Low-altitude fire fighting system and method based on urban CIM and Beidou positioning

Abstract

The invention relates to the technical field of urban fire safety rescue, in particular to a low-altitude fire fighting system and method based on urban CIM and Beidou positioning. According to the system and the method, full space-time accurate positioning is realized through the Beidou differential base station and the UWB indoor positioning equipment; integrating the urban information model system and unmanned aerial vehicle cluster data, and constructing a multi-source heterogeneous data set; a coordinate system is unified based on a space-time reference calibration module, and fire behavior prediction, path planning and resource scheduling are realized through AI computing nodes; the unmanned aerial vehicle task distribution module is used for realizing multi-vehicle cooperative rescue; and iteratively optimizing system parameters through a closed-loop adaptive optimization module. The method is suitable for fire emergency rescue in an urban complex building environment, and the rescue response speed, the decision-making precision and the cooperative combat capability are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of urban fire safety rescue, and in particular to a low-altitude fire-fighting system and method based on urban CIM and Beidou positioning. BACKGROUND

[0002] With the acceleration of urbanization, traditional fire-fighting positioning relies on GPS or single Beidou signal, which is invalid in satellite signal-free areas such as high-rise buildings, underground spaces, etc. Even outdoors, high-density building shielding and electromagnetic interference can cause positioning accuracy to drop to meters, which cannot meet the needs of unmanned aerial vehicle autonomous obstacle avoidance, real-time tracking of rescue personnel, and precise locking of fire points. At the same time, lacking coordination and calibration with urban space structures (such as building load-bearing walls and fire passages), positioning data is disconnected from actual rescue scenarios, which can easily lead to risks such as unmanned aerial vehicle collision and personnel entering dangerous areas.

[0003] Traditional fire-fighting decisions rely on the experience of commanders and can only obtain limited ground survey data, and cannot integrate the global information of "building structure-fire dynamic-rescue force position" in real time. On the one hand, urban building BIM, pipe network, and other static data are not linked with real-time fire data. On the other hand, dynamic data such as infrared thermal maps and gas concentrations collected by unmanned aerial vehicles exist in isolation and cannot be combined with the spatial model of the city information model (CIM) platform to generate accurate decision-making schemes, resulting in delayed fire prediction, unreasonable rescue path planning, and low resource scheduling efficiency.

[0004] In the prior art, although there are separate applications of Beidou positioning or unmanned aerial vehicle fire-fighting, the deep integration of Beidou, UWB positioning, CIM space coordination, and dynamic decision-making of unmanned aerial vehicle data has not been achieved, making it difficult to break through the core bottlenecks of inaccurate positioning and delayed decision-making in complex environments, and there is an urgent need for a low-altitude fire-fighting solution based on accurate positioning and data fusion decision-making. SUMMARY

[0005] The present application aims to solve the problems of insufficient positioning accuracy, low multi-source data fusion, lack of dynamic decision-making capability, and weak system self-evolution capability in the prior art, and provides a low-altitude fire-fighting precise positioning and navigation and dynamic decision-making system and method based on urban CIM and Beidou positioning. The system realizes centimeter / decimeter-level accurate positioning and time-space synchronization through the fusion of Beidou differential positioning and UWB indoor positioning; constructs a high-fidelity fire scene digital twin environment through multi-modal data collection of unmanned aerial vehicle clusters and dynamic model updating of the CIM platform; realizes dynamic decision-making of fire prediction, resource scheduling, and path planning through AI algorithms; continuously improves system performance through a closed-loop optimization mechanism, and ultimately realizes efficient low-altitude fire rescue through "air-ground-person" cooperation.

[0006] To achieve the above objectives, this invention provides a low-altitude firefighting system based on urban CIM and BeiDou positioning, including BeiDou positioning equipment, UAV swarm equipment, urban information modeling equipment, urban fire data center, and fire dispatch command terminal. The BeiDou positioning equipment is configured to generate and transmit BeiDou differential signals and UWB positioning signals, providing indoor and outdoor centimeter-level positioning and unified timing, transmitting UAV positioning data and personnel positioning data, receiving anti-interference commands, and triggering inertial navigation backup. The UAV swarm equipment transmits BeiDou differential signals and UWB positioning signals with the BeiDou positioning equipment. The UAV swarm equipment is also configured to collect and transmit building point clouds, infrared thermal maps, and building structure data, receive flight control commands for performing window breaking, firefighting, and material delivery tasks, and transmit first-person view video and task execution logs. The urban CIM equipment is data-connected to the UAV swarm equipment. The urban CIM equipment is configured to receive building structure data transmitted by the UAV swarm equipment, update the CIM model, mark building structures and risk areas, and transmit BIM models, building material parameters, obstacle data, and CIM 3D models. The city's fire protection data center is connected to BeiDou positioning equipment, sending anti-interference trigger commands to the equipment. It is also connected to drone swarm equipment, sending drone mission commands to control the drone swarm equipment to perform firefighting, window breaking, or scanning tasks. Furthermore, the city's fire protection data center is connected to urban CIM equipment, receiving static BIM models and dynamic structural damage data from the CIM equipment. The data center is also configured to integrate BeiDou positioning data, CIM models, building point clouds, infrared thermal maps, and building structural data to achieve spatiotemporal alignment. Based on AI algorithms, it generates fire prevention decision-making schemes, including fire prediction schemes, path planning schemes, and firefighting and rescue schemes. Finally, it receives full-process logs, evaluates accuracy, and outputs equipment and algorithm optimization parameters. The fire dispatch command terminal is connected to the Beidou positioning equipment to receive the location data of rescue personnel sent by the Beidou positioning equipment; the fire dispatch command terminal is connected to the drone swarm equipment to receive the first-person view video sent by the drone swarm equipment and send task dispatch instructions to the drone swarm equipment; the fire dispatch command terminal is connected to the city CIM equipment to receive the CIM 3D model sent by the city CIM equipment; the fire dispatch command terminal is connected to the city fire data center to receive the fire decision plan sent by the city fire data center and send decision adjustment requests to the city fire data center.

[0007] This invention also provides a low-altitude firefighting method based on urban CIM and BeiDou positioning, comprising the following steps: completing equipment self-inspection and spatiotemporal calibration, establishing a unified centimeter-level positioning and millisecond-level timing reference; collecting point cloud data of building facades and building interior structure data, transmitting the scanned data to the CIM system for point cloud registration and model updating, generating a digital twin model of the fire scene; integrating BeiDou positioning data, CIM model, and UAV sensor data, generating a firefighting decision-making scheme including fire prediction, personnel positioning, path planning, and resource scheduling through AI algorithms; receiving the firefighting decision-making scheme, triggering coordinated rescue by UAVs, fire trucks, and personnel, providing real-time feedback on execution status, and dynamically adjusting instructions; collecting full-process data including equipment status, rescue logs, fire data, and the execution effect of the decision-making scheme, evaluating the accuracy of the CIM model, BeiDou positioning accuracy, UWB positioning accuracy, and AI decision-making accuracy, optimizing equipment parameters and algorithm models, and updating historical databases and rescue plans.

[0008] The beneficial effects of this invention are as follows: 1) By fusing BeiDou differential and UWB multimodal positioning, the problem of insufficient positioning accuracy in complex environments is solved, achieving centimeter / sub-meter level positioning with seamless indoor and outdoor connections; 2) By deeply integrating UAV data with the CIM platform, data silos are broken down, providing a high-fidelity digital twin environment for dynamic decision-making; 3) By using AI algorithms, real-time fire prediction, dynamic resource scheduling, and adaptive path optimization are achieved, improving the accuracy and timeliness of decision-making; 4) By using a closed-loop optimization mechanism, the system achieves self-evolution, continuously improving rescue efficiency and safety, and providing an intelligent and precise solution for urban fire safety. Attached Figure Description

[0009] Figure 1 This is a block diagram of a low-altitude fire protection system based on urban CIM and BeiDou positioning.

[0010] Figure 2 This is a schematic diagram of a low-altitude firefighting method based on urban CIM and BeiDou positioning.

[0011] Figure 3 This is a flowchart of S10.

[0012] Figure 4 This is a flowchart of S101.

[0013] Figure 5 This is a flowchart of S102.

[0014] Figure 6 This is a flowchart of S103.

[0015] Figure 7 This is a flowchart of S104.

[0016] Figure 8This is a flowchart of S20.

[0017] Figure 9 This is a flowchart of S201.

[0018] Figure 10 This is a flowchart of S202.

[0019] Figure 11 This is a flowchart of S203.

[0020] Figure 12 This is a flowchart of S204.

[0021] Figure 13 This is a flowchart of S30.

[0022] Figure 14 This is a flowchart of S301.

[0023] Figure 15 This is a flowchart of S302.

[0024] Figure 16 This is a flowchart of S303.

[0025] Figure 17 This is a flowchart of S304.

[0026] Figure 18 This is a flowchart of S40.

[0027] Figure 19 This is a flowchart of S401.

[0028] Figure 20 This is a flowchart of S402.

[0029] Figure 21 This is a flowchart of S403.

[0030] Figure 22 This is a flowchart of the S50 process.

[0031] Figure 23 This is a flowchart of S501.

[0032] Figure 24 This is a flowchart of S502.

[0033] Figure 25 This is a flowchart of S503.

[0034] Figure 26 This is a flowchart of S504. Detailed Implementation

[0035] The following detailed description of embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0036] For fire emergency scenarios in complex urban environments (high-density high-rise areas, historical districts, mountain parks), this invention focuses on the precise positioning and navigation of firefighting elements (drone swarms, rescue personnel, ground vehicles), as well as dynamic rescue decision-making technology based on the fusion of city information model (CIM) and real-time drone data.

[0037] To overcome positioning limitations in complex environments (seamless indoor-outdoor connections, high-rise obstructions, electromagnetic interference), centimeter-level / sub-meter-level precise positioning and safe navigation for drone swarms, rescue personnel, and ground fire trucks are required, with positioning data and CIM building structure information being collaboratively calibrated. Furthermore, a deep fusion mechanism between CIM static data and drone dynamic data is being constructed. AI algorithms are used to generate real-time fire predictions, optimal rescue routes, and resource scheduling plans, upgrading traditional experience-based decision-making to data-driven dynamic decision-making, shortening decision response time and improving plan accuracy.

[0038] The BeiDou positioning signal mentioned in this article refers to the positioning service provided by China's independently developed Global Navigation Satellite System (BDS). In this system, satellite signals are received through BeiDou differential base stations, and the positioning accuracy is improved to within 1 meter by combining differential correction technology (RTCM3.3 format), which is used to unify outdoor spatial references (such as the positions of fire trucks and drones).

[0039] The UWB positioning described in this article uses ultra-wideband technology to achieve high-precision ranging through nanosecond-level narrow pulse signals. This system uses the Time-of-Flight (ToF) algorithm to calculate the distance between the base station and the tag, with a positioning accuracy of ≤3 cm. It is mainly used for indoor personnel or equipment positioning (such as the location of firefighters or tags of trapped personnel).

[0040] The City Information Modeling (BIM) data used in this paper is an extension of Building Information Modeling, integrating structured data such as building 3D geometry, materials, and fire protection facilities, stored in IFC format. This system uses BIM to acquire information such as building interior layout and combustible material distribution, supporting fire prediction and path planning.

[0041] The RTCM3.3 format used in this article is a differential protocol standard of Radio Technical Commission for Maritime Services, used to transmit correction data from BeiDou satellites. This system uses this format to achieve real-time data exchange between the base station and the terminal, ensuring positioning accuracy.

[0042] The MQTT protocol used in this paper is a lightweight IoT communication protocol based on a publish / subscribe model. This system uses MQTT to integrate heterogeneous data from multiple sources, such as BeiDou, UWB, and UAVs, to achieve efficient spatiotemporal alignment.

[0043] The timestamp alignment algorithm in this paper synchronizes data of different frequencies (such as BeiDou 1Hz and UWB 10Hz) to a unified time base (UTC) through linear interpolation or Kalman filtering. This system requires a time alignment error of ≤0.1 seconds to ensure the temporal consistency of decision data.

[0044] The thermal radiation propagation model in this paper is a mathematical model based on the principle of heat transfer to simulate flame spread. The input parameters include the material thermal diffusivity κ, wind speed v, and combustible density ρfuel, etc., which are used to predict the direction and speed of fire spread.

[0045] The improved A* algorithm presented in this paper is a path planning algorithm. It optimizes the traditional A* algorithm by using weight coefficients (w1 = path length, w2 = dangerous area) and dynamic adjustment factor λ. This system updates the path once per second, balancing efficiency and safety.

