-home energy consumption and safety.

[0069] Therefore, the definition of drones For the platform The utility function is:

[0070] ;

[0071] in, For drones Current remaining battery power; This is an estimate of the energy required to fly from the current location to platform CPj. This is an estimate of the flight time; In order to be on the platform Estimated waiting time in the queue; H align (k)(j) represents the high degree of matching score, and the larger the value, the more favorable the return flight; These represent the energy consumption weighting coefficient, the time cost weighting coefficient, and the vertical height penalty weighting coefficient, respectively.

[0072] The monitoring center monitors each drone that needs charging. Calculate its utility value across all available platforms and select the one with the highest utility value as the optimal charging platform:

[0073] ;

[0074] In order to enable the parking location of the automatic cruise mobile charging vehicle to adapt to the changes in the flight hotspot of the UAV, a flight heat map is constructed based on the historical trajectory of the UAV, and the cruise path of the automatic cruise mobile charging vehicle is dynamically adjusted accordingly.

[0075] Preferably, in step five, when a new task is added to the task waiting queue, the task is redistributed between the flying UAV and the standby UAV through redundancy capability assessment; specifically:

[0076] Task waiting queue detected If the queue is not empty, retrieve the highest priority task from the head of the queue. A redundancy capability scoring function is used to evaluate the suitability of each in-flight and standby UAV for the mission. The redundancy capability scoring function is expressed as follows:

[0077] ;

[0078] in, For drones Without disrupting the existing sampling constraints for high-priority tasks, new tasks can be... Insertion time margin score; The energy margin score for drone k, which is able to safely return to base after completing its original task and then performing the new task at the current power level; This indicates that new tasks will be inserted into the drone. The incremental marginal cost resulting from the existing path includes additional flight time and additional energy consumption; These are time weight, energy weight, and cost increment weight, respectively. To prevent extremely small positive constants with a denominator of zero; scoring value The larger the value, the higher the overall adaptability of the drone to the task.

[0079] On the other hand, this invention uses a smart safety helmet as a terminal for collecting the safety status of frontline personnel, and introduces information on worker location and fatigue level. Once the fatigue level is detected to reach the preset alarm level, the monitoring center automatically generates an active intervention task and includes it in the task waiting queue. Through the assessment of the redundancy capability of drones and the dynamic rescheduling mechanism, nearby drones or standby drones are assigned in real time to approach the workers for key monitoring, lighting and voice reminders, transforming from passive recording to active prevention and intervention.

[0080] Therefore, in step one, when constructing a computable digital twin scenario, each worker is equipped with a smart safety helmet. This smart safety helmet includes at least: a helmet structure, a main control module, a vital signs acquisition module, an environmental acquisition module, a positioning module, an activity intensity acquisition module, a communication module, and an alarm module. Among them, the vital signs acquisition module is used to collect heart rate and heart rate variability-related signals; the environmental acquisition module is used to collect thermal environment information such as skin temperature and ambient temperature and humidity; the inertial / activity intensity acquisition module is used to collect three-axis acceleration and generate activity intensity indicators; the positioning module is used to output the worker's position coordinates; and the communication module is used to report the safety helmet number, worker number, timestamp, position coordinates, and vital signs, environment, and activity intensity data to the monitoring center.

[0081] In step five, when the monitoring center enters the real-time monitoring and closed-loop control stage, it uses smart safety helmets to continuously collect the status of workers and report alarms. When the smart safety helmet detects that the worker's fatigue level reaches the preset alarm level, the monitoring center automatically generates an active intervention monitoring task based on the worker's real-time location and preset intervention strategy. The monitoring target of the active intervention monitoring task is the location of the worker who triggered the fatigue alarm. The monitoring actions include at least approaching and hovering, lighting, and voice reminders. The active intervention monitoring task is then added to the head of the task waiting queue with the highest scheduling priority.

[0082] Another objective of this invention is to provide a UAV collaborative multi-strategy dynamic monitoring system for multiple monitoring points at construction sites, used to execute the above-mentioned method, including:

[0083] The monitoring center is used to build digital twin scenarios, calculate the optimal monitoring area for monitoring points, select monitoring strategies, generate monitoring plans, and perform dynamic rescheduling.

[0084] Unmanned aerial vehicle (UAV) swarms are used to perform takeoff, cruise, hovering observation, return to base for refueling, and active intervention tasks according to the monitoring plan issued by the monitoring center.

[0085] Smart safety helmets are used to collect and report data related to the location and fatigue of workers.

[0086] Multiple types of charging and parking platforms, including at least ground-based fixed charging stations, tower crane-integrated solar-powered drone landing pads, and automated cruise mobile charging vehicles;

[0087] The monitoring center is configured to: during mission execution, perform mission release, mission relay, preemptive scheduling and path replanning based on the drone's battery level and mission queue status, and select a target charging platform among the various types of charging and parking platforms.

[0088] When the monitoring center retrieves a task from the task waiting queue, if the task at the top of the queue is the aforementioned proactive intervention monitoring task, it adopts a preemptive scheduling strategy that prioritizes personnel safety: under the premise of meeting flight safety constraints, it prioritizes selecting the drone closest to the operator to undertake the proactive intervention monitoring task. If necessary, it allows interruption of the low-priority monitoring task being executed by the drone, and the unfinished part of the interrupted task is released back into the task waiting queue. The monitoring sequence and flight path of the relevant drones are then replanned to shorten the response time of the proactive intervention task as much as possible.

[0089] During continuous monitoring, if the monitoring center detects that the drone's battery level is below a preset threshold, it will release any unfinished monitoring tasks for the drone, add these unfinished tasks to the task waiting queue, and select the optimal charging platform with the lowest overall cost for the drone based on the usage of various charging platforms, the distance from the drone's current location, and the platform type. It will then plan an obstacle avoidance flight path for the drone and assign it to the charging platform. The optimal charging platform can be a ground-based fixed charging station, a tower crane integrated solar-powered drone landing pad, or an automated mobile charging vehicle, thereby utilizing high-altitude and mobile platforms to shorten the return distance and reduce waiting time.

[0090] During continuous monitoring, if the monitoring center detects a task entering the task waiting queue, it retrieves the task from the queue and prioritizes nearby drones with redundancy capabilities to take over the task. The tasks entering the task waiting queue include not only incomplete routine monitoring tasks released due to battery warnings and newly added routine monitoring tasks, but also proactive intervention monitoring tasks triggered by smart helmet fatigue alarms. Based on the drone's current location, remaining battery power, and existing task load, the monitoring center determines whether there are nearby drones or drone assemblies capable of taking over the task. If so, the task is assigned to a drone with redundancy capabilities, and the time-division multiplexing monitoring sequence and flight path of the relevant drones are replanned. If not, a suitable backup drone is selected from the backup drone queue, the task is assigned to the backup drone, and an observation position, monitoring sequence, and flight path are planned for it, allowing it to serve as a supplementary resource to take over the execution.

[0091] During continuous monitoring, if the monitoring center detects that a certain drone has completed all its assigned tasks, it will assign the drone to the best charging platform for charging based on factors such as the usage and distance of each charging platform (including fixed charging stations, tower crane integrated solar-powered drone landing pads, and automatic cruise mobile charging vehicles) at that time, and will plan its return route to improve overall energy utilization efficiency and drone turnover efficiency.

[0092] During continuous monitoring, if the monitoring center detects that a drone has completed charging and reached a preset power threshold, it will add the drone to the standby drone queue and order it to enter standby mode, so that it can participate in dynamic scheduling as an optional resource for subsequent tasks waiting in the queue.

[0093] During continuous monitoring, if the monitoring center detects the completion of a specific monitoring task, it establishes a communication link with all drones involved in the task, collects drone monitoring videos related to the task, and extracts and merges the collected videos according to the monitoring objectives, time-division multiplexing monitoring sequences, and sampling requirements of the task. For tasks completed collaboratively by multiple drones or at different time slots, the multi-source video clips are spliced ​​and organized in chronological order to generate a complete and coherent monitoring video for archiving, retrospective analysis, and security evidence collection.

[0094] The smart safety helmet includes at least a main control module, a vital sign acquisition module, an environmental acquisition module, an activity intensity acquisition module, a positioning module, a communication module, and an alarm notification module; wherein:

[0095] The vital signs acquisition module is used to collect heart rate and heart rate variability-related signals;

[0096] The environmental data acquisition module is used to collect information on skin temperature and ambient temperature and humidity.

[0097] The activity intensity acquisition module is used to acquire triaxial acceleration and generate activity intensity indicators;

[0098] The positioning module is used to output the coordinates of the operator's position;

[0099] The communication module is used to report status information to the monitoring center, including at least the safety helmet number / worker identification, timestamp, location coordinates, vital signs data, environmental data, and activity intensity data.

[0100] The monitoring center performs a fatigue assessment based on the aforementioned status information to determine the fatigue level. The fatigue level is calculated and determined by the monitoring center based on a comprehensive fatigue score derived from the data reported by the smart safety helmet. for:

[0101] ;

[0102] in, Heart rate component scoring is based on the increase in real-time heart rate relative to resting baseline. calculate; Heart rate variability component scoring is calculated based on the RMSSD index; Scoring of heat load components based on heat load index Calculation, where The difference between skin temperature and ambient temperature; Score the activity intensity component; when When the threshold is exceeded, it is determined to be a fatigue alarm and an active intervention task is triggered.

[0103] The automatic cruise mobile charging vehicle includes a vehicle chassis and drive and braking module, an on-board positioning and obstacle avoidance perception module, a vehicle-to-cloud communication module, an on-board energy storage power supply and charging management module, and at least one drone take-off, landing, parking, and charging station set on the upper part of the vehicle body; the vehicle-to-cloud communication module is used to report the current location of the mobile charging vehicle, the number of available stations, and the estimated waiting time to the monitoring center, and to receive the stationing location or cruise route instructions issued by the monitoring center.

[0104] The integrated solar-powered drone landing pad for tower cranes includes a mounting bracket, the landing pad itself, at least one drone parking and charging station, a solar power supply and energy storage module, a platform status acquisition module, a platform safety assessment module, and a communication module; the platform status acquisition module is used to collect at least the tower boom rotation angular velocity. Hook load Safety-related parameters such as local wind speed; the platform safety determination module outputs a platform safety availability status indication based on a preset threshold, and reports the platform's available capacity and safety availability status to the monitoring center via the communication module, only when... and It outputs a safe and available status indication to allow the drone to take off, land, and charge.

