Intelligent fire-fighting robot system and control method

Abstract

The invention discloses an intelligent fire-fighting robot system and a control method. Comprising the following steps: acquiring wide-area environment image information through a fixed monitoring camera network deployed in an operation area, and acquiring approaching environment image information through a mobile sensing system carried by a fire-fighting robot; according to the intelligent fire-fighting robot system and the control method, the fire-fighting robot can automatically adapt to complex terrains such as oil stains and stairs and accurately recognize fire source properties, so that the optimal moving mode and the fire extinguishing agent are automatically selected. According to the system, sensing data of fixed monitoring and the robot body are fused, dynamic path planning and personnel guiding are achieved, the traffic capacity, the fire extinguishing precision and the personnel rescue efficiency in a complex fire scene are remarkably improved, and the risk of rescue personnel is effectively reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image recognition, and in particular to an intelligent fire-fighting robot system and a control method. BACKGROUND

[0002] Currently, fire-fighting robots mostly use wheel or track chassis, which can move quickly on flat roads, but when facing complex unstructured terrains such as oil spill, debris accumulation, and stair steps commonly seen in fire scenes, the traffic capacity is severely limited, and the robot is easy to be trapped or overturned. At the same time, the existing robots usually rely on preset programs or remote control to select the fire extinguishing method, lack the ability to quickly identify the properties of burning materials on site, and are difficult to automatically match the most effective fire extinguishing agent (such as misusing water to extinguish metal fires), resulting in low fire extinguishing efficiency and possible secondary disasters. SUMMARY

[0003] The present application aims to at least partially solve one of the technical problems in the related art.

[0004] To this end, one object of the present application is to provide an intelligent fire-fighting robot system and a control method, so that the fire-fighting robot can autonomously adapt to complex terrains such as oil spills and stairs, and accurately identify the nature of the fire source, so as to automatically select the optimal moving mode and fire extinguishing agent. The system integrates the perception data of fixed monitoring and robot body, realizes dynamic path planning and personnel guidance, significantly improves the traffic capacity, fire extinguishing accuracy and personnel rescue efficiency in complex fire scenes, and effectively reduces the risk of rescue personnel.

[0005] To achieve the above-mentioned object, the first aspect of the present application proposes a control method of a fire-fighting robot, comprising the following steps: acquiring wide-area environment image information through a network of fixed monitoring cameras deployed in the working area, and acquiring close-range environment image information through a mobile perception system carried by the fire-fighting robot; identifying and processing the wide-area environment image information and the close-range environment image information, and extracting fire scene environment information including at least terrain feature information and material attribute information; based on the terrain feature information and the material attribute information, making a linkage decision to generate a control instruction set, the control instruction set including a first instruction for controlling the robot to switch the moving mode and a second instruction for controlling the fire extinguishing strategy; issuing the control instruction set to the fire-fighting robot to control it to perform corresponding actions.

[0006] According to the intelligent fire-fighting robot system and the control method, the fire-fighting robot can autonomously adapt to complex terrains such as oil stains and stairs, accurately identify the nature of the fire source, and automatically select the optimal moving mode and fire extinguishing agent. The system fuses the sensing data of the fixed monitoring and the robot body, realizes dynamic path planning and personnel guidance, significantly improves the passing capacity, fire extinguishing accuracy and personnel rescue efficiency in a complex fire scene, and effectively reduces the risk of rescue personnel.

[0007] In addition, the control method of the fire-fighting robot according to the above-mentioned embodiments of the present application can have the following additional technical features: In an embodiment of the present application, the linkage decision step specifically includes: When it is identified according to the image information that the ground is covered with liquid oil or chemical leakage, a first instruction is generated to switch the moving mode of the robot from the wheeled mode to the multi-legged mode; When it is identified according to the image information that the fire source is a metal fire, a D-class dry powder extinguishing agent is selected as the second instruction, and a third instruction is generated to plan a path to approach the fire source from the upwind direction.

[0008] In an embodiment of the present application, the linkage decision is also based on pre-acquired spatial information of the work area, and a pre-planned path containing at least one mode switching point is generated as part of the control instruction set; The position of the mode switching point is determined according to the terrain feature information and / or material attribute information.