[0046] The DDPG algorithm presented in this paper is a deep deterministic policy gradient reinforcement learning algorithm used for resource scheduling (such as drone allocation). This system generates decisions based on personnel priority and fire severity, with a latency of ≤500 milliseconds.

[0047] The five-lens oblique camera system described in this paper, equipped with five lenses, can simultaneously acquire multi-angle images of buildings for 3D modeling. This system generates oblique photogrammetry models from 4K resolution images, supporting GIS+BIM visualization on command terminals.

[0048] The PWM signal used in this article is a Pulse Width Modulation signal, used to control the ejection module of the window-breaking drone. This system achieves precise window breaking by adjusting the blast delay (adjustable from 0.5 to 2 seconds) using PWM.

[0049] The CAN bus mentioned in this article refers to the Controller Area Network (CAN) protocol, used for communication between internal UAV devices. This system transmits dry powder spraying commands (dispensing volume adjustable from 0.5 to 5 kg) via the CAN bus.

[0050] The LoRa communication described in this article is a long-range, low-power wide-area network technology. This system supports remote communication with 4G / 5G dual-mode command terminals and is suitable for signal-shielded areas (such as underground parking garages).

[0051] The CFD model used in this paper is a computational fluid dynamics model, which improves the accuracy of fire prediction by numerically simulating parameters such as temperature field and oxygen concentration. The system requires a calculation time of ≤30 seconds per calculation.

[0052] The RRT* algorithm presented in this paper is a fast extended random tree optimization algorithm suitable for path planning in complex unstructured environments (such as collapsed buildings). This system optimizes path length by 10%–15% compared to the A* algorithm.

[0053] The millimeter-wave radar described in this article operates in the 30–300 GHz frequency band and is used to identify ground obstacles when UWB base stations are airdropped by drones, ensuring safe automatic deployment (networking time ≤ 3 minutes).

[0054] The Bayesian optimization algorithm presented in this paper is a parameter optimization method for probabilistic models. This system is used to automatically search for optimal fire prediction parameters (such as the κ value), which improves efficiency by 50% compared to manual adjustment.

[0055] On the one hand, this invention provides a low-altitude firefighting system based on urban CIM and BeiDou positioning. This system uses BeiDou high-precision positioning as the spatiotemporal reference, urban CIM as the digital foundation, UAV swarms as the execution carrier, and a data center as the decision-making core. Through hardware linkage of physical devices, information transmission via data links, and command interaction of control signals, it achieves low-altitude firefighting and rescue in complex environments.

[0056] Figure 1 This is a low-altitude firefighting system based on urban CIM and BeiDou positioning according to one embodiment of the present invention. The low-altitude firefighting system 100 includes: a BeiDou positioning device 10, a drone swarm device 20, an urban information modeling device 30, an urban fire data center 40, and a fire dispatch command terminal 50. In equivalent embodiments, the system 100 may include additional or fewer components.

[0057] The Beidou positioning device 10 is configured to generate and send Beidou differential signals and UWB positioning signals, provide indoor and outdoor centimeter-level positioning and unified timing, transmit UAV positioning data and personnel positioning data, receive anti-interference commands, and trigger inertial navigation backup.

[0058] The UAV swarm equipment 20 transmits BeiDou differential signals and UWB positioning signals with the BeiDou positioning equipment. The UAV swarm equipment is also configured to collect and send building point clouds, infrared thermal maps and building structure data, receive flight control commands for performing tasks such as breaking windows, fire fighting and material delivery, and send first-person view videos and task execution logs.

[0059] The city CIM device 30 is connected to the drone swarm device for data transmission. The city CIM device is configured to receive building structure data sent by the drone swarm device, update the CIM model, mark the building structure and risk areas, and send the BIM model, building material parameters, obstacle data and CIM 3D model.

[0060] The city fire data center 40 is connected to the Beidou positioning equipment, sending anti-interference trigger commands to the Beidou positioning equipment; the city fire data center is connected to the drone swarm equipment, sending drone mission commands to the drone swarm equipment, which are used to control the drone swarm equipment to perform fire fighting, window breaking, or scanning tasks; the city fire data center is connected to the city CIM equipment, receiving static BIM models and building dynamic structural damage data sent by the city CIM equipment; the city data center is also configured to integrate Beidou positioning data, CIM models, building point clouds, infrared thermal maps, and building structural data to achieve spatiotemporal alignment; it generates fire-fighting decision-making schemes based on AI algorithms, including fire prediction schemes, path planning schemes, and fire fighting and rescue schemes; and it receives full-process logs, evaluates accuracy, and outputs equipment optimization parameters and algorithm optimization parameters.

[0061] The fire dispatch command terminal 50 is connected to the Beidou positioning equipment to receive the location data of rescue personnel sent by the Beidou positioning equipment; the fire dispatch command terminal is connected to the drone swarm equipment to receive the first-person view video sent by the drone swarm equipment and send task dispatch instructions to the drone swarm equipment; the fire dispatch command terminal is connected to the city CIM equipment to receive the CIM 3D model sent by the city CIM equipment; the fire dispatch command terminal is connected to the city fire data center to receive the fire decision plan sent by the city fire data center and send decision adjustment requests to the city fire data center.

[0062] The Beidou positioning device 10 includes a Beidou positioning terminal 101, a Beidou differential reference device 102, a UWB indoor positioning device 103, and an inertial navigation backup device 104.

[0063] The Beidou positioning terminal 101 can be a UAV-borne Beidou-3 three-band receiver integrated into the UAV flight control system as a miniaturized module; it can also be a vehicle-mounted Beidou receiver fixed to the control panel of a ground fire truck; or it can be a handheld Beidou positioning terminal for firefighters, wrist-mounted, with a battery life of ≥12 hours. The Beidou positioning terminal 101 receives signals in the B1C / B2a / B3I frequency bands; its positioning accuracy is 1cm+1ppm horizontally and 2cm+1ppm vertically; and its anti-interference level is ≥40dB.

[0064] The BeiDou differential reference equipment 102 can be a fixed BeiDou differential base station, deployed in an open area within 5km of the fire scene, equipped with a solar power panel; or it can be a mobile BeiDou differential base station, mounted on an emergency command vehicle, capable of rapid deployment. The BeiDou differential reference equipment 102 has a coverage range of ≥5km, an output frequency range of 1Hz, and a differential signal format of RTCM3.3.

[0065] The UWB indoor positioning device 103 can be a wall-mounted UWB positioning base station, deployed in building corridors / refuge floors, requiring 220V power supply; it can be an airdropped UWB positioning base station, deployed by drones, with a battery life of ≥8 hours and an IP67 waterproof rating; it can be a wearable UWB tag, similar to a firefighter's badge, weighing ≤50g; or it can be an adhesive UWB tag, attached to the drone fuselage / rescue supply box. The positioning accuracy of the UWB indoor positioning device is ≤3cm; the communication frequency band is 3.1-10.6GHz; and the coverage radius is ≤50m (single base station).

[0066] The inertial navigation backup device 104 can be an UAV-borne IMU module (integrated into the BeiDou receiver and linked with the flight control); or a handheld IMU unit (built into the BeiDou terminal, with a drift rate ≤0.1° / h). The inertial navigation backup device 104 has a sampling rate of 100Hz, an attitude accuracy of ±0.5°, and supports data fusion with BeiDou / UWB.

[0067] The BeiDou differential reference device 102 is signal-connected to the BeiDou positioning terminal 101. For example, a fixed BeiDou differential base station and an UAV-borne BeiDou receiver are wirelessly connected via satellite signals (B1C / B2a / B3I band electromagnetic waves). The differential base station sends positioning correction signals to the receiver, improving the UAV's positioning accuracy from meter-level to centimeter-level. Similarly, a mobile BeiDou differential base station and a vehicle-mounted BeiDou receiver are connected via satellite signals, prioritizing the B2a band for stronger anti-interference capabilities. With an antenna height ≥2m, it provides dynamic differential positioning for ground fire trucks, ensuring accurate arrival at fire hydrant inlets (CIM-marked locations).

[0068] In addition, the wall-mounted UWB base station in the UWB indoor positioning device 103 connects wirelessly with the wearable UWB tag via UWB radio waves (5.8GHz band). The base station and tag automatically pair (triggered when the distance is ≤10m). When there is no BeiDou signal indoors, the tag sends a location signal to the base station, enabling firefighters to locate the target at centimeter-level accuracy. The air-dropped UWB base station connects wirelessly with the UWB tag attached to the drone's fuselage via UWB radio waves, for example, using the 3.5GHz band, which has strong wall penetration capabilities. After activation, the air-dropped UWB base station automatically scans the tag with a response time of ≤0.5s, providing positioning coverage for drones flying indoors and preventing collisions with load-bearing walls.

[0069] The Beidou positioning terminal 101 also connects to the city's fire data center 40 via a 5G wireless link, for example, in SA (Standalone) networking, using the NMEA0183 standardized positioning data format, with an update frequency of ≥10Hz. For instance, firefighters can use the handheld Beidou terminal to transmit their real-time location and movement speed to the data center for rescue route optimization.

[0070] UWB indoor positioning device 103 and drone 20 are connected via a UWB wireless link. For example, a UWB tag attached to relief supplies connects to the control module of drone 20, feeding back the location of the supplies to the group control system to ensure accurate delivery of relief supplies to the target personnel. The data format of these connections can be JSON, containing the supplies ID, remaining quantity, and real-time coordinates.

[0071] The Beidou differential reference device 102 is connected to the signal control of the city fire data center 40. For example, the fixed Beidou differential base station is connected to the edge server of the city fire data center through a wireless digital signal in the 433MHz industrial frequency band. The command format is 8-bit binary (high level 1 = normal, 0 = fault). When the fixed Beidou differential base station detects that the satellite signal-to-noise ratio is <10dB, it sends a "weak signal" warning to the edge server and triggers the anti-interference strategy.

[0072] The drone swarm equipment 20 includes a functional drone 201, drone-mounted sensors 202, drone actuators 203, and drone flight control unit 204.

[0073] The functional drone 201 can be a heavy-duty drone, such as a rotary-wing type, with a payload of 5kg, used for external scanning and material delivery; it can be a light-duty drone, such as a rotary-wing type, with a payload of 2kg, used for indoor scanning and light material delivery; it can be a window-breaking drone, a fixed-wing type equipped with window-breaking munitions, used for blasting high-rise glass curtain walls; it can be a fire-fighting drone, a multi-rotor type equipped with a dry powder spray device, with a capacity of 10L; or it can be a reconnaissance drone, a micro-type drone, weighing ≤500g, used for fire detection in confined spaces. Heavy-duty functional drones have a flight speed of 15m / s, an endurance of 40min, and a wind resistance rating of ≥6. Light-duty functional drones have a flight speed of 8m / s, an endurance of 30min, and a wind resistance rating of ≥6.

[0074] The UAV can be equipped with the following sensors: Sensor 202 can be a 1.48MP five-lens tilting camera (e.g., integrated into heavy UAVs for multi-angle shooting); a solid-state LiDAR with ±2cm accuracy (integrated into heavy / light UAVs); a through-wall radar with a detection depth ≥50cm (integrated into light UAVs); an infrared thermal imager with a resolution of 640×512 (integrated into reconnaissance / firefighting UAVs); a gas sensor for detecting CO / NO2 with a range of 0-1000ppm (integrated into reconnaissance UAVs); or an ultrasonic sensor with a detection radius of 10m (integrated into indoor light UAVs). The LiDAR point cloud density should be ≥300 points / m². The infrared thermal imager should have a temperature measurement range of -20 to 1500℃. The gas sensor should have an accuracy of ±5%FS.

[0075] The drone actuator 203 can be a dry powder spraying device, specifically for fire-fighting drones, with a spraying pressure ≥0.8MPa and a spraying device coverage range ≤5m; it can be a window-breaking projectile launching module, specifically for window-breaking drones, with a single projectile explosive force ≥5MPa and an effective distance of ≤10m; it can be a pneumatic ejection chamber, specifically for heavy drones, used for the delivery of rappelling equipment, with an ejection chamber delivery error ≤10cm; or it can be a rope rappelling package release device, specifically for light drones, used for the delivery of gas masks / towels.

[0076] Heavy-duty drones can be connected to photovoltaic charging stations for charging. For example, when a heavy-duty drone lands, it automatically connects to the Type-C fast charging port of the photovoltaic charging station (supporting 600W power) to replenish its power, reserving 15% of the power for return. A single charge can provide 30 minutes of flight time.

[0077] The lightweight drone is wired to the ultrasonic sensor. For example, the ultrasonic sensor is soldered onto the underside of the drone's arm and connected via an I2C communication interface. When flying indoors, the sensor detects obstacles (such as walls / pipes) in real time and triggers obstacle avoidance maneuvers.