[0105] Beneficial effects: Through the above-mentioned monitoring task planning and allocation, application of differentiated monitoring strategies, dynamic monitoring and scheduling during task execution, and linkage with the smart safety helmet and a multi-level charging and parking device network consisting of fixed charging stations, tower crane integrated solar-powered drone landing pads, and automatic cruise mobile charging vehicles, this invention can achieve efficient collaborative monitoring of drone swarms under multi-monitoring conditions, intelligent closed-loop management of energy replenishment, and real-time proactive safety intervention for front-line workers in complex and dynamic construction site environments, significantly improving monitoring efficiency, coverage, endurance, and system robustness. Attached Figure Description

[0106] Figure 1 This is a flowchart of a UAV collaborative multi-strategy dynamic monitoring method for multiple monitoring points on a construction site, as shown in the embodiment.

[0107] Figure 2 This is a schematic diagram of the structure of the smart safety helmet in the embodiment;

[0108] Figure 3 This is a schematic diagram of the structure of the automatic cruise mobile charging vehicle in the embodiment;

[0109] Figure 4 This is a schematic diagram of the structure of the integrated solar-powered drone landing pad for tower cranes in the embodiment. Detailed Implementation

[0110] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, embodiments of this invention, and should not be construed as limiting the invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention. In the description of this invention, it should be understood that the terminology used is for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0111] Example: Addressing the following prominent issues in construction site scenarios: First, the large number and wide distribution of monitoring points, coupled with significant dynamic changes during construction, make it difficult for traditional drone inspection methods based on fixed routes or simple "one drone per line" to simultaneously meet the sampling frequency and priority requirements of different monitoring points under limited fleet and power conditions. This easily leads to monitoring blind spots and time gaps. Second, existing scheduling schemes generally lack unified modeling of 3D obstacles, no-fly zones and height restrictions, charging platform capacity, and queuing factors. This prevents flexible selection of one-to-one or pair-to-pair scheduling based on the spatial relationships and task requirements of the monitoring points. Many-to-one strategies result in numerous repetitive flights and a high proportion of empty flights, leading to insufficient overall scheduling efficiency and system robustness. Third, the limited endurance of drones, which typically rely on a small number of fixed ground charging stations, makes it difficult to dynamically optimize configurations as construction progresses and shifts in focus. Long return paths and waiting times limit the effective operating radius and sustainability of drone swarms. Fourth, workers are prone to fatigue, heatstroke, and other physiological risks when working in high-temperature, high-intensity, and complex environments. Traditional monitoring systems lack the ability to proactively intervene in conjunction with workers' physiological states, making it difficult to promptly detect and address potential safety risks.

[0112] This embodiment provides a UAV-based collaborative multi-strategy dynamic monitoring method for multiple monitoring points on construction sites. The implementation of this method depends on the construction of a hardware platform; therefore, this embodiment combines... Figures 2-4 The hardware platform on which the method depends is described, namely, a drone-based collaborative multi-strategy dynamic monitoring system for multiple monitoring points on a construction site.

[0113] The system includes at least: a monitoring center, a drone swarm, smart safety helmets for collecting the location and fatigue status of workers, and multiple types of charging and parking platforms for recharging and parking drones; the multiple types of charging and parking platforms include at least ground-based fixed charging stations, tower crane integrated solar-powered drone landing pads, and autonomous mobile charging vehicles with autonomous mobility.

[0114] The monitoring center is used to build digital twin scenarios, calculate the optimal monitoring area for monitoring points, select monitoring strategies, generate monitoring plans, and perform dynamic rescheduling. During task execution, it performs task release, task relay, preemptive scheduling, and path replanning based on the drone's battery level and task queue status, and selects the target charging platform among multiple types of charging and parking platforms.

[0115] Smart helmets, such as Figure 2 As shown, the main body of the worker status acquisition terminal consists of a high-strength cap (a) for providing physical protection. A main control communication module (b) is integrated at the rear of the cap, containing a microprocessor, wireless communication unit, and positioning module, responsible for data processing and uploading. To achieve accurate monitoring of the worker's physiological characteristics, multiple sensor contacts are installed inside the cap liner; an environmental sensing contact (c) at the top collects the temperature and humidity of the microenvironment inside the cap to help determine the heat load; and physiological signal acquisition contacts (d) and (e) distributed on the forehead and sides preferably use flexible dry electrodes or PPG sensors, closely attached to the skin to acquire heart rate, blood oxygen, and EEG signals in real time, thereby assessing fatigue levels. Furthermore, an audible and visual alarm device (f) is installed at the front of the cap. When the fatigue level is detected to be excessive, this device immediately emits a high-frequency flashing and buzzing alert for local intervention. To ensure wearing comfort and good contact between the sensor and the skin, the cap is also equipped with a front cushioning layer g and a rear fitting layer h. These inner lining structures not only play a role in shock absorption, but also ensure that the data collection contacts d and e can still collect data stably when the worker is in motion through elastic pre-tightening force.

[0116] Automatic cruise mobile charging vehicle, such as Figure 3 As shown, this system provides flexible energy replenishment on the ground. An obstacle avoidance perception module (a) is located at the front of the vehicle, working in conjunction with onboard navigation algorithms to achieve autonomous path planning and obstacle avoidance in complex construction site environments. A high-gain communication antenna (b) is erected on one side of the vehicle to maintain a high-speed data link with the monitoring center and the drones in the air. The core functional area on the top of the vehicle consists of two parts: first, a drone take-off and landing charging station (c), which integrates an automatic centering guidance structure and a fast charging interface to support precise drone landing and recharging; second, a solar photovoltaic panel (d) located on the other side, which features an adjustable angle design to continuously replenish clean energy to the onboard battery and standby drones during vehicle stays, extending the system's field operation time.

[0117] Integrated solar-powered drone landing pad for tower cranes, such as Figure 4As shown, this embodiment utilizes a high-altitude tower crane structure to construct a power replenishment node, the foundation of which is a platform base a fixed to the tower body via supports. A drone landing pad b is located in the core area of ​​the platform, providing high-altitude parking and charging services for drones. To ensure the safety and stability of high-altitude operations, windproof railings c are installed around the platform, effectively mitigating the interference of high-altitude gusts on drone take-off and landing. Regarding communication and energy security, a long-distance communication antenna d and a directional relay antenna f are deployed at one end of the platform, respectively for wide-area signal coverage and high-bandwidth backhaul in specific areas; a large-area solar power array e is installed at the other end, enabling the platform to be energy self-sufficient. A meteorological and control workstation g is integrated on the central column of the platform, monitoring wind speed, wind direction, and vibration parameters in real time to ensure safe take-off and landing. In addition, an auxiliary monitoring and lighting arm h is installed above the platform for video verification of the landing pad status and nighttime supplementary lighting; a cable management and counterweight module i is located at the bottom, used to organize transmission cables and balance the platform's center of gravity.

[0118] With the support of the aforementioned hardware devices, the UAV collaborative multi-strategy dynamic monitoring method for multiple monitoring points on construction sites in this embodiment will be implemented in a unified digital twin scenario. The specific process is as follows: Figure 1 As shown, it includes the following steps:

[0119] By acquiring 3D maps, obstacle information, and no-fly zone information of the construction site to be monitored, and combining the monitoring requirements of all monitoring points with drone swarm status data and drone energy network status data, a computable digital twin scenario can be constructed.

[0120] S101: The monitoring center imports the latest 3D map of the construction site, no-fly zones, monitoring point information collection, monitoring requirements, the latest status data of drones, drone charging platform information, drone parking platforms, and information on operators and equipment.

[0121] In this embodiment, the construction site space is set as a large construction site with dimensions of 300m (X direction) × 500m (Y direction) × 80m (Z direction). The monitoring center first imports information on the main buildings and obstacles in the scene, no-fly zone information, monitoring point information and corresponding monitoring requirements, UAV status information, information on multiple types of charging and parking platforms, and information on operators and equipment, as shown in Tables 1-8.

[0122] Table 1. Information on main buildings and obstacles in the scene.

[0123]

[0124] In Table 1, object numbers O1 to O4 represent the main buildings and facilities on the construction site, which together constitute the obstacle set O.

[0125] Table 2 Information on Tower Cranes and Dynamic No-Fly Zones

[0126]

[0127] Table 2 shows the information on tower cranes and their corresponding dynamic no-fly zones. In Table 2, the boom rotation envelopes of tower cranes TC1 and TC2 during operation form zones numbered as follows: Together with the dynamic no-fly zone of NFZdynTC2, these two no-fly zones constitute the dynamic no-fly zone set NFZdyn.

[0128] Table 3 Static No-Fly Zone Information

[0129]

[0130] The static no-fly zone information is shown in Table 3. and Forming a set of static no-fly zones Combining Tables 2 and 3, the overall set of no-fly zones can be defined as follows:

[0131] ;

[0132] in, , .

[0133] After importing the aforementioned geometry and no-fly zone information, the monitoring center further imports monitoring point information and monitoring requirement information. The monitoring point information table is shown in Table 4.

[0134] Table 4 Monitoring Point Information Table

[0135] In Table 4, the monitoring point number uniquely identifies each monitoring target, the location coordinates give the location of the monitoring point in three-dimensional space, and the monitoring requirement number corresponds one-to-one with the monitoring requirements in Table 5.

[0136]

[0137] Table 5 Monitoring Requirements

[0138]

[0139] In Table 5, the sampling frequency indicates the number of times the monitoring task needs to be executed per unit time, the single monitoring duration is the minimum duration of each effective observation, and the monitoring weight is a value that measures the priority or importance of the task and will be used as a weighting factor in subsequent scheduling and rescheduling.

[0140] The monitoring center also needs to know the status of the drones performing the mission. This information is integrated into the drone status table, as shown in Table 6.

[0141] Table 6 UAV Status Table

[0142]

[0143] To ensure the safe, efficient, and continuous operation of the drone swarm, the monitoring center also imported information on charging and parking platforms, as shown in Table 7.

[0144] Table 7 Platform Information Table

[0145]

[0146] To extend the monitoring scope to the safety status of frontline workers, the monitoring center finally imported the smart safety helmet status table, as shown in Table 8.

[0147] Table 8 Smart Helmet Status Table

[0148]

[0149] Based on all the data in Tables 1-8, the monitoring center constructs a unified digital twin scene model. This model includes a three-dimensional geometric model. No-fly zone assembly Monitoring point set Monitoring requirements collection drone collection Charging and parking platform integration The data includes the set of workers {w} and their correspondence with smart safety helmets. This digital twin scenario will provide a unified environment for subsequent calculations, strategy selection, and path planning for all monitored areas.

[0150] Step S102: The monitoring center calculates the optimal monitoring area for each monitoring point based on the monitoring point information set and monitoring requirements.