[0009] The second aspect embodiment of the present application proposes an intelligent fire-fighting robot system, which includes a control platform and at least one fire-fighting robot. The control platform includes a communication module and a linkage decision engine, and the linkage decision engine is used to generate a control instruction set according to received environmental information containing spatial information of a work area. The fire-fighting robot includes: A moving chassis; A chassis frame installed on the bottom surface of the chassis frame through a lifting driving device, and four groups of driving wheels are arranged on the chassis frame; Six mechanical legs arranged around the hexagonal moving chassis, each of the mechanical legs includes a leg frame, a first joint, a second joint, a longitudinal adjusting mechanism, a transverse adjusting mechanism and multiple groups of folding and unfolding drivers; A visual perception module for real-time environmental verification of the fire-fighting robot; The fire extinguishing execution module comprises a fire extinguishing injection assembly, a mixing device, a plurality of groups of medicament bins, a plurality of connecting pipes and a plurality of electromagnetic valves, one input end of the mixing device is communicated with the plurality of groups of medicament bins through the plurality of connecting pipes, the one input end of the mixing device is connected with a water source through a connecting pipe, and the output of the mixing device is connected with the input of the fire extinguishing injection assembly; and the input port of each connecting pipe is provided with a corresponding electromagnetic valve; The main controller is in communication connection with the linkage decision engine, the lifting driving device, the driving wheel, the retractable drive of the mechanical leg, the visual perception module and the electromagnetic valve respectively; The main controller is configured to, in response to the first instruction from the linkage decision engine, control the lifting driving device and the retractable drive of the mechanical leg to act in coordination, so that the driving wheel and one of the mechanical legs contact the ground to switch the movement mode. In response to the second instruction, the corresponding electromagnetic valve is opened.

[0010] In addition, the intelligent fire-fighting robot system and the control method according to the above-mentioned application can have the following additional technical features: The visual perception module comprises: A multispectral imager is configured to collect spectral image data of a target; An infrared thermal imager is configured to collect infrared radiation image data of a scene, and the multispectral imager and the infrared thermal imager are installed side by side on a pan-tilt head that can be adjusted in pitch; A forward main camera is installed at a central axis position on the front part of the fire-fighting robot vehicle body and is configured to obtain the environment in front of the robot; A lateral auxiliary camera is symmetrically installed at positions close to the front and the rear on both sides of the vehicle body and is configured to obtain images of the environment on the sides of the robot.

[0011] In an embodiment of the present application, each of the mechanical legs comprises: A leg support for supporting the robot; A first joint, one end of which is hingedly connected to the proximal end of the leg support; A second joint, one end of which is hingedly connected to the other end of the first joint; A longitudinal adjustment mechanism, the adjustment shaft of which is hingedly connected to the other end of the second joint; A transverse adjustment mechanism, the adjustment shaft of which is connected to the adjustment seat of the longitudinal adjustment mechanism A plurality of groups of retractable drives, the rotation shafts of the first joint, the second joint, the longitudinal adjustment mechanism and the transverse adjustment mechanism are each provided with a retractable drive.

[0012] In an embodiment of the present application, the control platform further comprises a fixed monitoring camera network; The fixed monitoring camera network comprises a plurality of network cameras with a pan-tilt head, and the network cameras are in communication connection with the control platform, and the environmental information comprises wide-area visual information provided by the fixed monitoring camera network, and the fixed monitoring camera network comprises a plurality of network cameras with a pan-tilt head, and the network cameras are in communication connection with the control platform; A path planning engine, the job area spatial information comprises a pre-constructed park three-dimensional map, and the path planning engine accesses the pre-constructed park three-dimensional map, wherein the path planning engine is configured to: fuse the static spatial information of the park three-dimensional map and the real-time wide-area visual information provided by the fixed monitoring camera network, and plan a pre-planned path containing at least one mode switching point for the fire-fighting robot according to the decision result of the linkage decision engine, and the position of the mode switching point is determined based on the terrain feature information and / or material attribute information; The control instruction set comprises the pre-planned path.