[0078] Dry powder spraying devices are installed on fire-fighting drones. For example, the dry powder spraying device is bolted in place (made of 304 stainless steel) and connected to the drone's flight control system via a CAN bus. After receiving commands, the flight control system controls the spraying device to accurately deliver the dry powder extinguishing agent.

[0079] A window-breaking munition launching module is installed on the window-breaking drone. For example, the window-breaking munition launching module is fixed with a snap-fit ​​(allowing for quick magazine replacement) and is connected to the drone's fire control system via a PWM signal. After receiving the window-breaking coordinates, the fire control system outputs a trigger signal to the launching module to achieve directional detonation.

[0080] The reconnaissance drone can identify fires or risk areas using an infrared thermal imager. It connects to the edge server of the city's fire data center via dual 5G and satellite links, prioritizing 5G and switching to satellite when the public network is interrupted. The data protocol is MQTT and the data format is JPEG2000. It can transmit real-time thermal maps of the fire scene back to the edge server for core combustion identification.

[0081] The drone, equipped with sensor 202, is connected to the city information model server 30 via a wired Ethernet link. For example, a heavy-duty drone's solid-state LiDAR is connected to a LiDAR point cloud storage device via a CAT6 network cable, with data format LAS and a transmission rate ≥1GB / s. This data is uploaded to the CIM system for dynamic updates of the CIM model on the city information model server 30.

[0082] The drone, equipped with sensor 202, connects to the city's fire protection data center via a mesh network. For example, the through-wall radar on the lightweight drone transmits building internal structural data, including wall depth and crack width, to the city's fire protection data center through a mesh network that relays data between drones. The data protocol is TCP / IP. This data is used for load-bearing wall damage assessment.

[0083] The UAV flight control unit 204 is connected to the heavy UAV via wireless control signals. For example, the UAV flight control unit 204 includes an airspace management terminal, which connects to the heavy UAV via encrypted transmission in the 2.4GHz band. The command format is JSON (including flight path coordinates and flight altitude); the latency is ≤100ms. It sends serpentine flight path commands to the heavy UAV (e.g., altitude 100m, flight path spacing 5m) to control the external scanning path.

[0084] The drone flight control unit 204 is connected to the fire extinguishing drone via wireless digital signals. For example, the drone flight control unit 204 includes a cluster control terminal, which connects to the fire extinguishing drone via the 5.8GHz frequency band. The command is a high level 1 = spray, 0 = stop; the spray volume control is a PWM duty cycle (0-100%), which sends spray commands to the fire extinguishing drone and controls the amount of fire extinguishing agent released (adjusting the duty cycle according to the size of the fire).

[0085] The UAV flight control unit 204 includes an airspace management terminal, a flight path planning computer, a civil aviation airspace docking terminal, a cluster control terminal, a command transmission module, and a UAV status monitor. The airspace management terminal is used for airspace planning / avoiding no-fly zones, with a response time ≤0.5 seconds. The flight path planning computer generates UAV flight paths, with a calculation time ≤3 seconds for a single UAV. The civil aviation airspace docking terminal receives civil aviation no-fly zone data, with a data update frequency ≥1 time / minute. The cluster control terminal monitors the status of multiple UAVs, supporting simultaneous management of ≥20 UAVs. The command transmission module is high-powered, with a coverage radius ≥1km and a latency ≤100ms. The UAV status monitor monitors battery / sensor status in real time, with a sampling frequency ≥1Hz.

[0086] The airspace management terminal is wired to the route planning computer, for example via Ethernet (CAT6 cable). The airspace management terminal transmits airspace parameters (no-fly zone coordinates, flight altitude restrictions) to the planning computer to generate compliant routes.

[0087] The cluster control terminal is wired to the command transmission module, such as via an RS485 bus (two-wire system) with a transmission distance ≤100m and a command transmission rate of 9600bps. The cluster control terminal encodes the fire extinguishing / deployment commands and then wirelessly transmits them to the drone via the command transmission module.

[0088] The drone status monitor is wired to the battery management system of the heavy drone, such as a custom ribbon cable with an XT60 interface, to directly collect battery voltage / current / remaining power, monitor and read the drone battery status in real time, and trigger a return-to-home command when the remaining power is less than 15%.

[0089] The airspace management terminal connects to the light drone via signal control, for example, through a 2.4GHz wireless control signal. Commands control the drone's takeoff / landing / hovering, triggered by a high-level (3.3V) voltage. The airspace management terminal sends takeoff commands to the light drone, controlling it to enter the building through a window for indoor scanning.

[0090] The cluster control terminal connects with the window-breaking drone via signal control, for example, through an interference-resistant 5.8GHz wireless digital signal band. It commands and controls the window-breaking coordinates and blasting delay (0-5s), with an accuracy of ≤3cm. The cluster control terminal sends directional blasting commands to the window-breaking drone, ensuring the blasting location is aligned with a high-temperature area (infrared recognition).

[0091] The flight path planning computer is connected to the signal control system of multiple lightweight UAVs to achieve a mesh self-organizing network. It sends synchronization commands to the UAVs, partitions flight paths (to avoid collisions), synchronizes time using BeiDou timing, and maintains an error of ≤1ms. The flight path planning computer allocates indoor scanning areas to the UAVs to improve scanning efficiency; for example, two UAVs may be assigned to each of the east and west sides of the 30th floor.

[0092] Multiple reconnaissance drones are connected via a Mesh self-organizing network and synchronized through timestamp alignment (BeiDou time synchronization, error ≤1ms). Multiple drones work together to allocate scanning areas, avoiding duplicate scanning and improving fire detection efficiency.

[0093] The city information modeling equipment 30 includes a city information server, data storage equipment, and data access equipment.

[0094] The city information server can be a BIM model storage node, such as a rack-mounted server with a height of 2U and a storage node capacity of ≥100TB, used to store building BIM models; it can be a 3D rendering node, such as a GPU server equipped with an RTX4090 and a rendering node frame rate of ≥60FPS, used for CIM model visualization; or it can be a spatial index server, used for quickly retrieving building structure data, with an index server response time of ≤100ms.

[0095] Data storage devices can be LiDAR point cloud storage devices, distributed storage arrays, supporting hot backup, with a read / write rate of ≥1GB / s; or IoT data storage servers, supporting access for 100,000 devices, used to store real-time sensor data.

[0096] Data access devices can be IoT data access terminals, RS485 / Ethernet converters, with a baud rate of 9600-115200bps, used for accessing sensor data; or BIM model import terminals, supporting Revit / IFC formats, with import terminal format compatibility ≥95%, deployed on the CIM operation and maintenance console.

[0097] The BIM model storage node and the 3D rendering node are connected via Ethernet wired connection. For example, they can be directly connected to the same switch via a CAT6a network cable. The rendering node reads the BIM model from the storage node and generates a 3D visualization scene for display on the command platform.

[0098] The LiDAR point cloud storage device is wired to the spatial index server, for example, through a SAS interface. The storage array is connected to the server via a RAID card. The spatial index server reads point cloud data from the storage device, builds a spatial index, and accelerates the retrieval of structural data.

[0099] The IoT data access terminal is wired to the building smoke sensor (external device), for example, via RS485 bus (two-wire system, transmission distance ≤1000m), powered by 24V DC. The access terminal receives the fire alarm signal from the smoke sensor and forwards it to the CIM system to trigger an early warning.

[0100] The BIM model storage node is connected to the city fire data center 40, for example, to the multi-source data fusion gateway Ethernet (CAT6 cable) of the city fire data center 40. The data protocol is HTTP / HTTPS and the data format is IFC (BIM exchange format). It transmits static BIM data of the building (coordinates of load-bearing walls, width of fire lanes) to the data center for fire prediction.

[0101] The lidar point cloud storage device is connected to the city's fire protection data center via a fiber optic link (single-mode fiber, transmission distance ≥10km), with the data protocol being FTP (batch transfer of point cloud) and a transmission rate ≥10Gbps. The point cloud data is transmitted to the AI ​​node for registration with the BIM model and calculation of structural deviations.

[0102] The 3D rendering node connects to the fire dispatch platform via a data connection, for example, through an HDMI 2.1 interface (wired), with a video format of 4K@60Hz (supporting HDR) and a latency of ≤10ms. It outputs a CIM 3D model (overlaid with fire / personnel location) to the command display to assist the commander in decision-making.

[0103] The spatial index server is connected to the UAV flight control unit 204 via signal control. For example, it is connected via wireless digital signal (433MHz band), and the command format is XML (including fire lane coordinates and obstacle locations) to plan the UAV flight path and output building obstacle data for UAV obstacle avoidance path generation.

[0104] The BIM model import terminal is connected to the city fire data center via a 40 signal control system, such as a wired command (RS232 interface). The trigger command is a high level (1 = import complete, 0 = importing in progress). After the BIM model import is complete, a "data ready" signal is sent to the historical database to trigger historical fire matching.

[0105] The city fire data center 40 includes an edge server 401, an AI computing device 402, a database device 403, and a multi-source data fusion gateway 404.

[0106] Edge server 401, which can be rack-mounted, 4U in height, is deployed near the fire scene for real-time data noise reduction / stitching. CPU: Intel Xeon Gold 6426R; Memory: 128GB DDR5; End-to-end latency ≤50ms.

[0107] AI computing device 402 can be an AI computing node, such as a GPU cluster with 4 nodes, used for fire prediction / path planning; it can be a deep learning accelerator card (NVIDIA A100, used for model inference), with a single node computing power of ≥320TFLOPS; inference speed of ≥100FPS; and support for TensorFlow / PyTorch frameworks.

[0108] Database device 403 can be a historical fire database server, a relational database such as MySQL, with a historical database capacity of ≥50TB, used to store historical fire cases; or it can be a real-time database server, used to store real-time location / fire data, with a real-time database read / write throughput of ≥100,000 times / second; and a data backup cycle of ≤1 hour.

[0109] Multi-source data fusion gateway 404, industrial grade, supports multi-protocol conversion, used to integrate Beidou / CIM / UAV data, supported protocols: MQTT / NMEA0183 / HTTP; access channels ≥100; data fusion latency ≤200ms.

[0110] The edge server 401 is connected to the multi-source data fusion gateway 404 via Ethernet (CAT6 cable) and directly accesses the same local area network. The edge server transmits the pre-processed sensor data (noise-reduced point cloud, defogging infrared image) to the fusion gateway.

[0111] AI computing nodes are connected to deep learning accelerator cards via wired connections, such as through a PCIe 5.0 slot (directly installed on the AI ​​node motherboard); the data transfer rate is ≥32GB / s, and the accelerator card provides computing power support for AI nodes, improving the computing speed for fire spread prediction and personnel identification.

[0112] The historical fire data database server and the real-time database server are connected via a fiber optic link, with dual-fiber bidirectional transmission, redundant backup, and the database synchronization protocol is [missing information].

[0113] MySQLReplication synchronizes rescue data (personnel trajectories, firefighting effectiveness) to a historical database in real time for model training.

[0114] Edge server 401 connects to the gas sensor on the reconnaissance drone via, for example, a 5G wireless link, using the CoAP protocol. The data content includes CO concentration and detection time. The edge server receives the gas concentration data, removes outliers, and forwards it to the multi-source data fusion gateway 404. If the concentration is greater than 1000 ppm, it is considered a sensor malfunction.

[0115] The multi-source data fusion gateway 404 connects to the Beidou positioning terminal 101 via a 5G / Mesh dual-link data connection, for example, with a handheld Beidou terminal. The data protocol is NMEA0183, and the data content includes personnel latitude, longitude, elevation, and timestamp. The multi-source data fusion gateway 404 integrates firefighter positioning data and aligns it with CIM building coordinates for personnel location visualization.

[0116] The AI ​​computing node connects to the BIM model storage node via data, for example, through Ethernet, using the gRPC data protocol. The data content includes the ignition point of building materials and structural parameters. The AI ​​computing node reads building parameters from the BIM model storage node for calculating heat radiation propagation models and predicting fire spread.

[0117] The multi-source data fusion gateway 404 is connected to the fire dispatch command terminal 50 via signal control, for example, through a wired encrypted command network. The command format is JSON, and the commands include controls on fire level and rescue priority, with an update frequency of ≥1 time / 30 seconds. The multi-source data fusion gateway 404 outputs the fused disaster data to the fire dispatch command terminal 50, triggering the generation of a rescue plan.

[0118] The AI ​​computing node is connected to the drone flight control unit 204 via signal control, for example, through encrypted transmission of commands in the 2.4GHz frequency band using wireless digital signals. These commands include controls on fire extinguishing coordinates and deployment quantities, with an error of ≤5cm. The AI ​​computing node sends the calculated optimal fire extinguishing coordinates (based on the core location of the fire) to the drone flight control unit 204 to control the actions of the fire extinguishing drone.