[0151] In step S102, the monitoring center, based on the digital twin scenario, assigns each monitoring point in Table 1... (spatial coordinates are) Calculate the corresponding optimal monitoring area In this embodiment, Defined as a set of three-dimensional points that satisfy multiple constraints:

[0152] ;

[0153] in, Possible observation locations for the drone. The coordinates of the monitoring point; To constrain the observation distance, the observation distance must be within the minimum safe distance and based on the ground sampling resolution. Between the calculated maximum effective observation distances; To ensure unobstructed line of sight, a ray tracing algorithm is used to determine the set of observation lines and obstacles. No intersection; To constrain the observation perspective, the angle between the observation vector and the vertical direction and the normal vector of the measured surface is limited; To circumvent restrictions in no-fly zones and ensure that observation locations are not situated within static no-fly zones. or dynamic tower crane no-fly zone internal.

[0154] With monitoring points Taking the concrete pouring point on the top floor of building O2 as an example, the specific form of the above constraints is explained. The monitoring requirements for this point are as follows: The required feature size is approximately Detailed features. The monitoring center is... The assigned aircraft type is Its zoom camera has an equivalent focal length and pixel size First, consider the observation distance constraint. The relationship between ground sampling resolution GSD and observation distance D is defined as follows:

[0155] ;

[0156] Where D is the distance from the optical center of the drone camera to the monitoring point. spatial distance, This refers to the pixel size of the camera sensor. To ensure the length is... The goal is at least by l In this embodiment, each pixel represents a unit of data. =15, then the maximum required GSD is:

[0157] ;

[0158] Maximum effective observation distance Defined as satisfying The maximum distance, that is:

[0159] ;

[0160] From this, we can deduce... ≈40m. To ensure safe operation, this embodiment also sets a minimum safe observation distance. The observation distance constraint can be written as:

[0161] ;

[0162] in Indicates from the observation point To the monitoring point The distance.

[0163] Unobstructed view Required from arrive The lines do not intersect any obstacles. The monitoring center uses a ray tracing algorithm on the GPU to trace the line segments. The line segment polyhedrons of the 3D models of each obstacle in Table 1 are used for intersection detection. If an intersection point exists, then that point is considered an intersection point. It does not meet the no-occlusion condition. This can be formalized as:

[0164] ;

[0165] Observational perspective constraints For quality inspection tasks, an approximately vertical top-down view is required. Let's assume... point to The angle between the observation vector and the vertically downward direction is... , and the normal vector of the measured surface The included angle is This embodiment requires:

[0166] ;

[0167] in and These are used to measure the degree of view from above and the quality of observation relative to the surface normal, respectively. No-fly zone avoidance constraints. Required observation points Not located within any static or dynamic no-fly zone, can be written as:

[0168] ;

[0169] in and The results are given in Tables 2 and 3.

[0170] By discretizing the entire construction site space and simultaneously checking whether the above four types of constraints hold for each candidate point, the following can be obtained: ~ Each of their optimal monitoring areas These regions are generally non-convex three-dimensional point sets, which may consist of several disconnected sub-regions, providing a basis for subsequent observation location selection and monitoring strategy decisions.

[0171] Step S103: The monitoring center calculates a matching scheme between drones and monitoring points based on the set of monitoring point information, monitoring requirements, the monitoring area of ​​each monitoring point, and the set of available drones. The matching scheme specifies the set of monitoring drones for each monitoring point.

[0172] In step S103, the monitoring center estimates the minimum number of drones required to perform all monitoring tasks based on the monitoring point information, monitoring requirements, and the optimal monitoring area of ​​each monitoring point, and determines the set of available drones based on the status information in Table 6.

[0173] This embodiment models the minimum number of drones estimation problem as a set-coverage integer programming problem with time windows and energy constraints. Let the entire monitoring task cycle be... For monitoring points The sampling frequency is (Unit: times / hour), duration of a single monitoring session is If the number of seconds is less than the number of accesses required within a single task cycle, then the minimum number of accesses is:

[0174] ;

[0175] Each access request is recorded as a task unit. Let the set of candidate drones be... For each candidate drone Define binary variables Indicates enabling the first Deploy drones; for each task unit and drones Define variables Indicates by the first A drone was used for this visit.

[0176] The objective function is to minimize the number of drones activated.

[0177] ;

[0178] The constraints include task coverage constraints, single-machine time and energy constraints, and time window constraints. The task coverage constraints are:

[0179] ;

[0180] That is, each task unit must be carried out by one and only one drone.

[0181] In the single-machine time and energy constraints, let the unmanned aerial vehicle (UAV) be considered. The total battery capacity is The charge / discharge efficiency is The flight and hovering power model is The round trip time to the charging station is Then, the total effective working time available within the task cycle is... Constrained by these factors. For drones Following the planned path, fly from the previous task point to the monitoring point. Execute the The required flight time for this visit should meet the following requirements:

[0182] ;

[0183] The theoretical lower bound of the aforementioned integer programming problem can be obtained by solving it. In this embodiment, the monitoring center first calculates the theoretical lower bound based on the above model, and then combines the site geometry and no-fly zone to determine the transfer time. An estimate was made, and an approximate solution was obtained using heuristic algorithms (greedy algorithm and simulated annealing algorithm). The results show that to complete all monitoring tasks for each monitoring point in Table 1 within the agreed task period, considering energy and no-fly zones, at least four drones need to be used simultaneously. Subsequently, the monitoring center, in conjunction with Table 6 and the drone status table, screened the drones. Drones with a battery level of at least 90% and a standby status were selected as the usable drone set. While charging and as a backup In the initial stage of this embodiment, they are regarded as standby or reserve members of the set, and are used to take over the execution of tasks during subsequent dynamic scheduling.

[0184] Step S104: For monitoring points with high sampling frequency and remote locations or small, non-overlapping optimal monitoring areas, the monitoring center decides to adopt a one-to-one monitoring solution and assigns a dedicated drone to monitor the monitoring point within a specified time interval.

[0185] In step S104, based on the initial matching results and monitoring point characteristics of step S104, the monitoring center decides on the set of monitoring points for which a one-to-one dedicated monitoring strategy is adopted, and assigns a dedicated drone to it.

[0186] In this embodiment, the monitoring center selects one-to-one monitoring candidate points according to the following criteria: high sampling frequency (in Table 5). and The corresponding monitoring points, with high monitoring weights (weight not less than 90), are located in remote areas or their optimal monitoring areas have almost no overlap with the optimal monitoring areas of other monitoring points. After analysis, deep foundation pits... Temporary unloading area and the top floor concrete pouring point All meet the above characteristics. Based on the initial matching results, the monitoring center officially decided to adopt a one-to-one dedicated monitoring strategy, and will... Dedicated to ,Will Dedicated to ,Will Dedicated to .

[0187] Based on this, the monitoring center during the task cycle Within this system, access time series are generated for these monitoring points according to their sampling frequency. For example, ensure During the day shift, data is collected at least once every 3 minutes, with each collection lasting 45 seconds. Through a one-to-one monitoring strategy, the safety and quality monitoring of critical areas such as deep foundation pits, unloading areas, and concrete pouring can be prioritized under conditions of limited fleet and limited power.

[0188] Step S105: For monitoring points with low sampling frequency and overlapping optimal monitoring areas, the monitoring center decides to adopt a one-to-many static monitoring scheme, assigning a dedicated drone to statically monitor multiple monitoring points within a specified time interval.

[0189] In step S105, the monitoring center decides to adopt a one-to-many static monitoring strategy for the set of monitoring points with low sampling frequency and overlapping optimal monitoring areas.

[0190] material storage yard and Analysis shows that their monitoring requirements are all (Monitoring every 30 minutes, with each monitoring session lasting 60 seconds and a monitoring weight of 50), and based on the calculation results of step S102, its optimal monitoring area... and public intersection area It has a considerable spatial volume. From any feasible point in this public area, as long as constraints such as viewing angle and distance are met, effective observation of the two monitoring points can be achieved by rotating the pan-tilt unit.

[0191] Therefore, in this embodiment, the monitoring center will and Combined into a one-to-many static monitoring task group and will A dedicated drone will be assigned to this task group. In subsequent steps S109 and S112, Inside Select an optimal static hovering position, and based on... The sampling requirements are to develop a static time-division multiplexing monitoring sequence for this UAV.

[0192] Step S106: For monitoring points with low sampling frequency and no usable overlap in the best monitoring area but close proximity, the monitoring center decides to adopt a one-to-many dynamic monitoring scheme, assigning a dedicated drone to dynamically monitor multiple monitoring points within a specified time interval.

[0193] Two monitoring points for building O1 (Southwest corner curtain wall) and Analysis of (Northeast scaffolding) indicates that their monitoring requirements are... (Monitoring every 25 minutes, with each monitoring session lasting 50 seconds and a monitoring weight of 70), all locations are around the building. The layout is relatively close in space, but their respective optimal monitoring areas, after being solved in step S102, do not overlap. This type of task is not suitable for static hovering monitoring, but a single drone can dynamically travel between multiple monitoring areas, using time-division multiplexing to observe multiple points on the timeline.

[0194] Therefore, in this embodiment, the monitoring center will and Combined into a one-to-many dynamic monitoring task group Since after steps S104 and S105, the available drones are from the set. All have been assigned to other task forces, therefore, no immediate action will be taken at this stage. Instead of assigning a drone for execution, it is marked as a one-to-many dynamic monitoring task group to be assigned. In the subsequent step S110, depending on the existing task allocation, the redundancy of existing drones will be used to collaboratively complete the task group. If necessary, a backup drone will be activated during the dynamic scheduling phase.

[0195] Step S107: For monitoring points with low sampling frequency and remote locations or small, non-overlapping optimal monitoring areas, the monitoring center adopts a many-to-one monitoring scheme, assigning multiple drones with existing task assignments to monitor the monitoring point within a specified time zone.

[0196] In step S107, the monitoring center decides to adopt a many-to-one collaborative monitoring strategy for monitoring points with low sampling frequency, low monitoring weight, and remote locations or small optimal monitoring areas. This strategy does not allocate drones separately to these monitoring points, but instead utilizes drones with existing task assignments to complete relevant monitoring tasks "on the way" during their own task execution intervals.

[0197] For Table 1 and Analysis shows that the former corresponds to monitoring requirements. (Once every 15 minutes, 60 seconds per session, weight 80), the latter corresponds to (Once every 60 minutes, 30 seconds per session, weight 30), its importance in the overall task system is relatively lower than that of monitoring points such as deep foundation pits and concrete pouring. If a drone were to be configured separately for each of these monitoring points, the overall resource utilization rate would decrease significantly.