[0013] In an embodiment of the present application, the main controller is further configured to: during the movement of the fire-fighting robot along the pre-planned path, perform real-time environmental perception through the visual perception module; If the deviation between the perceived real-time terrain and material information and the pre-judgment information on which the pre-planned path is generated exceeds a safety threshold, the current path is interrupted, and a path segment is re-planned locally based on the real-time perception information to bypass an unpre-judged obstacle or dangerous area; Meanwhile, the deviation information and the re-planned path segment are returned to the control platform.

[0014] In an embodiment of the present application, the multi-spectral imager is configured to acquire spectral images of a target in a visible light and at least one non-visible light waveband, and the linkage decision engine identifies material attributes by analyzing characteristic spectral lines of the spectral images; The multi-spectral imager and the linkage decision engine are cooperatively configured to identify material attributes including at least one of the following types by analyzing spectral images: Ground liquid pollutant attributes, and the identification result thereof is used to trigger a switching instruction to a multi-legged movement mode; Combustion chemical attributes, and the identification result thereof is used to trigger a selection instruction of a specific fire extinguishing agent; Building structure material attributes, and the identification result thereof is used to trigger a passability decision instruction of impact or bypassing;

[0015] A second aspect of this application provides a computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements a control method such as that of a fire-fighting robot.

[0016] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0017] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a schematic diagram of the structure of the fire-fighting robot in an intelligent fire-fighting robot system and control method according to this application; Figure 2 This is a schematic diagram of another form of the fire-fighting robot in an intelligent fire-fighting robot system and control method according to this application; Figure 3 This is a rear view schematic diagram of the fire-fighting robot in an intelligent fire-fighting robot system and control method according to this application; Figure 4 This is a flowchart of a method in an intelligent fire-fighting robot system and control method according to another embodiment of this application; Figure 5 This is a system block diagram of an intelligent fire-fighting robot system and control method according to another embodiment of this application.

[0018] As shown in the figure: 1. Mobile chassis; 2. Chassis frame; 3. Lifting drive device; 4. Drive wheel; 5. Mechanical leg; 6. Fire spray assembly; 7. Mixing device; 8. Connecting pipe; 9. Infrared thermal imager; 10. Front main camera; 11. Side auxiliary camera; 12. Multispectral imager; 501. Leg frame; 502. First joint; 503. Second joint; 504. Longitudinal adjustment mechanism; 505. Lateral adjustment mechanism; 506. Retraction and extension drive. Detailed Implementation

[0019] Embodiments of this application are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application. Rather, embodiments of this application include all variations, modifications, and equivalents falling within the spirit and scope of the appended claims.

[0020] The following description, in conjunction with the accompanying drawings, illustrates an intelligent fire-fighting robot system and its control method according to an embodiment of this application.

[0021] like Figures 1-5 As shown in the figure, a control method for a fire-fighting robot according to an embodiment of this application includes the following steps: Wide-area environmental image information is obtained through a network of fixed surveillance cameras deployed in the work area, and close-range environmental image information is obtained through a mobile sensing system carried by the firefighting robot; The system identifies and processes wide-area and close-range environmental image information to extract fire scene environmental information, including at least terrain feature information and material property information. Based on terrain feature information and material property information, a coordinated decision is made to generate a control instruction set, which includes a first instruction for controlling the switching of robot movement modes and a second instruction for controlling fire extinguishing strategies. The control command set is sent to the fire-fighting robot to control it to perform the corresponding actions.

[0022] In this implementation, the coordinated decision-making steps specifically include: When the image information identifies the presence of liquid oil or chemical leaks on the ground, the first instruction is generated to switch the robot's movement mode from wheeled to multi-legged mode. When the fire source is identified as a metal fire based on the image information, a second instruction is generated to select Class D dry powder extinguishing agent, and a third instruction is generated to plan a path to approach the fire source from the upwind direction.

[0023] In this implementation, the linkage decision-making also generates a pre-planned path containing at least one mode switching point as part of the control instruction set based on the pre-acquired spatial information of the work area. The location of the mode switching point is determined based on terrain feature information and / or material property information.

[0024] The workflow of this implementation plan in practical application is as follows, which is used to specifically illustrate the process of "generating a pre-planned path based on the pre-acquired spatial information of the work area": Event Triggering and Information Integration: When the park's fixed surveillance camera network detects a suspected fire, such as smoke or an abnormal heat source, the system automatically issues an alarm and determines the initial location of the incident. The control platform immediately retrieves the corresponding high-precision 3D map of the park, i.e., the "pre-acquired work area spatial information," which includes static data such as building structures, passageways, staircases, and key facilities, such as chemical storage points.