[0119] The real-time database server is connected to the fire dispatch command terminal 50 via signal control, for example, through VHF radio frequency control that conforms to fire radio standards. This control directs the location of fire truck assembly points and water supply interfaces. The real-time database server sends ground fire truck dispatch instructions to the fire dispatch command terminal 50 to ensure vehicles arrive at their designated locations accurately.

[0120] The fire dispatch command terminal 50 includes a command terminal 501, a communication terminal 502, and a display terminal 503.

[0121] The command terminal 501 can be an industrial touch screen command terminal or a portable command tablet. The touch screen response time is ≤0.1 seconds; the tablet battery life is ≥8 hours; and it supports multi-touch operation.

[0122] The communication terminal 502 can be a 5G communication module integrated into the command terminal, supporting SA / NSA dual-mode; it can be a portable satellite communication terminal with a deployment time of ≤5 minutes, used for communication during public network outages; or it can be a dispatch radio, vehicle-mounted / handheld, operating on the VHF band, and compliant with fire safety standards. The 5G transmission rate of the communication terminal 502 is ≥1Gbps, and the satellite communication bit error rate is ≤10. -6 The communication distance of the dispatch radio is ≥5km.

[0123] The display terminal 503 can be a first-view display for drones, used to view fire scene images in real time, with a first-view display latency of ≤100ms; or it can be a multi-screen splicing screen (4×2 splicing, used to display CIM models / fire data), with a splicing screen brightness of ≥500cd / ㎡ and a contrast ratio of ≥3000:1.

[0124] Command terminal 501 is wired to communication terminal 502. For example, the 5G communication module is integrated on the motherboard of command terminal 501, with a built-in omnidirectional antenna. Command terminal 501 receives first-person view video from the drone and decision-making plans from the data center through the 5G module. Alternatively, a portable command tablet can wirelessly connect to a satellite communication terminal, for example via Bluetooth 5.0, with a pairing distance ≤10m and AES-256 encryption. In the event of a public network outage, the tablet sends a "plan adjustment" request to the data center via the satellite terminal. Another example is a vehicle-mounted dispatch radio connected to the central control unit of a ground fire truck via a wired connection, for example, directly connecting to the fire truck's radio host via a dedicated RF cable with an SMA interface. The dispatch radio sends assembly point and water supply instructions to the ground fire truck to ensure coordinated vehicle operations.

[0125] Display terminal 503 is connected to the drone via data transmission, for example, the drone's first-view display is connected to the visible light camera of the reconnaissance drone via a 5G wireless link, SA standalone networking, RTSP video protocol, and 1080P@60fps resolution. Display terminal 503 plays the drone's first-view footage in real time, allowing commanders to directly observe the fire situation.

[0126] The display terminal 503 is also connected to the city information modeling equipment 30 via data connection, for example, a multi-screen splicing screen and a 3D rendering node, using an HDMI 2.1 interface, 4K@60Hz video format, and a latency of ≤10ms. The multi-screen splicing screen displays an overlaid CIM 3D model of fire / personnel locations for global situational awareness.

[0127] Command terminal 501 connects to the city's fire data center 40 via an encrypted dedicated line, for example, to AI computing nodes. The data protocol is JSON-RPC. The data content includes rescue routes and resource scheduling tables. Command terminal 501 receives decision plans generated by AI nodes and supports commanders to make handwritten annotations and adjustments, such as modifying drone delivery points.

[0128] The portable command tablet and handheld Beidou terminal are connected via wireless command signals, such as the UHF band, to control evacuation / standby / rescue completion. Triggering methods include button presses and voice confirmation. The portable command tablet sends evacuation commands to firefighters, especially when personnel enter a danger zone (marked in the CIM).

[0129] The dispatch radio is wirelessly connected to the ambulance signal control system, enabling real-time voice communication and sending digital commands via VHF digital signals. These digital commands include the coordinates of the injured person's receiving point, the dispatch radio's transmission of the injured person's location and number to the ambulance, ensuring that medical resources are ready to be deployed in advance.

[0130] The command terminal 501 is connected to the UAV swarm control system (cluster control terminal) via signal control, for example, through an Ethernet connection. The command format is XML (including "execute / pause" commands and target priority). After the commander confirms the decision plan, the terminal sends an "execute firefighting" command to the cluster control terminal, triggering the UAV swarm action.

[0131] Furthermore, the system also includes communication support equipment, energy support equipment, and security early warning equipment.

[0132] Communication support equipment can be temporary 5G base stations, portable, deployed within 1km of the fire scene, supporting access for 100 users; or it can be drone-mounted Mesh self-organizing network modules, integrated into each drone, supporting multiple drones.

[0133] Energy backup equipment can be portable photovoltaic charging piles containing solar panels and energy storage batteries for drone charging; or emergency power supply boxes for powering UWB base stations / laser warning lines.

[0134] Safety warning devices can be laser warning lines deployed at building entrances to detect human approach; they can be audible and visual alarms, drone-borne or ground-based, to warn unauthorized personnel; or they can be emergency lighting devices deployed in fire exits with a high-intensity light mode of ≥5000 lumens.

[0135] Temporary 5G base stations are wirelessly connected to drone-borne Mesh self-organizing network modules, providing backbone links for the Mesh network and improving data transmission rates between drones.

[0136] The photovoltaic charging pile is wired to the emergency power box via AC cable, and the power is supplied in parallel. The emergency power box replenishes the photovoltaic charging pile with power, ensuring that drones can be charged even on cloudy days or at night.

[0137] The drones connect to each other via an onboard Mesh self-organizing network module, using a Mesh wireless link (multi-hop transmission) and the 802.11s Mesh standard protocol. The data content includes fire heat maps and obstacle avoidance information. For example, a reconnaissance drone transmits a real-time heat map to a firefighting drone, guiding the firefighting drone to accurately locate the core of the fire.

[0138] Temporary 5G base stations are connected to multi-source data fusion gateways via, for example, fiber optic links, single-mode, with a transmission distance of ≥10km, encrypted IPsec transmission protocol, and bandwidth of ≥1Gbps. After the temporary 5G base stations aggregate drone / personnel data, they transmit it at high speed to the multi-source data fusion gateway to avoid data congestion.

[0139] Emergency lighting equipment is connected to IoT data access terminals via, for example, a LoRa wireless link. The data protocol is LoRaWAN, and the data content includes lighting status (on / off) and remaining power. The status of fire escape lighting is fed back to the CIM system, and a "maintenance" warning is triggered if the lighting malfunctions.

[0140] The photovoltaic charging pile is connected to the signal control of heavy-duty drones, for example, by a wireless digital signal in the 433MHz band. The commands are charging request / charging complete / stop charging. After the drone lands on the charging pile, it sends a "charging request". The charging pile starts charging after confirming that the interface is compatible.

[0141] As the spatiotemporal reference center of the system, Beidou positioning equipment solves the pain point of insufficient positioning accuracy in complex environments (high-rise obstruction, no satellite signal indoors, electromagnetic interference), and provides "centimeter-level outdoors + sub-centimeter-level indoor" unified positioning and timing services for drone swarms, rescue personnel, and ground fire trucks, ensuring that the spatiotemporal coordinates of all fire-fighting elements are consistent and avoiding rescue errors caused by positioning deviations (such as drone collisions or personnel accidentally entering dangerous areas).

[0142] The Beidou-3 three-band receiver (B1C / B2a / B3I) receives raw satellite observation data and suppresses ionospheric delay and building obstruction interference through multi-band fusion, achieving a basic positioning accuracy of 1cm+1ppm horizontally and 2cm+1ppm vertically.

[0143] Fixed / mobile BeiDou differential base stations (coverage radius ≥ 5km) calculate satellite orbit errors and atmospheric delays in real time, generate RTCM3.3 format differential correction signals, and broadcast them to surrounding receivers via the B1C / B2a frequency band, improving the receiver positioning accuracy from "meter level" to "centimeter level".

[0144] Wall-mounted / airdropped UWB positioning base stations (3.1-10.6GHz band) measure the distance to UWB tags using the TOF (Time of Flight) algorithm. A single base station has a coverage radius of ≤50m, and after networking multiple base stations, triangulation positioning achieves an accuracy of ≤3cm. Wearable / device-adhesive UWB tags (weighing ≤50g) send location signals to the base station 10 times per second, solving the indoor positioning blind spots where BeiDou signals cannot penetrate walls.

[0145] The UAV-borne / personnel-handheld IMU module (drift rate ≤0.1° / h) is linked with the Beidou / UWB receiver via the I2C interface. When the Beidou signal signal-to-noise ratio is <10dB (electromagnetic interference) or the UWB base station fails, it automatically switches to inertial navigation and maintains positioning continuity through accelerometers and gyroscopes to avoid positioning interruption.

[0146] As a low-altitude execution carrier of the system, the drone swarm equipment solves the pain points of traditional fire fighting, such as "inaccessibility to ground forces, difficulty in covering high-rise buildings, and difficulty in indoor reconnaissance". Through functional division of labor (scanning / breaking windows / fire extinguishing / reconnaissance / material delivery), it realizes "air-ground integration" of fire detection, fire suppression and rescue support, while providing real-time dynamic fire data for CIM system and data center.

[0147] The heavy-duty drone, with a payload of 5kg, is equipped with a 48MP five-lens tilt camera (5 lenses for multi-angle shooting) and a ±2cm precision solid-state LiDAR. It collects point clouds (density ≥300 points / ㎡) of building facades via a serpentine flight path (100m height, 5m spacing). At the same time, it delivers rapid descent equipment via a pneumatic ejection pod.

[0148] Lightweight drone with a payload of 2kg, integrating a wall-penetrating radar with a detection depth of ≥50cm and an ultrasonic sensor with a radius of 10m. After entering the room through the window, it scans the fire escape / refuge floor along a spiral path (inter-floor dwell point ≤2m). The ultrasonic sensor detects obstacles in real time and triggers obstacle avoidance.

[0149] The window-breaking drone, a fixed-wing type, uses a window-breaking projectile launching module (single projectile explosive force ≥5MPa) to link with the fire control system via PWM signals. After receiving the window-breaking coordinates, it directionally blasts high-rise glass curtain walls (error ≤3cm).

[0150] The firefighting drone, with its multi-rotor and dry powder spraying device (capacity 10L, pressure ≥0.8MPa), is connected to the flight controller via a CAN bus. It accurately delivers dry powder according to the firefighting coordinates, with a coverage area of ​​≤5m.

[0151] The reconnaissance drone is miniature and equipped with a 640×512 resolution infrared thermal imager (temperature measurement from -20 to 1500℃) and a 0-1000ppm gas sensor to detect the combustion core and concentration of toxic gases in confined spaces (such as elevator shafts).

[0152] Infrared thermal imagers remove smoke interference through median filtering and extract thermal radiation intensity for combustion core identification; solid-state LiDAR generates effective point clouds by removing redundant points, and calculates building structure deviations after registration with the CIMBIM model; gas sensors are calibrated with ±5%FS accuracy, and output valid data after eliminating outliers (such as concentrations >1000ppm, which are judged as sensor malfunctions).

[0153] The urban CIM system hardware serves as the digital twin foundation of the system, addressing the pain points of traditional fire protection such as "lack of spatial benchmarks and disconnection of building information." It integrates static 3D data of urban buildings (BIM / pipeline / topography) with dynamic scanning data from drones to construct a high-fidelity digital environment for fire scenes, providing spatial support for positioning calibration, path planning, and structural damage assessment.

[0154] The BIM model storage node (capacity ≥ 100TB) imports the building BIM model in Revit / IFC format and labels core attributes (coordinates of load-bearing walls, width of fire lanes ≥ 1.2m, ignition point of building materials); the spatial index server constructs a spatial index of the building structure using R-tree indexing technology to ensure that the retrieval response time for key data such as fire lanes and refuge floors is ≤ 100ms.

[0155] The LiDAR point cloud storage device (read / write rate ≥1GB / s) receives the UAV LiDAR point cloud, registers it with the BIM model through the ICP (Iterative Closest Point) algorithm, and calculates structural deviations (such as the width of cracks in load-bearing walls); the IoT data access terminal (RS485 / Ethernet conversion) receives data from smoke sensors and emergency lighting equipment, and updates the "equipment status" attributes of the CIM model (such as lighting failure and smoke alarm).

[0156] The 3D rendering node (equipped with RTX4090) integrates static BIM with dynamic point cloud to generate a 3D visualization scene with over 60 FPS, allowing commanders to intuitively view building structures and fire situation. When outputting data to external devices, coordinate transformation (unified to CGCS2000 coordinate system) ensures alignment with BeiDou / UWB positioning data.