[0198] Based on the initial matching tendency and UAV in step S104 01 ~UAV 04 The path planning trend was observed by the monitoring center. On the way During the process, its path passes through The surrounding area, and in two consecutive pairs There is a certain time and energy redundancy between observations; at the same time On the way A detour is possible between it and its parking platform. Surrounding area. Therefore, in this embodiment, the monitoring center decides to... The monitoring task was handed over to In execution Completed collaboratively during breaks in the main task, The monitoring task was handed over to In execution Related paths are completed in a coordinated manner, thus respectively and A many-to-one collaborative monitoring relationship is formed.

[0199] Step S108: The monitoring center selects the best monitoring location for the assigned drone within the best monitoring area.

[0200] In step S108, the monitoring center, for monitoring points employing a one-to-one dedicated monitoring strategy, locates them in their corresponding optimal monitoring areas. Within the domain, the optimal monitoring location is determined for the assigned dedicated UAV under multi-objective conditions. With monitoring points For example, let's assume The takeoff position is In this embodiment, the monitoring center determines the following multi-objective optimization model. Optimal monitoring location:

[0201] ;

[0202] in, Indicates the drone's takeoff point The flight energy required to fly to the candidate observation point v can be obtained by time integration based on the dynamics model and power model of the UAV; Indicates the location The predicted path loss of the communication link between the UAV and the monitoring center is calculated using the free space propagation model and the shadow fading model. It is a penalty function that measures the stability of the wind field at that location, and can be calculated based on historical wind field data or CFD simulation results; These are the weights for energy consumption, communication link loss, and wind farm stability penalty, respectively.

[0203] Using numerical optimization methods, in Search for feasible points within the site and obtain drones. The best monitoring location is approximately Using the same method, the results can be calculated separately. In monitoring Optimal monitoring location at that time and UAV 03 In monitoring Optimal monitoring location at that time These optimal monitoring locations not only meet the constraints of obstacles and no-fly zones, but also take into account energy consumption, communication link quality, and local wind field stability.

[0204] Step S109: The monitoring center selects the best monitoring location for the assigned drone within the intersection of the best monitoring areas of multiple monitoring points.

[0205] In step S109, the monitoring center targets task groups that employ a one-to-many static monitoring strategy. In public areas Select an optimal static hovering point. To balance the observation quality of multiple targets with link performance, the overall observation quality function is defined as follows:

[0206] ;

[0207] in and These are normalized quality scores based on viewpoint and distance, respectively. To be based on monitoring weight The obtained normalized coefficients, These are the view quality weight, distance quality weight, and communication loss penalty weight, respectively. The monitoring center employs an improved particle swarm optimization or multi-starting-point gradient descent method. Maximize search The point, get The optimal static monitoring location is approximately (48, 88, 15). In subsequent tasks, Just stay Hovering nearby, the gimbal attitude control can be used to complete the task in time-sharing mode. and The inspection.

[0208] Step S110: The monitoring center formulates a monitoring sequence based on the specific sampling conditions and requirements of each monitoring point.

[0209] In step S110, the monitoring center targets the one-to-many dynamic monitoring task group G. dyn ={MP 05 ,MP 06 Based on parameters such as sampling frequency, single monitoring duration, and UAV speed, a monitoring sequence that reflects the time-division multiplexing principle is formulated.

[0210] In this embodiment, the final decision is made by In performing its primary task gaps Tasks, and execute them in a coordinated manner. The task. Therefore, for In step S110, the solution is obtained jointly through a unified optimization model. The monitoring center performs the following steps in step S110: The dynamic task access sequence involved is jointly optimized and solved; its collaborative monitoring sequence is further coordinated with the many-to-one task insertion strategy in step S111.

[0211] Step S111: The monitoring center formulates a monitoring sequence based on the allocation of multiple drone tasks and the sampling requirements of the monitoring point.

[0212] In step S111, the monitoring center, having determined the basic task assignment for each UAV, uniformly formulates a collaborative monitoring sequence for the many-to-one collaborative monitoring task and the one-to-many dynamic monitoring task group to be assigned.

[0213] In this embodiment, UAV is considered first. 03 Its task set is as follows:

[0214] ;

[0215] in Its primary task is to monitor the following requirements: Secondly, and This will form a one-to-many dynamic monitoring task group ; This is a many-to-one collaborative monitoring task. Let the planning period be... =30 minutes, monitoring center is the collection Each task within Planning access time series The optimization objective used in this embodiment can be written as:

[0216] ;

[0217] in, For the task The priority weights can be obtained by normalizing the monitoring weights in Table 5. For the task The sampling frequency; To conduct tasks within the planning period The Start time of the next access; for The flight energy required to perform all missions throughout the entire planning cycle; To balance the sampling interval bias and energy consumption, weighting parameters are used. Constraints include: the time interval between adjacent task visits must not be less than the shortest flight time obtained based on 3D path planning; at any given time, the UAV can only perform one monitoring task; for each task... The actual time interval between two adjacent visits should be equal to its ideal sampling period. Keep it as consistent as possible, and the deviation of this time interval must not exceed the preset allowable deviation.

[0218] By solving the above optimization problem, we can obtain a set of access time examples, such as those accessed at minutes 0, 10, and 20. Each visit lasts approximately 90 seconds; visits are made at the 4th and 19th minutes. Each visit lasts 50 seconds; visit at the 14th minute. Stop for 50 seconds; visit at the 24th minute. Hold for 60 seconds. This ensures both... The high-frequency rigid sampling requirement is met, and reasonable insertions are made on the time axis. , and It achieves a balance between one-to-many dynamic monitoring and many-to-one collaborative monitoring.

[0219] for This many-to-one collaborative monitoring point, in this embodiment, is... Responsible for completing this task along the way. This is known to the monitoring center. The sampling frequency is once every 60 minutes, corresponding to the ideal access time point sequence. In determining round trip After determining the basic path and task sequence, the monitoring center calculates the set of gaps that can be inserted on its timeline. and will An ideal access time window over a relatively long period is discretized into several candidate access times. Then, by solving a small-scale bipartite graph matching problem, each candidate access time is assigned to... Given an insertable time gap, the requirements are:

[0220] ;

[0221] in for Maximum allowable sampling time deviation. Ultimately, the result obtained in this embodiment is: at the 30th minute... Monitor along the way In the 90th minute by The second visit can be completed along the way, thus satisfying the requirement without increasing the fleet size. Sampling requirements.

[0222] Through the above sequence design, the one-to-many dynamic monitoring group G, which has not yet been fully developed in step S111, is now... dyn The access time series was thus established.

[0223] Step S112: The monitoring center formulates a monitoring sequence based on the sampling requirements of each monitoring point.

[0224] In step S112, the monitoring center targets the one-to-many static monitoring task group G. stat ={MP 03 ,MP 04}, in the already determined Static hovering position is Under the premise of this, a static time-division multiplexing monitoring sequence is formulated based on the sampling frequency and single monitoring duration of each monitoring point.

[0225] Assume the task The sampling frequency is The duration of a single monitoring session is For monitoring points The access time series is The cycle length of a static time-division multiplexing round is In this case, the sampling interval should meet the following requirements:

[0226] ;

[0227] In this embodiment, corresponding The requirements are sampling every 30 minutes and a single monitoring duration of 60 seconds, therefore, we can take... Minutes. For example, it can be within each 30-minute cycle. The sequence is planned as follows: At the 5th minute, the gimbal points to... The observation continued for 60 seconds; at the 6.5-minute mark, the gimbal turned. Observe continuously for 60 seconds; observe again at the 35th minute. Repeat the above steps and observe again at 36.5 minutes. This process repeats cyclically. Through appropriate phase design, the overlap of observation times between two monitoring points can be avoided, thus strictly meeting the requirements. Sampling time requirements.

[0228] Step S113: The monitoring center plans the path for the assigned drone based on the established monitoring sequence and the specific location of each monitoring point.

[0229] In step S113, the monitoring center performs single-machine path planning for each assigned UAV based on the previously established monitoring sequence and optimal observation position. The goal of path planning is to generate a continuous, wingless flight path in the 3D digital twin scene that satisfies obstacle and no-fly zone constraints.

[0230] by For example, its task path can be simplified to flying from the starting position to the static hovering point. It performs periodic static monitoring tasks and returns to the charging or parking platform after the task cycle is completed. In this embodiment, the monitoring center is in the state space. The above adopts a hybrid approach The algorithm performs path search, where The coordinates of the drone's position. For each search node, the heading angle is used. Define the cost function:

[0231] ;

[0232] in From the starting point to the node The actual cumulative energy consumption can be estimated based on the UAV's flight distance and its power model; For the node Heuristic estimates of the target location can be obtained using a distance field based on obstacle distribution. (Hybrid) The search generates a smooth path from the current starting position to the target observation position or charging platform, while ensuring that the kinematic constraints of the UAV (maximum climb angle, minimum turning radius) are met.

[0233] Similarly, the monitoring center... , , The drones also perform individual path planning to ensure that they do not cross the obstacles in Table 1 or the no-fly zones in Tables 2 and 3 when visiting the best observation positions of their respective monitoring points.

[0234] Step S114: The monitoring center performs path planning for the assigned drones based on the established monitoring sequence.

[0235] In step S114, based on the single-machine path planning, the monitoring center performs joint optimization of the paths of multiple drones for the many-to-one collaborative monitoring scenario, ensuring that they do not conflict with each other in time and space, and forming a feasible multi-agent path planning scheme.

[0236] For a swarm of drones participating in many-to-one collaborative monitoring, the following constraint must be satisfied: no drone at any given time and position falls into the obstacle set. or no-fly zone Any two drones and At any given moment The distance should not be less than the preset no-collision safety distance. ,Right now:

[0237] ;

[0238] in For drones at all times The three-dimensional position coordinates; at the same time, each drone needs to access the corresponding mission's access time. a certain time window before and after The system arrives at the designated observation point within the specified time to ensure that the monitoring sequence planned in step S112 is executed.

[0239] In this embodiment, the monitoring center employs a combination of priority planning and conflict-based search. First, based on individual drone paths, the paths of high-priority drones are fixed sequentially according to their priority. Then, the paths of low-priority drones are corrected. When a path conflict occurs (i.e., when the distance between two drones is less than a certain value at a given moment), the conflict is resolved. Conflicts are eliminated by inserting waiting actions on the timeline or by performing local path replanning within a limited space. After several iterations, a set of multi-machine cooperative paths that satisfy all constraints is finally obtained.

[0240] Step S115: The monitoring center will distribute the established monitoring plan to the assigned drones.

[0241] In step S115, the monitoring center will provide... The generated complete monitoring solution is encapsulated as a task description data structure and distributed to the corresponding drones via wireless communication links.