[0025] Generate a pre-planned path: The collaborative decision engine combines the following information to plan the path: Starting point and target point: The starting point is the current location of the standby firefighting robot, and the target point is the initial fire location point.

[0026] 3D map data: Provides global path constraints such as passable areas, passage widths, and stair locations.

[0027] Real-time information identification: During planning, the system simultaneously accesses real-time footage from a fixed camera network for rapid analysis, extracting key environmental dynamic information, such as visible smoke in a passage or apparent obstruction at an intersection. The system then overlays and verifies this dynamic terrain feature information, such as "passage obstructed," with material property information, such as "smoke composition is burning plastic," onto a 3D map.

[0028] Intelligent embedding of mode switching points: In the generated pre-planned path, the system does not simply connect the start and end points, but actively and intelligently embeds "mode switching points" based on the fused environmental information. For example: Path analysis shows that the robot first needs to traverse a long, straight, and flat driveway, then enter a cargo yard filled with miscellaneous items, and finally arrive at the stairs at the warehouse entrance.

[0029] Based on the terrain feature of "flat lane", the decision engine sets the road segment as a "wheeled high-speed mode" segment.

[0030] Before entering the cargo yard, the system determined that the wheeled vehicle had poor passability based on the "unhardened ground" attribute marked on the map and the "scattered debris" information recognized by the real-time image. Therefore, a "mode switching point" was set at the entrance of the cargo yard, instructing the robot to switch to multi-legged mode in advance to stably cross the debris.

[0031] At the warehouse staircase entrance, the system sets a second "mode switching point" based on the "staircase" terrain features clearly identified by the 3D map, instructing the robot to maintain or switch to multi-legged mode for climbing.

[0032] Command Issuance and Robot Execution: This "pre-planned path," containing a complete sequence of path coordinates and precise mode switching point instructions, is issued to the firefighting robot as the core of the control command set. After the robot departs, its main controller will automatically navigate along this path and automatically trigger the wheel / leg mechanism to switch modes when approaching the predetermined switching point, achieving optimal maneuverability throughout the entire process.

[0033] This application proposes an intelligent firefighting robot system, comprising: a control platform and at least one firefighting robot; The control platform includes a communication module and a linkage decision engine. The linkage decision engine is used to generate a set of control instructions based on the received environmental information, which includes spatial information of the work area. Firefighting robots include: Mobile chassis 1; Chassis frame 2, chassis frame 2 is installed on the bottom surface of chassis frame 2 by lifting drive device 3, and four sets of drive wheels 4 are provided on chassis frame 2; Six mechanical legs 5 are arranged around the hexagon of the mobile chassis 1. Each mechanical leg 5 includes a leg frame 501, a first joint 502, a second joint 503, a longitudinal adjustment mechanism 504, a transverse adjustment mechanism 505, and multiple sets of retractable drivers 506; A visual perception module for real-time environmental verification of the fire-fighting robot; A fire-extinguishing execution module includes a fire-fighting spray assembly 6, a mixing device 7, multiple sets of chemical agent bins, multiple connecting pipes 8, and multiple solenoid valves. One of the input ends of the mixing device 7 is connected to the multiple sets of chemical agent bins through the multiple connecting pipes 8. One of the input ends of the mixing device 7 is connected to a water source through a connecting pipe 8. The output of the mixing device 7 is connected to the input of the fire-fighting spray assembly 6. A corresponding solenoid valve is provided at the input port of each connecting pipe 8; A main controller, which is respectively communicatively connected to the linkage decision-making engine, the lifting drive device 3, the drive wheels 4, the retractable drivers 506 of the mechanical legs 5, the visual perception module, and the solenoid valves; Among them, the main controller is configured to: in response to a first instruction from the linkage decision-making engine, control the lifting drive device 3 and the retractable drivers 506 of the mechanical legs 5 to act in coordination, so that one of the drive wheels 4 and the mechanical legs 5 touches the ground to switch the movement mode; In response to a second instruction, control the corresponding solenoid valve to open.