[0157] As the intelligent decision-making hub of the system, the urban fire data center addresses the pain points of traditional fire protection, such as "isolated data and decision-making relying on experience." By integrating multi-source data (BeiDou positioning / CIM / BIM / drone sensors) and AI algorithms (fire prediction / path planning / personnel identification), it generates accurate and real-time rescue plans, providing data-driven decision support for the fire dispatch platform.

[0158] The edge server (Intel Xeon Gold 6426R CPU, 128GB DDR5) preprocesses drone sensor data: infrared thermal image defogging, LiDAR point cloud noise reduction, and removal of outliers in gas concentration, with an end-to-end latency of ≤50ms.

[0159] The multi-source data fusion gateway (supporting ≥100 access channels) integrates BeiDou positioning, CIMBIM, and UAV data through protocol conversion (MQTT / NMEA0183 / HTTP) to achieve spatiotemporal alignment (time synchronization error ≤1ms, coordinates are unified to CGCS2000).

[0160] AI computing nodes are used for fire spread prediction, personnel identification, and path planning.

[0161] The real-time database server (Redis, read / write throughput ≥ 100,000 times / second) stores drone status, personnel location, and fire data, supporting millisecond-level retrieval; the historical fire database server (MySQL, capacity ≥ 50TB) stores rescue logs and historical cases, and synchronizes real-time data through MySQLReplication for AI model training and contingency plan optimization.

[0162] As the central hub for managing drone swarms, drone management and control hardware addresses the pain points of traditional drone firefighting, such as "chaotic airspace and poor multi-drone coordination." Through airspace planning (avoiding no-fly zones / collisions), swarm task allocation (scanning / firefighting / delivery), and real-time status monitoring, it ensures that drone swarms can safely and efficiently carry out rescue missions.

[0163] The airspace management terminal (15-inch industrial touch screen, response time ≤0.5 seconds) obtains civil aviation no-fly zone data through the civil aviation airspace docking terminal (4G link, data update ≥1 time / minute) and generates "flight height restriction (e.g., 50m above the building)" in combination with the CIM building height.

[0164] The flight path planning computer (high-performance desktop computer, calculation time ≤ 3 seconds / unit) is based on the A* algorithm, integrates CIM obstacle data (load-bearing walls / pipes) and Beidou positioning reference to generate serpentine / spiral flight paths for UAVs, ensuring that the flight path redundancy coefficient is ≤ 1.15.

[0165] The cluster control terminal (4-screen console, supporting the management of ≥20 drones) issues task commands to the drones through the command transmission module (coverage radius ≥1km, latency ≤100ms): window breaking drone (blasting coordinates / latency), fire extinguishing drone (spray volume / coordinates), and material drone (delivery point).

[0166] The drone status monitor (sampling frequency ≥1Hz) collects real-time data on the drone's battery voltage / current and sensor status. It reads the remaining battery power through the XT60 interface and triggers an "automatic return" command when the remaining power is <15%.

[0167] Time synchronization between drones is achieved through a mesh self-organizing network (BeiDou time synchronization, error ≤1ms), avoiding overlapping scans by multiple drones;

[0168] The scanning area is allocated based on the distributed PSO algorithm, maximizing information entropy (complete coverage) and minimizing scanning time.

[0169] As the system's operational command terminal, the fire dispatch command terminal addresses the pain points of traditional fire services, such as "dispersed command and low coordination efficiency." It integrates functions such as visualized decision-making schemes, one-click linkage of multiple devices, and real-time situation adjustment to achieve collaborative command among "commanders, drones, ground resources, and medical forces," ensuring that rescue orders are implemented rapidly.

[0170] The industrial touch screen command terminal (21.5-inch, IP65 waterproof and dustproof) receives decision-making solutions from the data center and supports handwritten annotations (such as modifying drone delivery points). The annotations are synchronized to the data center via the JSON-RPC protocol.

[0171] The portable command tablet (10.1-inch, battery life ≥8 hours) supports mobile command and connects to a satellite communication terminal via Bluetooth 5.0 (deployment time ≤5 minutes). In the event of a public network outage, communication is maintained via a satellite link (bit error rate ≤10%). -6 ).

[0172] The drone's first-person view display (4K resolution, latency ≤100ms) receives visible light video from the reconnaissance drone via the RTSP protocol, allowing commanders to visually observe details of the fire scene (such as the location of trapped personnel).

[0173] Multi-screen splicing (4×2 splicing, brightness ≥500cd / ㎡) overlays CIM 3D model, Beidou positioning heat map, and fire prediction range to form a two-layer display of "global situation - local details".

[0174] The dispatch radio (VHF band, communication distance ≥ 5km) sends dispatch instructions (assembly point / patient receiving point) to ground fire trucks and ambulances via DTMF signal (dual tone multi-frequency);

[0175] One-click linkage function: After the commander clicks the "Execute Rescue" button, the terminal simultaneously sends trigger signals to the drone swarm control, ground dispatch and medical standby system, with a linkage response time of ≤1 second.

[0176] Communication and auxiliary equipment serve as the system's support system, addressing the pain points of traditional fire protection such as "communication interruption, insufficient energy, and lack of safety warnings." Through multi-link communication (5G + satellite + Mesh), data / command transmission is guaranteed; through energy supply (photovoltaic charging / emergency power), equipment endurance is maintained; and through safety warnings (laser warning lines / audio-visual alarms), unauthorized personnel are prevented from entering, ensuring the system's stable operation in extreme environments.

[0177] Temporary 5G base stations (portable, bandwidth ≥1Gbps) are deployed within 1km of the fire site, supporting access for 100 users. They are connected to the data center via fiber optic links, with 5G prioritizing the transmission of drone video and location data.

[0178] The drone carries a Mesh self-organizing network module (supporting the 802.11s protocol) to enable multi-hop transmission between drones. When the 5G signal is weak (such as when it is blocked by high-rise buildings), the Mesh relay ensures that the data is not interrupted.

[0179] The portable satellite communication terminal automatically switches to Ku-band data transmission when the public network is interrupted (bit error rate ≤10). -6 To maintain communication between the command terminal and the data center.

[0180] The photovoltaic charging pile (power ≥600W, energy storage battery ≥100Ah) is powered by both solar panels and energy storage batteries. When the drone lands, it automatically connects via a Type-C interface (positioning error ≤5cm). A single charge provides 30 minutes of flight time.

[0181] The emergency power supply box (220V AC output, battery life ≥8h) supplies power to UWB base stations and laser warning lines via 2.5 square millimeter copper wire and supports leakage protection (trip response time ≤0.1 seconds).

[0182] The laser warning line device (deployed at building entrances) forms a warning line using a 650nm red laser. The sensing response time is ≤0.1 seconds, and an audible and visual alarm is triggered when the line is blocked for ≥1 second.

[0183] The audible and visual alarm (80dB buzzer + 500-meter video flashing light) is divided into drone-borne and ground-based fixed types. The flashing light is activated when the drone takes off, and the ground alarm is triggered by a laser warning line.

[0184] Next, refer to Figure 2 To explain Figure 1 The low-altitude fire-fighting method of the low-altitude fire-fighting system 100 shown. Figure 2 This is a schematic flowchart of an application method of a low-altitude fire protection system according to an embodiment of the present invention.

[0185] like Figure 2As shown, the low-altitude firefighting method includes: completing equipment self-inspection and spatiotemporal calibration, and establishing a unified centimeter-level positioning and millisecond-level timing reference (S10); collecting point cloud data of building facades and building interior structural data, transmitting the scanned data to the CIM system for point cloud registration and model updating, and generating a digital twin model of the fire scene (S20); integrating Beidou positioning data, CIM model, and UAV sensor data, and generating a firefighting decision-making plan including fire prediction, personnel positioning, path planning, and resource scheduling through AI algorithms (S30); receiving the firefighting decision-making plan, triggering coordinated rescue by UAVs, fire trucks, and personnel, providing real-time feedback on the execution status, and dynamically adjusting instructions (S40); collecting full-process data including equipment status, rescue logs, fire data, and the execution effect of the decision-making plan, evaluating the accuracy of the CIM model, Beidou positioning accuracy, UWB positioning accuracy, and AI decision-making accuracy, optimizing equipment parameters and algorithm models, and updating the historical database and rescue plan (S50).

[0186] like Figure 3 As shown, some embodiments provide step S10, which includes: performing hardware status detection on core equipment such as Beidou receivers, UWB base stations, UAV flight controllers, and CIM servers, troubleshooting, and initiating backups (S101); deploying Beidou differential base stations in open areas within 5km of the fire scene, receiving satellite data to generate differential parameters, broadcasting them to each positioning terminal, and establishing an outdoor centimeter-level positioning reference (S102); deploying UWB base stations through a combination of UAV airdrop and manual deployment to build an indoor positioning network, calibrating UWB coordinates using Beidou differential parameters, and establishing an indoor sub-centimeter-level positioning reference (S103); integrating Beidou differential parameters, UWB conversion parameters, and CIM model coordinates, unifying the time and coordinates of all devices, generating an initial CIM calibration model, and completing system initialization (S104).

[0187] like Figure 4As shown, some embodiments provide step S101, which includes: receiving device parameter data and generating a device status signal. The device status signal includes a high-level signal and a low-level signal. The high-level signal indicates that the device is in a normal state, and the low-level signal indicates that the device is in a fault state. The device includes a Beidou receiver, a UWB base station, a UAV flight controller, a CIM server, and communication equipment. The device parameter data includes the Beidou receiver frequency band, the UWB base station coverage radius, and the UAV's maximum speed (S111). If the status signal of at least one of the Beidou receiver, UWB base station, UAV flight controller, and CIM server is detected to be a low-level signal, a fault handling step is triggered, including self-testing and repairing the faulty device or replacing it with a backup device, and re-performing the device self-test after repair (S121). If the status signal of the communication equipment is detected to be a low-level signal, a satellite communication switching step is triggered (S131). The device parameter data is simultaneously stored in the real-time database of the urban fire protection data center (S141). The device parameter data includes the Beidou receiver frequency band, the UWB base station coverage radius, and the UAV speed. These data are derived from the device's factory parameter configuration file and the CIM system device library.

[0188] For example, after each device starts up, it sends a hardware ready signal to the "system control center." A high level (1) indicates that the device is in normal condition, while a low level (0) indicates that the device is in a fault state. If the Beidou receiver can only receive two frequency bands, a fault handling step is triggered, which involves device self-testing or device replacement, such as replacing it with a three-band receiver or starting the drone's backup battery, and re-testing until all signals are high. If all device status signals are high, the candidate step is then performed.

[0189] like Figure 5 As shown, some embodiments provide step S102, which includes: if all device status signals are detected to be high-level signals, then sending a base station deployment start command to the BeiDou differential base station deployment team (S112); deploying the BeiDou differential base station according to the received environmental basic data and receiving the original satellite observation data (S122); generating BeiDou differential correction parameters based on device parameter data using the differential correction formula Δ=ρ^-ρtrue, where Δ is the differential correction amount, ρ^ is the satellite distance measured by the base station, and ρtrue is the true distance based on known coordinates and ephemeris, used to correct receiver positioning errors (S132); outputting the BeiDou differential correction parameters and base station deployment coordinates, broadcasting differential signals to drones, fire trucks, and personnel terminals to establish an outdoor positioning reference (S142); if the satellite signal-to-noise ratio is detected to be <10dB, then outputting an anti-interference warning signal and synchronously triggering satellite communication backup of communication and auxiliary equipment to ensure that the differential signal is not interrupted (S152). The environmental data includes the latitude and longitude of the fire scene, the location coordinates of the outdoor area, and the coordinates of the indoor building entrance. The location coordinates of the outdoor area are used to deploy BeiDou differential base stations. This data comes from the fire alarm receiving platform and handheld BeiDou terminals.

[0190] like Figure 6 As shown, step S103 in some embodiments includes: acquiring BeiDou differential correction parameters and basic environmental data, and sending a command to the UAV control terminal to deploy a UWB base station (S113); controlling the spacing between adjacent base stations based on equipment parameter data, automatically forming a network through the UWB frequency band, and generating UWB base station coordinates (S123); outputting UWB coordinate transformation parameters and UWB base station status signals (S133); if the UWB base station positioning error is >3cm, calling the BeiDou differential correction parameters for recalibration until the error is ≤3cm, and then outputting the UWB coordinate transformation parameters (S143). The UWB coordinate transformation parameters are used to transform the UWB coordinate system into the CGCS2000 coordinate system.