[0242] Each task description data structure includes: monitoring strategy type (e.g., one-to-one, one-to-many static, one-to-many dynamic, many-to-one collaborative), the set of monitoring points to be accessed and their optimal observation locations, the access time series for each monitoring point, the planned 3D flight path sequence, applicable charging and return-to-home strategies, and relevant safety parameters. The monitoring center establishes a communication connection with each UAV via an encrypted 4G / 5G or dedicated industrial wireless network, sending the task package to the onboard task management module of each UAV. The UAV performs semantic parsing of the task package locally and performs a secondary verification of the path's dynamic feasibility and safety constraints based on its own flight control parameters. Only when the verification passes does the UAV enter a takeoff-ready state, ready to execute subsequent monitoring tasks.

[0243] Step S116: The monitoring center continuously monitors the task execution status until all tasks are completed.

[0244] In step S116, starting from the takeoff of the drone cluster, the monitoring center enters the real-time monitoring and closed-loop control stage, continuously monitoring and estimating the status of the drones in flight, charging and parking platforms, mobile charging vehicles, and operators.

[0245] For each drone in flight, the monitoring center receives its position, attitude, speed, and battery level data at a high frequency. Status information such as the current task number, task completion progress, and link quality. For fixed charging stations... and The monitoring center collects parameters such as the number of berths occupied, available capacity, queue length, voltage, and current. This applies to integrated solar-powered drone landing pads for tower cranes. The monitoring center obtains the tower boom angular velocity in real time. Hook load Information such as platform tilt angle and local wind speed is used to calculate the platform safety status indicator variables:

[0246] ;

[0247] in The angular velocity of the tower arm rotation; The maximum permissible safe angular velocity threshold; This represents the current hook load. This represents the maximum allowable load limit. When =1, It can be considered a usable charging platform; when When the value is 0, it cannot be used for drone take-off and landing or charging.

[0248] For mobile charging vehicles The monitoring center estimates the state vector of the vehicle based on sensor data from the Global Navigation Satellite System, inertial measurement unit, wheel speedometer, and lidar, using the extended Kalman filter method.

[0249] ;

[0250] in The position of the mobile charging vehicle in the ground coordinate system; This refers to the vehicle's heading angle; Linear velocity; ω represents the angular velocity. This state vector will be used for subsequent optimization of the mobile charging vehicle's cruising path and stopping position.

[0251] Regarding worker fatigue monitoring, the monitoring center continuously receives physiological and environmental data from smart safety helmets and executes fatigue assessment algorithms at fixed time intervals. This includes HR, RMSSD, and [other parameters not specified in the original text]. , and Depend on Figure 2 The smart safety helmet shown uses a vital sign acquisition module, an environmental acquisition module, and an inertial acquisition module to acquire and report location data. The output is from the positioning module and corresponds to the site coordinate system. In this embodiment, the fatigue assessment uses a length of... =A 5-minute sliding time window. For each worker wearing a smart safety helmet. The monitoring center calculates the following indicator within each time window: average heart rate within the time window. Compared with the worker's resting baseline heart rate The difference in heart rate relative elevation index Heart rate variability index RMSSD; Skin temperature difference with ambient temperature ,in For skin temperature, The ambient temperature; the average value of the acceleration modulus within the time window. .

[0252] Based on the above indicators, the monitoring center defines fatigue scores with different components. Heart rate component score:

[0253] ;

[0254] in and These are preset lower and upper threshold values ​​used to normalize the relative increase in heart rate to a range of 0 to 1. Heart rate variability component scoring:

[0255] ;

[0256] in and These represent the lower and upper thresholds of the heart rate variability index. To comprehensively consider the effects of thermal environment and increased heart rate, a simplified heat load index is introduced:

[0257] ;

[0258] in It is a linear combination coefficient that combines ambient temperature, relative increase in heart rate, and the temperature difference between skin and the environment into a heat load. Through the analysis of... Piecewise linear normalization yields the heat load component scores. Meanwhile, based on the average acceleration modulus... The magnitude of the value is mapped to the activity intensity component score. The score was also normalized to the 0-1 range using a piecewise linear method. Finally, the comprehensive fatigue score was defined as:

[0259] ;

[0260] in The weighting coefficients for each component are: weight for increased heart rate, weight for decreased heart rate variability, weight for thermal load, and weight for activity intensity, satisfying the following:

[0261] ;

[0262] Based on the comprehensive fatigue score, the worker's condition is divided into four levels: when A value <0.25 is considered normal; when 0.25 ≤ A value <0.5 indicates mild fatigue; when 0.5 ≤ A value <0.75 indicates moderate fatigue; when A value ≥0.75 indicates severe fatigue. When the threshold of 0.75 is first crossed, the smart safety helmet immediately reports the worker's three-dimensional position and fatigue level to the monitoring center, which uses this as a trigger signal for the generation of proactive intervention tasks in subsequent steps.

[0263] During the aforementioned continuous monitoring process, this embodiment will exhibit two key events 20 minutes after task execution: one is Zhang Gong's overall fatigue score. A jump from a normal value to a value no lower than 0.75 triggers a severe fatigue alarm; secondly, monitoring tasks are performed above deep foundation pits. When the battery level drops to 29.8%, below the preset 30% safety threshold, a battery warning is triggered. These two types of events will respectively activate the task rescheduling and proactive intervention processes in the corresponding steps S119 to S131.

[0264] Step S117: The monitoring center detects that a new task has entered the task waiting queue.

[0265] In step S117, when the monitoring center detects a new task that needs to be scheduled, regardless of whether the task is a newly added routine monitoring task, an unfinished task released due to a power warning, or an active intervention task triggered by a worker fatigue alarm, it will be encapsulated into a unified task description structure and inserted into the task waiting queue. .

[0266] In the timeline of this embodiment, when step S121 detects that staff fatigue exceeds the threshold, the monitoring center generates an active intervention task and adds it to the queue in step S126; subsequently, in step S117, a severe fatigue alarm is detected when a new task is added to the queue, and the monitoring center immediately generates an active safety intervention task. This task is based on Zhang Gong's current location. The target area is (90, 120, 30). The task involves hovering over it within a safe area, activating a searchlight, and repeatedly broadcasting safety reminders and rest prompts via loudspeaker. This task is assigned the highest scheduling priority, and its description includes fields such as target location, task type, priority weight, and expected response time. Subsequently, the monitoring center will... Insert task waiting queue The leader of the team.

[0267] Almost simultaneously, with the occurrence of the power shortage warning, the monitoring center will, in subsequent step S124, […]. Unfinished and The remaining monitoring tasks are packaged into task release items, and these task items are inserted according to their original or elevated priorities. The task waiting queue is sorted according to a combination of task priority weight and earliest deadline, ensuring that the most important and urgent tasks are processed first when retrieving tasks from the queue.

[0268] Step S118: The monitoring center detects that the drone has finished charging.

[0269] In step S118, the monitoring center monitors the status of drones charging on each charging platform in real time. When the charging process of a drone meets the preset termination conditions, the monitoring center triggers a "charging complete" event.

[0270] In this embodiment, On the platform The device was charged, and the monitoring center detected its battery level around the 25th minute of the task. It has reached approximately 95%, and the charging current has dropped to the termination threshold. The following conditions are met for charging completion as defined in this embodiment. If the drone is docked on the tower crane charging platform... Then the contact resistance also needs to be verified. Not greater than the threshold This confirms that the contact is safe and reliable. Once all conditions are met, the platform controller reports to the monitoring center. The charging complete event will be used for subsequent... Prepare to include them in the standby drone queue.

[0271] Step S119: The monitoring center detected that the drone's battery level was below the threshold.

[0272] In step S119, the monitoring center uses the drone's battery information and energy consumption model to detect and determine the event that "the drone's battery is below the threshold".

[0273] Let the total battery capacity of drone k be... The safe buffer energy is The current battery percentage is B. k(t) The remaining available energy is:

[0274] ;

[0275] Based on the current task list and path planning, the monitoring center predicts the energy required from the current moment until the remaining tasks are completed and the user returns to any available charging platform, denoted as . When the following conditions are met:

[0276] ;

[0277] This indicates that after deducting the safety buffer energy, the remaining available energy is insufficient to support the drone in completing its current mission and returning to base under safe conditions. At this point, the monitoring center determines that the drone... Trigger an event indicating that the battery level is below a threshold.

[0278] In this embodiment, The battery level gradually decreased from 98% at takeoff, dropping to 29.8% after 20 minutes of mission execution, below the set 30% safety threshold. According to the aforementioned energy inequality, of Lower than expected Therefore, a low battery threshold event is triggered. This event will drive S124 to release its unfinished task, and S128 to select the best charging platform and plan the return route.

[0279] Step S120: The monitoring center detects that a task has been completed.

[0280] In step S120, when the monitoring center detects that a specific monitoring task unit has been completed as required, a task completion event is triggered. A task completion event records the corresponding monitoring point number, the set of executing drone numbers, and the start and end times of the task execution.

[0281] In this embodiment, In executing its joint task set J (03) At that time, it will be within a certain time window. (Tower crane TC1 base) performs a 60-second inspection. The monitoring center... The reported task execution log confirmed that the patrol task was completed within the specified time window and duration, and therefore the task was... The patrol is marked as mission completion and triggers event S125, which is used for subsequent video data collection in S125 and video processing and compositing in S139.

[0282] Step S121: The monitoring center detected that some staff members' fatigue levels exceeded the threshold.

[0283] In step S121, during the continuous monitoring cycle of step S116, the monitoring center monitors each worker wearing a smart safety helmet. According to the preset time window Fatigue assessments are performed periodically, and a comprehensive fatigue score is calculated based on vital signs, thermal environment, and activity intensity data from the smart helmet. The monitoring center calculates the fatigue level and its corresponding fatigue level, then writes it into the worker's status table, and simultaneously records the timestamp 't' corresponding to the alarm determination and the worker's real-time location. And key component indicators used for traceability.

[0284] In this embodiment, the severe fatigue alarm threshold is set to =0.75. The monitoring center monitors the workers. The alarm triggering logic can be formalized as follows: if the operator is in a non-alarm state and meets the following conditions...

[0285] ;

[0286] If the data validity and de-jitter conditions are met, the monitoring center determines that the worker's fatigue value exceeds the threshold and generates a fatigue alarm event. The event must include at least the following fields: safety helmet number, worker number, alarm level, trigger timestamp, 3D location, confidence level, data quality identifier, and suggested intervention strategy type. Subsequently, the monitoring center will... As a trigger signal, step S126 is initiated to automatically generate an active intervention task and include it in a unified scheduling queue, thereby enabling the system to respond quickly and coordinate resources after detecting personnel risks.

[0287] Step S122: The monitoring center retrieves tasks from the task waiting queue and prioritizes nearby drones with redundancy capabilities to undertake the tasks.