[0034] The following is a detailed description of the specific working process of this technical solution in an actual fire-fighting scenario: 1. Fire situation receiving and instruction generation stage: When the fixed monitoring network in the park detects a fire and uploads it to the control platform, the linkage decision-making engine starts to work. It integrates the real-time video from the monitoring camera, shows the flame position, spreading direction, static information from the park digital map, building layout, passage position, and a preset fire scene knowledge base, such as the location of chemical warehouses and the characteristics of combustibles, and generates a structured control instruction set within seconds. For example, for an electrical fire deep in the warehouse, the engine may decide: "The path planning is to enter the wheeled mode along the main passage from the west gate → switch to the multi-legged mode at the entrance of the shelf area → use carbon dioxide fire extinguishing agent for local spraying".

[0035] 2. Robot instruction receiving and initialization stage: The above instruction set is sent to the main controller of the designated fire-fighting robot through the communication module. The main controller then analyzes the instruction and enters the preparation state. For example, it first calculates the required initial rotation speed and direction of the relevant drive wheels 4 according to the first section of the path in the instruction.

[0036] 3. Movement mode execution and switching stage: Wheel-type mobile start: The main controller controls the lifting drive device 3 according to the instruction to lower the entire chassis frame 2 to a low position, so that the four sets of drive wheels 4 are in full contact with the ground. At the same time, it controls the retraction and extension drive 506 of all mechanical legs 5 to completely retract and lift the six mechanical legs 5 off the ground and lock them within the outline of the vehicle body.

[0037] Subsequently, the main controller drives four sets of drive wheels and four motors, and the robot begins to drive at high speed towards the fire in wheel mode.

[0038] Predictive mode switching: When the robot recognizes the "shelf area entrance" mode switching point preset in the instruction through its own visual perception module, such as the forward-facing main camera 10, and the local perception confirms that the terrain has become uneven or there are obstacles, the main controller triggers the mode switching program.

[0039] It executes in sequence: a) Controls drive wheel 4 to decelerate to a stop; b) Control the lifting drive device 3 to lift the chassis frame 2, so that the drive wheels 4 are suspended off the ground; c) Simultaneously, the retraction and extension actuators 506 control the six robotic legs 5 to extend them sequentially to the ground contact position. During this process, the longitudinal and lateral adjustment mechanisms 505 fine-tune the length and angle of each leg based on pre-loaded terrain data and real-time sensor feedback to ensure a stable landing. The entire process is completed within seconds, and the robot seamlessly transitions from a "four-wheeled" state to a "six-legged" state.

[0040] Multi-legged obstacle crossing: After switching, the main controller controls the six mechanical legs 5 to move in a coordinated manner. Through the cooperation of the first joint 502, the second joint 503, etc., the machine walks across the complex and uneven ground between the shelves and maintains the stability of the machine body.

[0041] Firefighting mission execution phase: Once the robot arrives near the designated fire extinguishing location, its visual perception modules, such as the multispectral imager 12 and the infrared thermal imager 9, perform a final confirmation of the fire source. For example, it confirms that the burning material is electrical equipment. The main controller then responds to the second instruction in the command, controlling the opening of a specific solenoid valve connected to the carbon dioxide agent compartment in the fire extinguishing execution module, while keeping the solenoid valves supplying water and other agents closed.

[0042] Under pressure, the carbon dioxide agent flows through the connecting pipe 8 into the mixing device 7. In this case, mixing may not be necessary; the agent may pass directly through and is ultimately sprayed precisely onto the fire source area by the fire spray assembly 6, such as a steerable nozzle. The main controller can control the agent flow rate by adjusting the opening of the solenoid valve based on fire feedback.

[0043] In one embodiment of this application, the visual perception module includes: Multispectral imager 12 is used to acquire spectral image data of the target; Infrared thermal imager 9 is used to collect infrared radiation image data of the scene. Multispectral imager 12 is mounted side by side with infrared thermal imager 9 on a tilt-adjustable gimbal. The front-facing main camera 10 is installed on the central axis of the front of the fire-fighting robot body to obtain the environment in front of the robot. Side auxiliary cameras 11, at least two side auxiliary cameras 11 are symmetrically installed on both sides of the vehicle body at the front and rear positions, for acquiring images of the robot's side environment.