[0191] like Figure 7 As shown, step S104 provided in some embodiments includes: unifying the time and coordinates of all devices according to the BeiDou differential correction parameters, UWB coordinate transformation parameters and CIM model (S114); aligning the coordinates, BeiDou coordinates and UWB coordinates of the CIM model to generate a CIM initial calibration model containing positioning calibration marks (S124); outputting a spatiotemporal reference dataset containing BeiDou differential parameters, UWB transformation parameters, time synchronization signals and CIM initial calibration model (S134); and simultaneously outputting a system initialization completion signal (S144).

[0192] According to step S10, a three-in-one benchmark system of positioning, time and space is established to solve the problems of traditional fire positioning deviation, time asynchrony and coordinate inconsistency. At the same time, spatiotemporal benchmark data and CIM initial model are output to provide core data support for subsequent steps and ensure the spatiotemporal consistency of data throughout the process.

[0193] like Figure 8 As shown, some embodiments provide step S20, which includes: combining UAV parameters, CIM model and no-fly zone data to calculate the width of the safety isolation zone and generate the external serpentine and internal spiral flight paths of the UAV (S201); scheduling a heavy UAV to fly along a compliant flight path, activating the camera and LiDAR to collect point cloud and texture data of the building facade, and generating external structural data of the building after denoising (S202); scheduling a light UAV to enter the building through the building window, activating the radar and sensors to collect structural data of fire escape routes and refuge floors, calculating structural priority, and generating internal structural data of the building (S203); transmitting the external structural data and internal structural data of the building scanned by the UAV to the CIM system, registering the model through the ICP algorithm, marking structural damage, and generating a CIM model containing dynamic damage (S204).

[0194] like Figure 9As shown, some embodiments provide step S201, which includes: calling the UAV's maximum speed parameters, combining the pre-approved obstacle avoidance response time and sensor coverage radius to determine the UAV isolation zone width (S211); receiving the spatiotemporal reference dataset, the CIM initial calibration model, and civil aviation no-fly zone data, and generating a flight path coordinate file containing UAV safe flight path information based on the UAV isolation zone width (S221); if interference between the UAV safe flight path and the no-fly zone is detected, reconfirming the UAV isolation zone width and generating an adjusted flight path coordinate file (S231); if the UAV density is detected to be higher than a preset density, outputting a UAV waiting signal (S241); if the UAV density is detected to be not higher than a preset density, triggering a UAV scan start signal (S251). The formula for determining the UAV isolation zone width is S... buffer =V max ×t react +R sensor S buffer For the width of the buffer zone for moving objects, for heavy drones, its V max 15 m / s, t react =0.2s, R sensor =20m, ensuring no risk of collision between the drone and moving objects. In some embodiments, the preset density is 5 per 100m. 2 .

[0195] like Figure 10 As shown, some embodiments provide step S202, which includes: receiving a scan start signal, a flight path coordinate file, and BeiDou differential correction parameters, and generating a flight path command for the UAV (S212); collecting point cloud data and texture data of the building facade (S222); denoising the point cloud data and texture data of the building facade to generate a point cloud of the effective facade, and stitching the texture data to generate a facade texture map (S232); if the point cloud density of the building facade is detected to be not less than 300 points / ㎡, outputting a building facade scan completion signal and transmitting the point cloud data and facade texture map of the effective facade to the city CIM device (S242); if the point cloud density of the building facade is detected to be less than 300 points / ㎡, outputting a supplementary scan command to control the UAV to adjust the flight path and rescan the building facade (S252); and synchronously outputting the external scan status of the UAV to the flight management platform, wherein the external scan status of the UAV includes the remaining battery power and sensor working status (S262).

[0196] like Figure 11As shown, step S203 in some embodiments includes: receiving an external scan completion signal, UWB coordinate transformation parameters, and indoor entrance coordinates from the initial CIM data; generating control commands for the lightweight UAV to enter the building; and activating the through-wall radar and ultrasonic sensor (S213); performing a spiral scan according to the scan task parameters; collecting data on the width of the fire escape route and the structure of the refuge floor; and determining the indoor structure recognition priority based on LiDAR integrity, infrared temperature difference, and radar depth (S223); outputting the indoor structure dataset and the internal scan completion signal, wherein the indoor structure dataset includes the passage width, refuge floor model, and structure priority (S233); and generating an obstacle avoidance trigger signal if the UAV's ultrasonic sensor detects an obstacle to control the UAV to adjust its attitude, and continuing scanning after obstacle avoidance (S243). Wherein, the indoor structure recognition priority Ps = αP LiDAR +βΔT+γD,P LiDAR ΔT and D are LiDAR integrity, infrared temperature difference and radar depth, respectively, and α, β and γ are LiDAR integrity weight, infrared temperature difference weight and radar depth weight, respectively. For example, for an indoor load-bearing wall, α = 0.3, β = 0.4 and γ = 0.3.

[0197] like Figure 12 As shown, step S204 provided in some embodiments includes: registering the effective facade point cloud and interior structure dataset with the CIM model and calculating the structural deviation (S214); if the structural deviation is >10cm, marking it as a high-risk structure and updating the structural damage attributes of the CIM model (S224); if the structural deviation is ≤10cm, only updating the texture data of the CIM model (S234); outputting the updated CIM model and the model update completion signal (S244); and simultaneously storing the effective facade point cloud, interior structure dataset, and CIM LiDAR point cloud storage device (S254).

[0198] Step S20 completes a full-dimensional scan of the building's external and internal structures, generating structural and sensor data with centimeter-level accuracy, thus solving the problem of blind spots in information within traditional fire-fighting buildings. Simultaneously, the output CIM model and scanned data provide spatial and dynamic dual-dimensional data support for intelligent decision-making, ensuring that the decision-making plan aligns with the actual fire situation.

[0199] like Figure 13 As shown, some embodiments provide steps S30 including: spatiotemporal alignment and preprocessing of multi-source data (S301); fire spread prediction (S302); location and status assessment of trapped personnel (S303); rescue route planning and resource scheduling (S304).

[0200] like Figure 14As shown, some embodiments provide step S301, which includes: calling the time synchronization signal and UWB coordinate transformation parameters to unify the UWB personnel coordinates, Beidou UAV coordinates, and CIM model coordinates into the same spatiotemporal system (S311); preprocessing the positioning data, UAV sensor data, and CIM dynamic data to output a spatiotemporally aligned dataset (S321); synchronously distributing the spatiotemporally aligned dataset to the fire prediction module and personnel positioning module of the AI ​​computing node (S331); and outputting a data replenishment command if the data missing rate of the spatiotemporally aligned dataset is >5% (S341). During data preprocessing, for example, outliers with gas concentrations exceeding 1000 ppm are removed, linear interpolation is performed, dehazing of infrared thermal images is performed using a degradation-restoration model, and smoothing filtering of positioning data is performed using a sliding window. The data replenishment command can schedule the reconnaissance UAV to re-collect data until the data missing rate of the spatiotemporally aligned dataset is ≤5%.

[0201] like Figure 15 As shown, step S302 provided in some embodiments includes: receiving a preprocessed dataset containing infrared thermal images, building material parameters of a CIM dynamic model, and fire case data of similar buildings in a historical fire database from the fire prediction module of the AI ​​computing node, and determining the heat flux change rate (S312); generating fire spread information based on the heat flux change rate and the thermal diffusivity of building materials, the fire spread information including a fire spread trend map and fire spread coverage speed information, determining the future fire coverage radius based on the fire spread information, overlaying CIM high-risk structure data, and marking collapse risk information, the collapse risk information including the coordinates of the collapse risk area and the collapse risk level (S322); integrating the fire spread information trend map and the coordinates of the collapse risk area to generate a fire prediction result containing a visual chart; determining the fire warning level based on the fire spread speed and the collapse risk level, the fire warning level being used to trigger the fire dispatch command terminal to generate a corresponding alarm signal (S332).

[0202] Among them, the rate of change of heat flux is based on the formula It is confirmed that, among them, ρ represents the rate of change of heat flux, with positive values ​​indicating increased fire intensity and negative values ​​indicating decreased fire intensity; κ is the thermal diffusivity of the material, for example, 0.12 W / (m·K) for wood; f uel q is the density of the combustible material. comb ν represents the heat of combustion, and v is the wind speed vector.

[0203] Among them, according to the fire coverage radius based on R impact =τ is determined; R impact Let τ be the radius of the fire's coverage, and τ be the time variable.

[0204] The fire warning level can be displayed using different colors, such as red for extreme risk, orange for high risk, yellow for medium risk, and blue for low risk.

[0205] like Figure 16 As shown, step S303 provided in some embodiments includes: the personnel positioning module of the AI ​​computing node receives a preprocessed dataset containing RGB-D camera images, infrared thermal images, and UWB indoor positioning data; loads a lightweight YOLOv7-Tiny model; performs inference on the image data; detects personnel target boxes; and outputs personnel pixel coordinates and inference confidence (S313); the personnel pixel coordinates, combined with the UAV's BeiDou positioning data and camera intrinsic parameters, are converted into personnel spatial coordinates in the CGCS2000 coordinate system using a pixel coordinate to spatial coordinate formula; and then the extended Kalman filter is used... The wave formula integrates IMU data and visual data to optimize the positioning accuracy of personnel spatial coordinates (S323); combining the optimized personnel spatial coordinates, CO concentration data and infrared body temperature data in the preprocessed data, personnel priority is marked according to preset standards to generate a personnel positioning and priority table (S333); comparing the personnel positioning and priority table with the initial personnel statistics list, if a new personnel not in the initial list is detected in the personnel positioning and priority table, an emergency addition signal is sent to the city fire data center and fire dispatch command terminal (S343) so as to notify subsequent resource scheduling to give priority to the new personnel.

[0206] The Kalman filter formula is: The personnel target bounding box was optimized and the positioning accuracy was improved to ±0.3m.

[0207] F represents the optimal coordinates of the personnel at time k. k Let B be the state transition matrix. k u k To control the quantity (step size), K k For Kalman gain, z k These are sensor observations (e.g., infrared sensor observations) used to fuse multi-source positioning data and reduce errors.

[0208] Personnel priority: Level 1 (seriously injured, body temperature > 38℃ or CO concentration > 800ppm), Level 2 (emergency, located within 10m of the predicted fire coverage radius of S322), Level 3 (normal, other situations).

[0209] like Figure 17As shown, some embodiments provide step S304, which includes: the path planning module receiving fire prediction results, personnel location and priority table, and CIM dynamic model, and generating personnel evacuation paths and drone firefighting paths respectively using the improved A* algorithm and path cost function (S314); the resource scheduling module receiving path coordinates, personnel location and priority table, and drone status report, and generating resource scheduling table using target priority formula (S324); integrating fire prediction results, personnel location and priority table, path coordinates, and resource scheduling table to generate a fire-fighting decision-making scheme including a visualized fire map, path map, and resource distribution map; extracting drone mission parameters from resource scheduling results to control the fire-fighting drone's fire-fighting coordinates and deployment quantity, the window-breaking drone's blasting coordinates and delay, and the material drone's delivery point and material type (S334); extracting the three-dimensional coordinates of high-risk areas from the collapse risk area coordinates to generate a risk warning signal, which includes risk area coordinates, risk level, and avoidance suggestions, and synchronously sending it to the fire dispatch command terminal and the rescue personnel's handheld Beidou terminal via a 5G link (S344).

[0210] The path cost function is f(n) = g(n) + h(n) + λ[w1D(n) + w2T(n)], where g(n) is the actual movement cost, h(n) is the Euclidean distance heuristic, λ = 0.8, D(n) is the population density, w1 = 0.4, T(n) is the hazard weight, which uses risk area data, and w2 = 0.6.

[0211] Through the target priority formula T p =argmin( xi,yi )(α∣Pdrone-Ptarget∣ / Vmax+βR r / C e Assign drones and fire trucks, where α is the distance weight (0.6), β is the urgency weight (0.4), and Rr represents remaining resources. e Prioritize personnel.

[0212] Step S30 breaks down data silos, achieving deep integration of multi-source data from positioning, sensing, CIM, and historical data, thus solving the problem of traditional fire-fighting decision-making relying on experience. Simultaneously, the output AI decision-making solution provides quantitative and precise operational basis for candidate steps, ensuring that rescue resources are allocated to high-priority targets, improving rescue efficiency and safety.

[0213] The fire dispatch platform receives fire-fighting decision-making plans and triggers coordinated execution by a drone swarm, ground fire trucks, and rescue personnel with a single click. Drones complete tasks such as breaking windows, extinguishing fires, and delivering supplies according to mission parameters. Fire trucks arrive at designated locations to provide water and conduct rescue operations. Personnel evacuate or are rescued according to planned routes, with real-time feedback on the execution status and dynamic adjustments to instructions throughout the process.