[0288] In step S122, when the monitoring center detects the task waiting queue... If the queue is not empty, retrieve the highest priority task from the head of the queue. Furthermore, a quantitative assessment was conducted on the redundancy capabilities of each candidate drone in undertaking the mission.

[0289] In this embodiment, the task at the head of the queue is an active intervention task triggered by a severe fatigue alarm. The monitoring center needs to assess the suitability of each in-flight drone, and, if necessary, some backup drones, for the mission, using a redundancy capability scoring function:

[0290] ;

[0291] in, For drones Without disrupting the existing sampling constraints for high-priority tasks, new tasks can be... Insertion time margin score; The energy margin score for drone k, which is able to safely return to base after completing its original task and then performing the new task at the current power level; This indicates that new tasks will be inserted into the drone. The incremental marginal cost resulting from the existing path includes additional flight time and additional energy consumption; These are time weight, energy weight, and cost increment weight, respectively. To prevent extremely small positive constants with a denominator of zero. (Score value) The larger the value, the higher the overall adaptability of the drone to the task.

[0292] In the scenario of this embodiment, due to Currently in a static hovering state, it is close to the alarm point and has an existing task. and As a low-frequency task, therefore and The rating is high, and it is currently hovering. Fly to Zhang Gong's location and return Relatively small, the final calculation yields of The value is significantly higher than and Based on the scoring results, the monitoring center will... As a preferred candidate successor, it provides a basis for decision-making in subsequent steps S128 and S131.

[0293] Step S123: The monitoring center adds the drone to the standby drone queue and commands the drone to enter standby mode.

[0294] In step S123, the monitoring center updates the status of the drones that were determined to have completed charging in step S118, adds them to the standby drone queue, and puts them into standby mode.

[0295] In this embodiment, The set charging completion conditions were met approximately 25 minutes into the task. The monitoring center first updated its status from charging to standby, and then... Add to the standby drone queue Meanwhile, the monitoring center sent... The command to remain in standby mode was issued, requiring it to remain on the charging platform. The system can stop or hover in a low-power mode and periodically report heartbeat and critical status information. This way, if at some point S127 determines that none of the currently flying drones have sufficient redundancy to take on a high-priority task, it can preferentially select from the backup queue. Wait for the fully charged drone to carry out a new mission.

[0296] Step S124: The monitoring center detects that the drone has not completed its mission and adds it to the mission waiting queue.

[0297] In step S124, after step S119 determines that the battery level of a certain drone is below the threshold, the monitoring center immediately removes the unfinished monitoring tasks from the drone's task list and adds them to the task waiting queue as new task items. .

[0298] For this embodiment In addition, its task list still includes several unfinished deep foundation pits. Periodic monitoring tasks and (Curtain wall) related collaborative monitoring tasks. The monitoring center detects... Once the battery level drops below the safety threshold, these unfinished tasks will be removed from the system. Removed from the schedule and packaged into one or more new task units T. rel And maintain or appropriately increase its original priority. For example, for deep foundation pits These types of tasks are of high importance and should be given high weight during rescheduling. Subsequently, the monitoring center inserts these task units into the task waiting queue. The system is designed to be carried out by other drones in subsequent steps S123, S128, S131 or S132, thereby avoiding the interruption of critical point monitoring due to the power problem of a single drone.

[0299] Step S125: The monitoring center establishes a communication connection with all drones participating in the mission and collects drone monitoring videos.

[0300] In step S125, the monitoring center establishes or maintains a high-bandwidth communication link with all drones participating in the task execution for the monitoring task corresponding to the task completion event in step S120, and instructs them to upload relevant monitoring video data.

[0301] In this embodiment, Taking the inspection task as an example, this task is carried out by It is executed sequentially within its joint operation sequence. When the task is completed within the predetermined time (60 seconds), the monitoring center determines, based on the task execution log, that the ensemble of drones participating in the task consists of only [number missing]. Subsequently, the monitoring center... Send a video upload request, requesting it to retrieve the video from the local cache. The raw video stream (including visible light and infrared channels) recorded during the mission start and end time interval, along with timestamps, UAV attitude information, and camera intrinsic and extrinsic parameter data, is uploaded to the video processing server at the monitoring center. While receiving the video data, the monitoring center records the metadata of each video segment, providing temporal and spatial alignment for subsequent steps S129, including video extraction, registration, image stabilization, and compositing.

[0302] Step S126: The monitoring center immediately adds a top-weighted task to the task list, which will assign a drone to the worker's location to issue an early warning.

[0303] In step S126, when step S121 determines that a fatigue alarm event has occurred... Subsequently, the monitoring center automatically generates a proactive intervention task based on the preset proactive security intervention strategy template. And insert it into the task waiting queue with the highest scheduling priority. The team leader. The monitoring target for proactive intervention tasks is the real-time location of the personnel receiving the alarm. or its short-term predicted location The task actions include at least: approaching and hovering near the target area while meeting safety constraints, turning on the searchlight and supplementing the lighting, playing rest and evacuation reminders in a loop through the loudspeaker, and conducting a short-term key video check of the target's surroundings to assist the safety officer in confirming the on-site status.

[0304] To ensure that proactive intervention is both close enough to provide alerts without impacting construction and personnel safety, the monitoring center preferentially calculates a set of feasible intervention observation spaces for this task within a digital twin scenario, and selects the optimal intervention location within it. The feasible area for worker intervention can be defined as:

[0305] ;

[0306] in, To constrain the distance for personnel intervention, To ensure unobstructed visual constraints and guarantee that the drone can maintain an effective line of sight to the personnel in this location; To circumvent restrictions in no-fly zones; Constraints are in place for personnel safety and construction safety. Within the monitoring center, the optimal intervention location can be selected using a multi-objective criterion primarily based on minimizing response time. :

[0307] ;

[0308] in To estimate the response time, To predict the increase in energy consumption, Penalties for risks of conflict involving personnel, equipment, and airspace; These are priority levels: response time, energy consumption increment, and risk penalty.

[0309] At the task data structure level It should include at least: task type, target personnel identification, and target location. Recommended intervention observation points Expected latest response time Minimum intervention duration The system includes a preemptible flag and a set of actions to be executed. After a task is generated, the monitoring center adds it to the queue and simultaneously updates the operator's status to an alert state. If the same operator repeatedly triggers alarms within a short period, the monitoring center preferentially adopts a task merging and target location update strategy to avoid queue congestion and ensure the timeliness and stability of the intervention response. Subsequently, the process enters the queue scheduling logic in steps S117 and S122, where a suitable UAV is selected to execute the proactive intervention task by a redundancy capacity assessment and preemptive scheduling mechanism.

[0310] Step S127: Are there any drones or drone assemblies nearby that can undertake this task?

[0311] In step S127, the monitoring center, based on the redundancy capability score obtained in step S122, assesses the current lead task. Make a Boolean decision on whether the in-flight drone should undertake the task.

[0312] The determination rule is: if there exists a certain drone set The maximum redundancy score must be no less than a preset threshold. If inserting the task into the corresponding drone's task sequence does not cause any existing high-priority task to violate its sampling frequency and deadline constraints, then it is considered that there is a drone or drone set nearby that can take over the task. If the result is yes, the program proceeds to step S130; otherwise, if the result is no, the program proceeds to step S131 and selects a taker from the backup drones.

[0313] In this embodiment, for active intervention tasks ,because Redundancy rating Much higher than the threshold And insert the task into its original or Static monitoring sequences will not be destroyed Due to the low-frequency sampling requirement, S127 returns "yes" to the task, and the task will be handled by [the relevant entity] in step S130. Continuing. And once this task is completed, the new first task in the task waiting queue may be... Released deep foundation pit The subsequent monitoring tasks. A reassessment of the redundancy capabilities of all flying drones revealed that, under current energy and time constraints, no flying drone could take over without affecting existing high-priority tasks. Due to the monitoring requirements, S127 returns no to this situation, and the task will be handed over to a backup drone in step S131.

[0314] Step S128: The monitoring center selects the best charging platform for the drone based on the usage and distance of the charging platform, plans its path, and assigns it to charge.

[0315] In step S128, the monitoring center selects the best charging platform with the lowest overall cost from all charging and parking platforms for drones that need to return to base to recharge due to low battery levels or completion of all tasks, and plans an obstacle avoidance return path for the drone.

[0316] In this embodiment, the candidate platform set is as follows: The monitoring center first determines whether each platform is available based on its capacity and current available capacity as shown in Table 7; for LPs... 01 It is also necessary to calculate based on the results in step S116. Determine its safety and usability. Then, for each pair of drones and platform combinations... The monitoring center estimated that from the drone Fly to the platform from the current location Required energy and time And based on the queue theory model, the expected waiting time of the platform is estimated. Meanwhile, to more accurately characterize the energy consumption and risks caused by the height difference, a height matching degree function is introduced. This is used to measure the impact of the difference between the drone's current operating altitude and the platform altitude on return-to-home energy consumption and safety.

[0317] Based on the above quantities, the monitoring center defines drones For the platform The utility function is:

[0318] ;

[0319] in, For drones Current remaining battery power; This is an estimate of the energy required to fly from the current location to platform CPj. This is an estimate of the flight time; In order to be on the platform Estimated waiting time in the queue; H align (k)(j) represents the high degree of matching score, and the larger the value, the more favorable the return flight; These represent the energy consumption weighting coefficient, used to adjust the system's sensitivity to the drone's flight energy consumption; the time cost weighting coefficient, used to adjust the system's sensitivity to flight time; and the vertical altitude penalty weighting coefficient, which aims to reduce the frequency of unnecessary large-scale altitude layer switching by the drone.

[0320] The monitoring center monitors each drone that needs charging. Calculate its utility value across all available platforms and select the one with the highest utility value as the optimal charging platform:

[0321] ;

[0322] In this embodiment, when In deep foundation pit When the vehicle in the air needed to return to base due to insufficient power, the mobile charging vehicle... Based on the hotspot distribution, the UAV has moved to the high-frequency activity area to the north (approximately at coordinates (140, 400, 1)) and is currently not in a queue. Considering factors such as flight distance, waiting time, and altitude matching, the UAV... 01 For MCV 01 The utility value is the highest, therefore the monitoring center chooses... As The optimal charging platform. Then, a 3D obstacle avoidance path planning algorithm is invoked to provide [the necessary path] without traversing obstacles or no-fly zones. Generate a flight from the current monitoring location Find the optimal path and issue the execution command.