[0214] Therefore, the application method of the low-altitude fire-fighting system provided by the present invention further includes step S40. For example... Figure 18 As shown, some embodiments provide step S40, which includes: allocating drone tasks according to the fire-fighting decision plan, issuing instructions to control the drone to perform window breaking, fire extinguishing, and material delivery, monitoring the status in real time, and handling abnormal information such as return and loss of contact (S401); ground fire trucks and personnel coordinating the execution (S402); and dynamic instruction adjustment (S403).

[0215] like Figure 19 As shown, some embodiments provide step S401, which includes: the fire dispatch command terminal receiving the fire decision plan and the drone task allocation result, sending a task trigger signal to the cluster control terminal of the drone flight control unit, and starting the drone task execution process (S411); after receiving the task trigger signal, the cluster control terminal, according to the drone task allocation result and drone task parameters, sends a special task instruction to the corresponding drone, the window breaking drone receiving a window breaking task instruction including blasting coordinates and delay time, the fire extinguishing drone receiving a fire extinguishing task instruction including fire extinguishing coordinates and dry powder dosage, and the material drone receiving a material task instruction including delivery point coordinates and material type; after the drone executes the task, it transmits the information via a 5G link. Feedback on task execution status to the cluster control terminal (S421); if the drone's battery level is detected to be <15%, a return-to-home command is sent to the drone, which is used to instruct the drone to return to the nearest photovoltaic charging pile and simultaneously dispatch a backup drone to take over, and output the device replacement log (S431); if the cluster control terminal does not receive positioning data or status feedback from a drone within 5 seconds, a disconnection warning signal is output to the city fire data center and the fire dispatch command terminal, and an inertial navigation start command is sent to the drone to start the drone's onboard inertial navigation backup equipment to maintain positioning; attempt to re-establish communication, and after communication is restored, a task recovery command is sent to the drone to continue executing the original task (S441).

[0216] The group control terminal allocates drones according to task type. Window-breaking drones receive blasting coordinates and delay information, calibrate using BeiDou positioning, fly to the target window, trigger the window-breaking projectile, and output a window-breaking completion signal. Firefighting drones receive firefighting coordinates and the amount of inactivating agent to be delivered, break the window, spray dry powder, and output a firefighting progress signal. Material delivery drones receive delivery coordinates and material type information, and output a material delivery confirmation signal after delivery. Every second, the drones send task execution status information to the group control terminal, such as a fire extinguishing completion rate of 60%, output drone execution logs, and synchronize them to the fire dispatch platform.

[0217] like Figure 20As shown, some embodiments provide step S402, which includes: the fire dispatch command terminal receiving the fire truck dispatch results, encoding the dispatch instructions into DTMF dual-tone multi-frequency signals according to the fire radio standard, and sending the assembly point coordinates and water supply instructions to the vehicle-mounted radio of the ground fire truck (S412); the Beidou terminal receiving the personnel evacuation path coordinates and risk warning signals, and calibrating its own position in real time through the UWB positioning module built into the Beidou terminal; if a deviation from the evacuation path > 5m is detected, the terminal immediately triggers an audible and visual warning and displays a path correction signal; simultaneously, the fire-resistant flying vehicle flies along the evacuation path (S422); the ground fire truck navigates to the designated assembly point through the vehicle-mounted Beidou receiver, and according to the CIM dynamic model... The fire water supply interface is marked with a type, and the fire hose is connected to complete the water supply. After the water supply is completed, a water supply log is sent to the fire dispatch command terminal. After the rescue personnel arrive at the refuge floor or outdoor safe area, the handheld terminal automatically sends an evacuation completion signal. The fire truck water supply log, personnel evacuation completion signal and drone execution log are summarized to generate a rescue full process log, which is synchronized to the fire dispatch command terminal and the city fire data center (S432). When the drone flight control unit detects that the drone fire-fighting mission execution completion rate is ≥90%, the fire dispatch command terminal automatically sends an internal fire attack command to the ground fire truck. This internal fire supply command controls the fire truck to enter the building and cooperate with the drone fire-fighting results (S442).

[0218] Through steps S401 and S402, after the drone completes firefighting, it can trigger the fire attack command inside the fire truck, achieving coordinated air and ground suppression of the fire area.

[0219] like Figure 21 As shown, some embodiments provide step S403, which includes: the fire dispatch command terminal receiving the drone execution log, fire truck status signal, personnel status signal and real-time dynamic information of the fire, and evaluating the extinguishing effect (S413); if the fire spread exceeds the prediction by 10%, a scheme recalculation request is output, which is used to trigger the AI ​​node to recalculate the path and task parameters, and the new parameters are sent to the execution device through the 5G link (S423); if a new seriously injured person is detected, a priority adjustment signal is output, which is used to trigger S401 to reallocate the drone and prioritize the rescue of the newly seriously injured person (S433); if a sudden collapse of the indoor building structure is detected, an emergency evacuation signal is output, which is used to trigger S402 (S443); a dynamic adjustment log containing the adjustment reason, new parameters and execution results is output and incorporated into the rescue process log (S453); when all personnel targets are detected to have been evacuated or rescued and the fire control rate is ≥90%, a rescue mission completion signal is output, triggering closed-loop optimization preparation (S463).

[0220] The effectiveness of the fire extinguishing operation is evaluated using preset assessment indicators, such as the fire control rate = extinguished area / total fire area, and the personnel evacuation rate = evacuated number / total number of trapped people.

[0221] Step S403 enables multi-dimensional collaboration between air, ground, and human resources, translating AI decision-making solutions into concrete rescue actions and addressing the issues of dispersed and weak coordination among traditional firefighting forces. Real-time dynamic adjustments ensure the solutions adapt to changes in the fire scene, and the output of full-process logs and evaluation data provides complete data support for closed-loop optimization, achieving seamless integration of execution and evaluation.

[0222] After the rescue mission is completed, collect full-process data including equipment status, rescue logs, fire data, and the effectiveness of decision-making plans. Evaluate the accuracy of the CIM model, the accuracy of BeiDou / UWB positioning, and the accuracy of AI decision-making. Optimize equipment parameters and algorithm models, update historical databases and rescue plans, and achieve system self-evolution.

[0223] Therefore, the low-altitude firefighting method based on urban CIM and BeiDou positioning provided by this invention also includes step S50. Specifically, as... Figure 22 As shown, step S50 includes: evaluating and optimizing the accuracy of the CIM model (S501); verifying and optimizing the accuracy of BeiDou positioning and UWB positioning (S502); optimizing the AI ​​decision-making algorithm (S503); and optimizing equipment parameters and rescue plans (S504).

[0224] Some embodiments provide step S501, which evaluates the accuracy of the CIM model based on the original scanning data and the results of the on-site survey, analyzes the causes of errors and optimizes the modeling process to improve the accuracy of subsequent models.

[0225] like Figure 23As shown, some embodiments provide step S501, which includes: the urban CIM device retrieves external and internal building structure data from the lidar point cloud storage device, extracts actual building structure data from the on-site survey from the rescue process log, imports the CIM dynamic model and actual structure data into the accuracy assessment tool, and calculates the spatial matching degree between the model and the actual building using the intersection-union ratio formula (S511); the CIM modeling accuracy is determined based on the intersection-union ratio (S521); if the intersection-union ratio is detected to be ≥90%, a CIM accuracy compliance report is output (S531); if the intersection-union ratio is detected to be <90%, the cause of the error is analyzed by comparing the original scan data and the model data, and targeted CIM optimization suggestions are output, and the suggestions are converted into executable CIM optimization parameters (S541); the CIM accuracy assessment report, CIM optimization suggestions, and CIM optimization parameters are stored in the historical fire database of the urban fire data center through the database interface to establish a historical archive of CIM accuracy assessment; at the same time, a CIM parameter update command is sent to the management terminal of the urban CIM device to write the CIM optimization parameters into the configuration file of the CIM modeling system (S551).

[0226] The Intersection over Union (IoU) formula is: IoU = 100% × (Model-to-actual overlap volume) / (Model-to-actual merged volume).

[0227] like Figure 24 As shown, step S502 provided in some embodiments includes: the urban fire data center extracts BeiDou positioning-related data from the rescue process log, calculates the BeiDou positioning track offset standard deviation using the track offset standard deviation formula; simultaneously extracts UWB indoor positioning logs, combines them with the measured indoor coordinates on site, and statistically analyzes the UWB indoor positioning error (S512); determines the optimization direction based on the positioning accuracy assessment report (S522): if the BeiDou track offset standard deviation is detected to be greater than 1cm, analyzes the cause of the deviation, and outputs BeiDou optimization parameters (S532); if the UWB indoor positioning error is detected to be greater than 3cm, analyzes the cause, and outputs UWB optimization parameters (S542); stores the positioning accuracy assessment report, optimization suggestions, and optimization parameters in the positioning system archive of the historical fire database, compares them with historical assessment data, and analyzes the trend of positioning accuracy changes; simultaneously sends a positioning parameter update command to the management terminal of the BeiDou positioning device and the UWB device, and writes the optimization parameters into the device configuration (S552).

[0228] The formula for the standard deviation of the track offset is σ = , where di is the distance of the i-th sampling point from the planned path, μ is the mean, and n is the number of sampling points.

[0229] like Figure 25As shown, step S503 provided in some embodiments includes: the urban fire data center extracts AI decision error data from the rescue process log; calculates the fire prediction error using the fire prediction error formula; statistically analyzes the deviation between the actual time taken for the rescue path and the predicted time; and simultaneously organizes dynamic adjustment records as supplementary basis for algorithm optimization (S513); if the fire prediction error is detected to be >10%, the κ value of different subdivision materials is added (S523); if the path time deviation is detected to be >20%, the cost function weight of the improved A algorithm is adjusted; at the same time, new logical relationships are extracted from the rescue process log to update the fire knowledge graph (S533); the AI ​​algorithm optimization report, optimization parameters, and updated knowledge graph are stored in the AI ​​algorithm archive of the historical fire database, compared with historical optimization data, and the improvement trend of algorithm accuracy is analyzed; at the same time, an AI parameter update instruction is sent to the AI ​​computing node of the urban fire data center, the optimized algorithm parameters are written into the configuration file of the AI ​​model, and the updated knowledge graph is imported into the inference engine (S543).

[0230] The fire prediction error formula is Error_Q = 100% × |Qpred - Qactual| / Qactual, where Error_Q is the fire prediction error, Qpred is the predicted heat flux, and Qactual is the actual heat flux. If the error is greater than 10%, the model parameters need to be optimized. If Error_Q > 10%, it indicates that the classification of the thermal diffusivity coefficient κ of building materials is not refined enough (e.g., "subdividing 'wood' into 'pine' and 'fir'"). The κ values ​​for different subdivided materials should be supplemented (e.g., pine κ = 0.11 W / (m·K), fir κ = 0.13 W / (m·K)). If the path time deviation is greater than 20%, adjust the weight of the cost function of the improved A algorithm (e.g., increase the personnel density weight w1 from 0.4 to 0.5); at the same time, extract new logical relationships from the "rescue process log" (e.g., "when the wind speed is 5m / s, the hovering stability threshold of the drone is ±0.5m" and "when the steel temperature reaches 600℃, the risk of collapse increases sharply"), update the fire protection knowledge graph, and enrich the decision basis of the algorithm.

[0231] like Figure 26As shown, step S504 provided in some embodiments includes: the urban fire data center integrates CIM optimization parameters, positioning optimization parameters, AI algorithm optimization parameters, and equipment fault records in the rescue process log, and generates equipment parameter optimization results and rescue plan optimization results in two categories. The equipment parameter optimization results include updating the drone sensor parameters, the anti-interference frequency band of the Beidou receiver, and the sensitivity of the UWB base station. The rescue plan optimization results include updating the plan table containing fire type and resource configuration based on the effect of this rescue, and adding dynamic adjustment trigger conditions (S514); the optimized system parameter configuration file is distributed to all core devices through the device management network, and the devices receive... Upon receiving the data, the system automatically updates the local parameter configuration and sends a parameter update confirmation signal to the data center. Simultaneously, it imports the iterated rescue plan into the fire dispatch command terminal's plan library, overwriting the old plan and setting it as the default plan (S524). The optimized system parameter configuration file, the iterated rescue plan, the entire rescue process log, all assessment reports, and optimized data are synchronously stored in the historical fire database via the database backup interface, completing the closed-loop archiving of this rescue data. Furthermore, the system startup configuration is set to automatically load optimized parameters upon next startup, ensuring that the optimized system parameter configuration file is automatically called as the initial equipment parameters during the next system initialization (S534).