[0323] In order to make mobile charging vehicles The stationary position can adapt to changes in the drone's flight heat map. The monitoring center also constructs a flight heat map based on the drone's historical trajectory and dynamically adjusts the location accordingly. The patrol path. In this embodiment, the monitoring center divides the construction site into grids. For each location point in the historical flight trajectory Apply time decay weights ,in For the current moment, This represents the time decay coefficient. The value for each grid cell is obtained through kernel density estimation. Popularity value:

[0324] ;

[0325] In the formula The kernel function bandwidth determines the smoothness of the heatmap. Choose one that satisfies... The grid points form high-frequency activity areas, and several hotspot centers are obtained through clustering algorithms. Furthermore, for each candidate residence site... Define the residency utility function:

[0326] ;

[0327] in The heat value of the hotspot center; Candidate residence sites With Hotspot Center The ground distance between them; The environmental cost of the hotspot area (terrain ruggedness, construction interference); The cost of potential future congestion or conflict near this outpost; The monitoring center assigns corresponding weights to each candidate location. It then evaluates the utility of each location and selects the area with the highest utility as the MCV (Mean Value). 01 The mobile charging vehicle identifies the navigation target and invokes a ground-based path planning algorithm to generate a cruising path and new stopping point locations. Through this heatmap-driven strategy, the mobile charging vehicle can dynamically follow the drone's flight hotspots, improving overall charging efficiency.

[0328] Step S129: The monitoring center extracts and merges the collected videos according to the monitoring task, monitoring sequence and sampling requirements.

[0329] In step S129, the monitoring center performs fine processing on the multi-source monitoring videos collected in step S125, extracting, correcting, and synthesizing them into structured standard monitoring videos to meet the requirements of monitoring recording and security evidence collection.

[0330] In this embodiment, the monitoring center first extracts the corresponding time period from the original video stream uploaded by the drone according to the monitoring sequence and sampling requirements of the corresponding task. Within the effective monitoring footage, segments including takeoff and landing times, flight transitions, and other irrelevant segments are removed. Then, based on the UAV's attitude information and camera intrinsic parameters, geometric distortion correction and image stabilization are performed on the video to reduce image shift caused by aircraft shake and tilt, ensuring image stability and geometric accuracy. For tasks requiring multiple UAVs or coordinated operation across different time periods, such as... and The monitoring center then stitches together valid segments uploaded by different drones at different time slots in chronological order, and performs brightness and color matching as needed, to generate a standard monitoring video that is continuous in time and has uniform image quality. Finally, this video, along with metadata such as task number and monitoring point number, is archived for subsequent project quality acceptance, accident review, and security evidence collection.

[0331] Step S130: The monitoring center assigns tasks to drones with redundancy capabilities and replans the relevant drone monitoring sequences and path planning.

[0332] In step S130, when step S127 determines that there is an in-flying UAV or UAV group that can take over the current task of the lead UAV, the monitoring center selects the UAV with the highest redundancy score as the takeover and performs local replanning of its existing monitoring sequence and flight path.

[0333] In this embodiment, for the active intervention task triggered by the severe fatigue alarm... The monitoring center confirmed this in S123 and S128. Therefore, in S130, the monitoring center assigns the task to the most suitable recipient. And temporarily suspend its static monitoring task group. ,Right now and Regular inspections were conducted. Subsequently, in Current static hover point With worker Zhang Gong's position Local path replanning is performed between (90, 120, 30) and the return position after intervention to construct a new flight path: The route, under the constraints of a 3D raster map and no-fly zones, is determined by a hybrid approach. Alternatively, a local graph search algorithm can be used for calculation, with the goal of minimizing flight time and energy consumption while ensuring safety, thereby shortening the response time from fatigue alarm to the drone arriving at the scene as much as possible.

[0334] After completing the intervention task, the monitoring center, based on and The sampling requirements and time window deviation tolerance, for The existing static monitoring sequence will be appropriately shifted and adjusted to ensure that it still meets the requirements throughout the entire monitoring cycle. The specified sampling frequency and single-session duration requirements are followed. In this way, this embodiment achieves rapid and proactive safety intervention for fatigue risks among frontline workers without sacrificing the monitoring quality of high-risk points such as deep foundation pits, unloading areas, and concrete pouring.

[0335] Step S131: The monitoring center assigns the task to the backup drone and plans the observation position, monitoring sequence, and path planning of the backup drone.

[0336] In step S131, when step S127 determines that none of the currently flying drones have sufficient redundancy to take over the current lead drone task, the monitoring center selects from the backup drone queue. The most suitable drone was selected as the carrier, and a complete monitoring scheme was redesigned for the drone.

[0337] In this embodiment, when The deep foundation pit was released due to insufficient power. and curtain wall After the monitoring task was not completed, the monitoring center determined through S128 that if any flying drone took over the task, it would cause existing high-priority tasks to violate their sampling constraints. Therefore, it was decided to activate the backup drone in S131. The monitoring center determines the status based on the battery level of the backup drone. (Battery level is 100%), current location, and distribution of mission objectives. Selected as a drone receiving provider. Due to... and The optimal monitoring area and optimal observation location have been calculated in steps S102 to S109. The monitoring center directly reuses these results. and Combined into a one-to-many dynamic monitoring task group, by They shuttle between two monitoring points to perform monitoring.

[0338] Subsequently, the monitoring center Developing new monitoring sequences can be achieved by addressing the simplified traveling salesman problem or its variant with time window constraints, to monitor access. and The order and interval are optimized to ensure that it meets the requirements. and While meeting the requirements for sampling frequency and single monitoring duration, efforts should be made to minimize flight energy consumption. Furthermore, under three-dimensional no-fly zones, [the following is necessary:]... The plan starts from the base and visits in sequence. and The mission plan involves determining the optimal observation location and, upon completion, returning to a designated optimal charging platform via a closed-loop flight path. Finally, this new mission plan is distributed to [the relevant authority / organization] through step S115. He was ordered to take off from the base to relieve [him / her]. All unfinished tasks were completed, enabling seamless integration and dynamic resource optimization at the system level for monitoring deep foundation pits and curtain walls during the drone replacement process.

[0339] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0340] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for collaborative multi-strategy dynamic monitoring of multiple monitoring points on a construction site using unmanned aerial vehicles (UAVs), characterized in that: Includes the following steps: Step 1: Obtain a 3D map, obstacle information, and no-fly zone information of the construction site to be monitored. Combine this with the monitoring requirements of all monitoring points, as well as the drone swarm status data and drone energy network status data, to construct a computable digital twin scenario. Step 2: Based on the monitoring requirements of each monitoring point, calculate the three-dimensional feasible observation space of each monitoring point that satisfies the observation distance constraint, the line of sight unobstructed constraint, and the no-fly zone avoidance constraint, and use it as the optimal monitoring area for that monitoring point. Step 3: Based on the monitoring requirements and optimal monitoring areas of each monitoring point, divide all monitoring points into independent monitoring points consisting of a single monitoring point, and monitoring point groups formed by combining multiple monitoring points. And develop monitoring strategies for each independent monitoring point and monitoring point group; Step 4: Based on the current drone cluster status data, determine the set of available drones that can perform monitoring tasks, and generate a matching scheme between drones and each independent monitoring point or monitoring point group according to the established monitoring strategy. Plan the monitoring sequence and flight path of each drone, and assign the corresponding drone to perform the monitoring task. Step 5: During the monitoring mission, continuously acquire drone cluster status data and drone energy network status data; when a drone's battery level is detected to be below a preset threshold, release the drone's unfinished monitoring mission and add it to the mission waiting queue. Through redundancy assessment, redistribute the mission between the flying drone and the standby drone, and replan the monitoring sequence and flight path of the relevant drones to achieve mission relay; at the same time, based on the current location, battery status, and mission urgency level of the low-battery drone, as well as the drone's energy network status data, designate the target refueling platform for the drone, and plan a three-dimensional obstacle avoidance return path for the drone to the target refueling platform.

2. The UAV collaborative multi-strategy dynamic monitoring method for multiple monitoring points on a construction site according to claim 1, characterized in that, Step two involves calculating the three-dimensional feasible observation space for each monitoring point, based on its monitoring requirements, while satisfying the constraints of observation distance, unobstructed line of sight, and no-fly zone avoidance. This space serves as the optimal monitoring area for that monitoring point. Specifically: Best monitoring area Defined as a set of three-dimensional points satisfying multiple constraints, expressed as: ； in, Possible observation locations for the drone. The coordinates of the monitoring point; To constrain the observation distance, the observation distance must be within the minimum safe distance and based on the ground sampling resolution. Between the calculated maximum effective observation distances; To ensure unobstructed line of sight, a ray tracing algorithm is used to determine the set of observation lines and obstacles. No intersection; To constrain the observation perspective, the angle between the observation vector and the vertical direction and the normal vector of the measured surface is limited; To circumvent restrictions in no-fly zones and ensure that observation locations are not situated within static no-fly zones. or dynamic tower crane no-fly zone internal.

3. The UAV collaborative multi-strategy dynamic monitoring method for multiple monitoring points on a construction site according to claim 2, characterized in that, Step three specifically involves: When the sampling frequency or priority of a monitoring point is high, and the best monitoring area of ​​the monitoring point has no usable intersection with the best monitoring areas of other monitoring points, the monitoring point is treated as an independent monitoring point, and a one-to-one dedicated monitoring strategy is selected. Based on the sampling frequency of the monitoring point, an access time series is generated, and a dedicated drone is designated to monitor the monitoring point according to the access time series. When multiple monitoring points have low sampling frequencies and their optimal monitoring areas have available intersections, the multiple monitoring points are divided into the same monitoring point group, a one-to-many static monitoring strategy is selected, and a single hovering observation position is determined in the available intersection. The UAV is then deployed to that position, and time-division multiplexing observations are performed on multiple monitoring points through gimbal attitude adjustment. When multiple monitoring points have low sampling frequencies and their optimal monitoring areas have no usable overlap but are spatially close to each other, the multiple monitoring points are divided into the same monitoring point group, a one-to-many dynamic monitoring strategy is selected, and a round-trip flight path is generated between multiple monitoring areas, and the drone is designated to observe in sequence according to the monitoring sequence. When the sampling frequency of a monitoring point is low, and its location is remote or its optimal monitoring area is small and has no usable overlap with the optimal monitoring areas of other monitoring points, the monitoring point is treated as an independent monitoring point. A many-to-one collaborative monitoring strategy is selected, and the observation task of the monitoring point is inserted into the task gap of multiple assigned task drones to complete collaboratively.

4. The UAV collaborative multi-strategy dynamic monitoring method for multiple monitoring points on a construction site according to claim 3, characterized in that, Step four is as follows: First, based on the monitoring strategies formulated by each independent monitoring point or monitoring point group, the minimum number of drones required to perform all monitoring tasks is estimated, and based on the current drone swarm status data, the set of available drones is determined. For monitoring points employing a one-to-one dedicated monitoring strategy, their corresponding optimal monitoring area Internally, numerical optimization methods are used to solve for the optimal monitoring position for the assigned dedicated UAV under multi-target conditions. , is represented as: ； in, Indicates the drone's takeoff point The flight energy required to reach the candidate observation point v is obtained by time integration based on the dynamics model and power model of the UAV; Indicates at candidate observation points The predicted path loss of the communication link between the UAV and the monitoring center is calculated using the free space propagation model and the shadow fading model. It is a penalty function that measures the stability of the wind field at that location, calculated based on historical wind field data or CFD simulation results; These are the weights for energy consumption, communication link loss, and wind farm stability penalty, respectively. For a monitoring point group employing a one-to-many static monitoring strategy, select an optimal static hovering point within the public area. To balance the observation quality of multiple targets with link performance; let the comprehensive observation quality function be: ； in and These are normalized quality scores based on viewpoint and distance, respectively. To be based on monitoring weight The obtained normalized coefficients, These are, respectively, the view quality weight, the distance quality weight, and the communication loss penalty weight; An improved particle swarm optimization or multi-starting-point gradient descent method is employed to search for the maximum value within a common region. The point is used to obtain the optimal static monitoring position for the drone. In subsequent missions, the drone only needs to remain in the optimal static monitoring position. The system can hover nearby and perform time-sharing patrols of each monitoring point in the monitoring point group by controlling the attitude of the gimbal. For a monitoring point group that adopts a one-to-many dynamic monitoring strategy, a monitoring sequence that reflects the time-division multiplexing principle is formulated based on the sampling frequency, single monitoring duration and UAV flight speed parameters. A round-trip flight path is generated between multiple monitoring areas, and the UAV is designated to observe sequentially according to the monitoring sequence. For monitoring points that select a many-to-one collaborative monitoring strategy, a monitoring sequence that reflects the time-division multiplexing principle is formulated based on the assignment of multiple drone tasks and the sampling requirements of the monitoring points, and path planning is performed for the assigned drones based on the formulated monitoring sequence.

5. The UAV collaborative multi-strategy dynamic monitoring method for multiple monitoring points on a construction site according to claim 4, characterized in that, The minimum number of drones required to perform all monitoring tasks is estimated based on the monitoring strategy formulated by each independent monitoring point or monitoring point group, and the available drone set is determined based on the current drone cluster status data. Specifically: The problem of estimating the minimum number of drones is modeled as a set-coverage integer programming problem with time windows and energy constraints. Let the entire monitoring task cycle be... For monitoring points The sampling frequency is The duration of a single monitoring session is If the time limit is seconds, then the minimum number of accesses required within a task cycle is: ； Each access request is recorded as a task unit. ; Let the set of candidate drones be For each candidate drone Define binary variables Indicates enabling the first Deploy drones; for each task unit and drones Define variables Indicates by the first A drone was used for this visit; The objective function is to minimize the number of drones activated, expressed as: ； The constraints include task coverage constraints, single-machine time and energy constraints, and time window constraints; The task coverage constraint is expressed as follows: ； That is, each task unit must be carried out by one and only one drone; In the single-machine time and energy constraints, let the unmanned aerial vehicle (UAV) be considered. The total battery capacity is The charge / discharge efficiency is The power model for flight and hovering is The round trip time to the charging station is Then, the total effective working time available within the task cycle is... Constrained by these factors; For drones Following the planned path, fly from the previous task point to the monitoring point. Execute the The required flight time for this visit should meet the following requirements: ； The theoretical lower bound for the above integer programming problem is obtained by solving it. After calculating the theoretical lower bound based on the above model, and then combining the site geometry and no-fly zone to determine the transfer time... An estimate was made, and an approximate solution was obtained using a heuristic algorithm. This yielded the minimum number of drones required to simultaneously operate within the agreed task period to complete all monitoring tasks at each monitoring point, taking into account energy and no-fly zones. ; Subsequently, based on the drone status data, the drones were screened, and those with a battery level of at least 90% and a standby status constituted a set of usable drones. Meanwhile, the drones that are charging and the rest of the drones serve as backup or standby members of the group, which are used to take over the task during subsequent dynamic scheduling.

6. The UAV collaborative multi-strategy dynamic monitoring method for multiple monitoring points on a construction site according to claim 1, characterized in that, The drone energy network mentioned in step one includes: Fixed charging stations provide a reliable parking and charging platform for drones in fixed locations; The integrated solar-powered drone landing pad for tower cranes is installed on the tower of a high-altitude tower crane structure, providing high-altitude parking and charging services for drones. The autonomous mobile charging vehicle can autonomously plan its path and avoid obstacles in complex construction site environments. It has a photovoltaic power generation mechanism and can provide a flexible parking and charging platform for drones on the ground.

7. The UAV collaborative multi-strategy dynamic monitoring method for multiple monitoring points on a construction site according to claim 6, characterized in that, Step five describes continuously acquiring drone cluster status data and drone energy network status data during the drone's monitoring mission; specifically: From the moment the drone swarm takes off to perform monitoring tasks, the monitoring center enters the real-time monitoring and closed-loop control phase, continuously monitoring and estimating the drones in flight, fixed charging stations, tower crane integrated solar-powered drone landing pads, and automatic cruise mobile charging vehicles. For each drone in flight, the monitoring center receives its position, attitude, speed, and battery level data at a high frequency. The status information includes the current task number, task completion progress, and link quality. For fixed charging stations, tower crane integrated solar-powered drone landing pads, and automated mobile charging vehicles, the monitoring center collects the number of berths occupied, available capacity, queue length, and voltage and current parameters for each platform. In addition, for the integrated solar-powered drone landing pad for tower cranes, the monitoring center obtains the tower boom angular velocity in real time. Hook load The platform tilt angle and local wind speed information are used to calculate the safety status indicator variables of the integrated solar-powered UAV landing pad for tower cranes. ； in, The maximum permissible safe angular velocity threshold; The maximum allowable load limit; when When =1, the integrated solar-powered drone landing pad of the tower crane is considered a usable charging platform; when When the value is 0, it cannot be used for drone take-off and landing or charging. For autonomous cruise mobile charging vehicles, the monitoring center estimates their state vectors using the extended Kalman filter method, based on data from the Global Navigation Satellite System, inertial measurement unit, wheel speedometer, and lidar sensors. ； in, The position of the mobile charging vehicle in the ground coordinate system; This refers to the vehicle's heading angle; Linear velocity; ω represents the angular velocity; this state vector will be used for subsequent optimization of the mobile charging vehicle's cruising path and stopping position.

8. The UAV collaborative multi-strategy dynamic monitoring method for multiple monitoring points on a construction site according to claim 7, characterized in that, In step five, when the drone's battery level is detected to be below a preset threshold, based on the drone's current location, battery status, mission urgency markers, and drone energy network status data, a target refueling platform for the drone is designated, and a three-dimensional obstacle avoidance and return path for the drone to the target refueling platform is planned; specifically: First, determine the availability of each platform based on its capacity and current available capacity; for integrated tower crane solar-powered drone landing pads, further calculations are needed. To determine whether it is safe and usable; Subsequently, for each pair of drones and platform combinations The monitoring center estimated that from the drone Fly to the platform from the current location Required energy and time And based on the queue theory model, the expected waiting time of the platform is estimated. Meanwhile, to more accurately characterize the energy consumption and risks caused by the height difference, a height matching degree function is introduced. This is used to measure the impact of the difference between the drone's current operating altitude and the platform altitude on return-to-home energy consumption and safety. Thus, the definition of drones For the platform The utility function is: ； in, For drones Current remaining battery power; This is an estimate of the energy required to fly from the current location to platform CPj. This is an estimate of the flight time; In order to be on the platform The estimated waiting time in the queue; H align (k)(j) represents the high degree of matching score, and the larger the value, the more favorable the return flight; These represent the energy consumption weighting coefficient, the time cost weighting coefficient, and the vertical height penalty weighting coefficient, respectively. The monitoring center monitors each drone that needs charging. Calculate its utility value across all available platforms and select the one with the highest utility value as the optimal charging platform: ； In order to enable the parking location of the automatic cruise mobile charging vehicle to adapt to the changes in the flight hotspot of the UAV, a flight heat map is constructed based on the historical trajectory of the UAV, and the cruise path of the automatic cruise mobile charging vehicle is dynamically adjusted accordingly.

9. The UAV collaborative multi-strategy dynamic monitoring method for multiple monitoring points on a construction site according to claim 1, characterized in that, In step five, when a new task is added to the task waiting queue, a redundancy assessment is used to reallocate the task between the in-flying drone and the standby drone; specifically: Task waiting queue detected If the queue is not empty, retrieve the highest priority task from the head of the queue. A redundancy capability scoring function is used to evaluate the suitability of each in-flight and standby UAV for the mission. The redundancy capability scoring function is expressed as follows: ； in, For drones Without disrupting the existing sampling constraints for high-priority tasks, new tasks can be... Insertion time margin score; The energy margin score for drone k, which is able to safely return to base after completing its original task and then performing the new task at the current power level; This indicates that new tasks will be inserted into the drone. The incremental marginal cost resulting from the existing path includes additional flight time and additional energy consumption; These are time weight, energy weight, and cost increment weight, respectively. To prevent extremely small positive constants with a denominator of zero; scoring value The larger the value, the higher the overall adaptability of the drone to the task.

10. The UAV collaborative multi-strategy dynamic monitoring method for multiple monitoring points on a construction site according to claim 1, characterized in that, In step one, when constructing a computable digital twin scenario, each worker is equipped with a smart safety helmet. This smart safety helmet includes at least: a helmet structure, a main control module, a vital signs acquisition module, an environmental acquisition module, a positioning module, an activity intensity acquisition module, a communication module, and an alarm notification module. The vital signs acquisition module collects heart rate and heart rate variability-related signals; the environmental acquisition module collects thermal environment information such as skin temperature and ambient temperature and humidity; the inertial / activity intensity acquisition module collects triaxial acceleration and generates activity intensity indicators; the positioning module outputs the worker's position coordinates; and the communication module reports the safety helmet number, worker number, timestamp, position coordinates, and vital signs, environmental, and activity intensity data to the monitoring center. In step five, when the monitoring center enters the real-time monitoring and closed-loop control stage, it uses smart safety helmets to continuously collect the status of workers and report alarms. When the smart safety helmet detects that the worker's fatigue level reaches the preset alarm level, the monitoring center automatically generates an active intervention monitoring task based on the worker's real-time location and preset intervention strategy. The monitoring target of the active intervention monitoring task is the location of the worker who triggered the fatigue alarm. The monitoring actions include at least approaching and hovering, lighting, and voice reminders. The active intervention monitoring task is then added to the head of the task waiting queue with the highest scheduling priority.

Images