[0232] The "optimization parameters" in this step will update the initial configuration in step S10 in reverse, and the "iterative plan" will serve as the new benchmark for decision-making in step S30, forming a self-evolving closed loop of "execution-evaluation-optimization-re-execution". This step establishes a complete data closed loop of rescue, evaluation, and optimization, solving the problem of traditional fire protection "lacking self-evolving capabilities"; through quantitative optimization, it continuously improves the system's positioning accuracy, decision-making accuracy, and equipment reliability, transforming single rescue experience into general system capabilities, and promoting the upgrade of urban fire safety from "passive response" to "proactive precision".

[0233] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0234] Although only some specific embodiments of the invention have been described, those skilled in the art will understand that the invention can be practiced in many other forms without departing from its spirit and scope. Therefore, the examples and embodiments shown are to be considered illustrative rather than restrictive, and the invention may encompass various modifications and substitutions without departing from the spirit and scope of the invention as defined by the appended claims.

Claims

1. A low-altitude fire suppression system based on urban CIM and BeiDou positioning, characterized in that, include: The BeiDou positioning equipment is configured to generate and send BeiDou differential signals and UWB positioning signals, provide indoor and outdoor centimeter-level positioning and unified timing, transmit UAV positioning data and personnel positioning data, receive anti-interference commands, and trigger inertial navigation backup. The drone swarm equipment transmits BeiDou differential signals and UWB positioning signals with the BeiDou positioning equipment. The drone swarm equipment is also configured to collect and send building point clouds, infrared thermal maps and building structure data, receive flight control commands for performing tasks such as breaking windows, fire fighting and material delivery, and send first-person view videos and task execution logs. The city CIM device is connected to the drone swarm device. The city CIM device is configured to receive building structure data sent by the drone swarm device, update the CIM model, mark the building structure and risk areas, and send the BIM model, building material parameters, obstacle data and CIM 3D model. The city's fire protection data center is connected to BeiDou positioning equipment, sending anti-interference trigger commands to it. It also connects to drone swarm equipment, sending drone mission commands to control the drone swarm equipment to perform firefighting, window-breaking, or scanning tasks. Furthermore, it connects to city CIM equipment, receiving static BIM models and dynamic structural damage data from the city CIM equipment. The data center is also configured to integrate BeiDou positioning data, CIM models, building point clouds, infrared thermal maps, and building structural data to achieve spatiotemporal alignment. Based on AI algorithms, it generates fire prevention decision-making schemes, including fire prediction schemes, path planning schemes, and firefighting and rescue schemes. Finally, it receives full-process logs, evaluates accuracy, and outputs equipment and algorithm optimization parameters. The fire dispatch command terminal is connected to the Beidou positioning equipment to receive the location data of rescue personnel sent by the Beidou positioning equipment; the fire dispatch command terminal is connected to the drone swarm equipment to receive the first-person view video sent by the drone swarm equipment and send task dispatch instructions to the drone swarm equipment; the fire dispatch command terminal is connected to the city CIM equipment to receive the CIM 3D model sent by the city CIM equipment; the fire dispatch command terminal is connected to the city fire data center to receive the fire decision plan sent by the city fire data center and send decision adjustment requests to the city fire data center.

2. The system according to claim 1, characterized in that, Beidou positioning equipment includes: The Beidou positioning terminal is connected to the city's fire protection data center and transmits positioning data back to the city's fire protection data center. The Beidou positioning terminal is also configured to receive Beidou differential signals, output centimeter-level positioning data, and receive inertial navigation backup data. The BeiDou differential reference equipment is configured to generate a differential correction signal higher than the signal-to-noise ratio and transmit the differential correction signal to the BeiDou positioning terminal; the BeiDou differential reference equipment is connected to the city fire data center and is configured to detect the satellite signal-to-noise ratio and send the satellite signal-to-noise ratio to the city fire data center, as well as receive anti-interference commands sent by the city. The UWB indoor positioning device transmits UWB positioning signals with the BeiDou positioning terminal; the UWB indoor positioning device is configured to generate indoor positioning signals and send indoor positioning signals to the UAV swarm device; the UWB indoor positioning device is also configured to receive base station coordinate calibration instructions sent by the city CIM device; The inertial navigation backup device is configured to output inertial navigation data to the Beidou positioning terminal. The inertial navigation backup device is also connected to the city fire data center to receive drift calibration commands sent by the city fire data center. The inertial navigation backup device is also connected to the drone swarm device to send drone attitude adjustment data to the drone swarm device.

3. The system according to claim 1, characterized in that, Drone swarm equipment includes: The functional drone connects to the city's CIM (Construction Information Management) equipment, sending building structure data to the equipment. It is also configured to perform external and internal building scans, break windows, extinguish fires, deliver supplies, and transmit information on remaining circuits and mission progress, all according to task instructions. The drone is equipped with sensors configured to collect and preprocess multimodal data, transmit preprocessed multimodal data, and receive sampling frequency adjustment commands. The sensors are fixed to the fuselage of the functional drone and connected to its data. The drone's sensors are also connected to the city's fire protection data center to transmit preprocessed multimodal data. The drone's sensors are also connected to the city's CIM equipment for data storage and backup. The drone actuator is fixed to the fuselage of the functional drone and connected to the signal of the functional drone. It sends action trigger signals to the functional drone to perform actions such as breaking windows, spraying dry powder, and dropping materials. The drone actuator is also connected to the fire dispatch command terminal for data transmission of action execution logs. The drone actuator is also configured to receive execution parameter adjustment instructions. The UAV flight control unit connects to the functional UAV, sending mission commands and providing flight status information. It also connects to the UAV's onboard sensors, sending sampling frequency adjustment commands. Furthermore, it connects to the UAV's actuators, sending execution parameter adjustment commands. The UAV flight control unit also connects to the city's CIM (Construction Information Modeling) equipment, providing building obstacle coordinates. Additionally, it connects to the city's fire data center, receiving mission parameters from the data center. Finally, it connects to the fire dispatch command terminal, receiving adjustment commands from it.

4. The system according to claim 1, characterized in that, The city's fire protection data center includes: The edge server connects to the drone cluster equipment to acquire raw sensor data, preprocess the raw sensor data, and send the preprocessed sensor data. The AI ​​computing device connects to an edge calculator to receive pre-processed sensor data and send pre-processing parameter adjustment instructions; it also connects to urban CIM equipment to acquire building material and building structure data; it is configured to generate AI decision-making schemes and send them to the fire dispatch command terminal; it connects to drone swarm equipment to send drone mission parameters; and it is configured to receive full-process logs to optimize algorithm parameters. The database device is configured to store historical fire data, real-time location data, and sensor data; provide historical cases to AI computing devices to support model training; connect to the city's CIM equipment to back up historical CIM models; and connect to the fire dispatch command terminal to provide historical rescue plans and receive full-process logs sent by the fire dispatch command terminal. The multi-source data fusion gateway connects to BeiDou positioning devices, urban CIM devices, drone swarm devices, edge servers, and AI computing devices. Based on the conversion data protocol, it integrates BeiDou positioning data, CIM models, and drone sensor data to achieve spatiotemporal alignment; and distributes the fused data to edge servers and AI computing devices.

5. The system according to claim 1, characterized in that, The fire dispatch command terminal includes: The command terminal is configured to receive fire-fighting decision plans generated by the city's fire data center and send plan adjustment requests to the city's fire data center, receive execution status feedback signals and send adjustment instructions; the command terminal is also connected to Beidou positioning equipment to obtain the location data of rescue personnel; The communication terminal is configured to establish multi-link communication, 5G, satellite, and Mesh networking; the communication terminal is electrically connected to the command terminal for transmitting adjustment commands and video streams; the communication terminal is also data-connected to the drone swarm equipment to acquire first-person view video from the drones and send emergency evacuation commands; the communication terminal is also configured to receive communication quality data and send link switching suggestions to the command terminal. The display terminal is electrically connected to the command terminal, receives display parameter adjustment instructions from the command terminal, and displays the CIM 3D model, the first-person view of the UAV, the personnel positioning heat map, and the fire prediction map.

6. A low-altitude firefighting method based on the system described in any one of claims 1 to 5, characterized in that, Includes the following steps: Complete equipment self-test and spatiotemporal calibration, and establish a unified centimeter-level positioning and millisecond-level timing reference; Collect point cloud data of the building facade and the building interior structure data, and transmit the scanned data to the CIM system for point cloud registration and model update to generate a digital twin model of the fire scene. By integrating BeiDou positioning data, CIM models, and UAV sensor data, AI algorithms are used to generate fire prevention decision-making schemes that include fire prediction, personnel location, path planning, and resource scheduling. Receive fire-fighting decision-making plans, trigger coordinated rescue operations involving drones, fire trucks, and personnel, provide real-time feedback on execution status, and dynamically adjust instructions; Collect full-process data including equipment status, rescue logs, fire data, and the effectiveness of decision-making plans; evaluate the accuracy of CIM models, BeiDou positioning accuracy, UWB positioning accuracy, and AI decision-making accuracy; optimize equipment parameters and algorithm models; and update historical databases and rescue plans.

7. The low-altitude firefighting method according to claim 6, characterized in that, Complete equipment self-test and spatiotemporal calibration, and establish a unified centimeter-level positioning and millisecond-level timing reference, including: Perform hardware status checks on core equipment such as Beidou receivers, UWB base stations, drone flight controllers, and CIM servers, troubleshoot faults, and initiate backups. Deploy BeiDou differential base stations in an open area within 5km of the fire scene to receive satellite data, generate differential parameters, and broadcast them to each positioning terminal to establish an outdoor centimeter-level positioning benchmark. UWB base stations were deployed by combining drone airdrop and manual deployment to build an indoor positioning network. UWB coordinates were calibrated using BeiDou differential parameters to establish an indoor sub-centimeter-level positioning benchmark. By integrating BeiDou differential parameters, UWB conversion parameters, and CIM model coordinates, and unifying the time and coordinates of all devices, an initial CIM calibration model is generated, and system initialization is completed.

8. The low-altitude firefighting method according to claim 6, characterized in that, The system collects point cloud data of the building's exterior facade and interior structure, transmits the scanned data to the CIM system for point cloud registration and model updating, and generates a digital twin model of the fire scene, including: By combining drone parameters, CIM model and no-fly zone data, the width of the safety buffer zone is calculated, and the external serpentine and internal spiral flight paths of the drone are generated to ensure that the flight path is compliant and has no risk of collision. The heavy-duty drones are scheduled to fly along compliant routes, and cameras and LiDAR are activated to collect point cloud and texture data of the building facade. After noise reduction, the external structural data of the building is generated. Lightweight drones are deployed to enter the building through windows, activate radar and sensors to collect structural data on fire escape routes and refuge floors, calculate structural priorities, and generate interior structural data of the building. The external structural data of the building scanned by the drone and the internal structural data of the building are transmitted to the CIM system. The model is registered using the ICP algorithm, structural damage is marked, and a CIM model containing dynamic damage is generated.

9. The low-altitude firefighting method according to claim 6, characterized in that, Integrating BeiDou positioning data, CIM models, and UAV sensor data, AI algorithms generate fire prevention decision-making schemes that include fire prediction, personnel location, path planning, and resource allocation, including: The BeiDou positioning data, CIM model, and UAV sensor data are unified into the same spatiotemporal system, pre-processed, and then distributed to AI computing devices. Based on preprocessed data and historical cases, the fire spread rate and coverage are calculated using a thermal radiation model, collapse risk areas are marked, and fire prediction results and warning levels are output. Based on visual and UWB data, AI models are used to detect people and optimize positioning accuracy. Combined with sensor data, personnel priority is evaluated, and a personnel positioning and priority table is generated. Based on fire prediction and personnel priority, the improved A* algorithm is used to generate personnel evacuation and drone firefighting paths. Combined with equipment status, drones and fire trucks are allocated, and decision-making schemes and task parameters are output.

10. The low-altitude firefighting method according to claim 6, characterized in that, Receive fire-fighting decision-making plans, trigger coordinated rescue operations involving drones, fire trucks, and personnel, provide real-time feedback on execution status, and dynamically adjust instructions, including: Based on the fire safety decision-making plan, drone tasks are assigned, instructions are issued to control drones to perform tasks such as breaking windows, extinguishing fires, and delivering supplies, and the status is monitored in real time to handle anomalies such as returning to base or losing contact. Dispatch ground fire trucks to the assembly point and water source interface, guide rescue personnel to evacuate along the planned route, and use sound and light and flying drones to assist in guidance to achieve air-ground coordinated fire fighting; Real-time monitoring of rescue efforts and dynamic adjustments to rescue plans and resource allocation based on changes in fire intensity, new personnel, structural collapses, etc., to ensure that rescue operations adapt to the dynamic changes in the fire scene.

Images