[0044] In actual firefighting operations, firefighting robots transform raw image data into environmental understanding that is crucial for rescue operations.

[0045] Startup and Global Scan: Upon receiving the task command, the robot's visual perception module is activated. The forward-facing main camera 10, located on the central axis of the front of the vehicle, immediately begins operation, providing a color 3D view of the robot's front. This is the primary basis for the robot's autonomous navigation and avoidance of large obstacles. Simultaneously, the lateral auxiliary cameras 11, symmetrically mounted on the front and rear sides of the vehicle, are activated. Together, they form a lateral field of view ring around the vehicle, ensuring that the robot can perceive collision risks or spatial constraints from the sides when navigating narrow passages or performing delicate maneuvers, thus avoiding scrapes.

[0046] In-depth reconnaissance and professional analysis: When the robot approaches the fire area or the system instructs it to focus on a specific location, the tilt-adjustable gimbal located above the robot begins to function. A multispectral imager 12 and an infrared thermal imager 9 are mounted side-by-side on the gimbal.

[0047] The gimbal adjusts its angle according to the instructions, aligning the two precision sensors with the target area.

[0048] The infrared thermal imager 9 quickly scans and generates a temperature distribution map, identifying potentially trapped heat sources, such as the human body.

[0049] At the same time, the multispectral imager 12 captures reflectance spectral data of the target in multiple specific bands. These unique spectral "fingerprints" are immediately sent to the processing unit for rapid comparison with a pre-stored material spectral database, thereby directly identifying the material of the burning material, such as wood, plastic, textiles, or the type of chemicals spilled on the ground, such as oils and solvents.

[0050] Data Fusion and Intelligent Decision Support: The data acquired by all cameras and sensors are not isolated. Real-time video streams from forward and side cameras are used to construct an instant 3D map of the robot's surroundings and continuously detect dynamic obstacles. Temperature field data from the infrared thermal imager 9 and material identification results from the multispectral imager 12 are superimposed on this 3D map as key semantic information. For example, the map not only shows "there is an obstacle ahead," but also clearly marks it as "a metal shelf with extremely high temperature" or "a pool of leaked oil with flammable properties." This comprehensive environmental perception result, which integrates geometric, temperature, and chemical information, is uploaded in real time to the linkage decision engine of the control platform or used for the robot's own instant decision-making, becoming the most direct and reliable basis for determining whether to switch movement modes, select fire extinguishing agents, and plan safe approach paths.

[0051] Dynamic Adaptation and Closed-Loop Verification: Throughout the mission, the visual perception module remains operational. For example, as the robot moves along a pre-planned path, the forward-facing main camera 10 continuously compares the actual scene with the expected view, enabling real-time environmental verification. If an unexpected obstacle is detected, the robot can immediately initiate local replanning. Sensors on the gimbal also dynamically scan the fire according to its development, continuously updating fire scene information to ensure the adaptability and safety of the rescue operation.

[0052] In one embodiment of this application, each mechanical leg 5 includes: Leg support 501, used for robot support; First joint 502, one end of first joint 502 is hinged to the proximal end of leg frame 501; The second joint 503, one end of which is hinged to the other end of the first joint 502; The longitudinal adjustment mechanism 504 has its adjustment shaft hinged to the other end of the second joint 503. The lateral adjustment mechanism 505 has an adjustment shaft that is connected to the adjustment seat of the longitudinal adjustment mechanism 504. Multiple sets of retraction and extension actuators 506 are provided, and the rotating shafts of the first joint 502, the second joint 503, the longitudinal adjustment mechanism 504, and the lateral adjustment mechanism 505 are all equipped with retraction and extension actuators 506.

[0053] In actual firefighting missions, this mechanical leg 5 structure serves as the actuator for the robot to switch between wheeled high-speed movement and six-legged obstacle-crossing mode. Specifically, Initial state wheel mode: When the robot is in wheeled movement mode, all mechanical legs 5 are fully retracted. At this time, the first joint 502 and the second joint 503 of each leg rotate to an angle that makes the leg frame 501 fit tightly against or retract within the outline of the vehicle body, driven by their respective retraction and extension actuators 506.

[0054] The longitudinal adjustment mechanism 504 retracts to its shortest stroke, and the lateral adjustment mechanism 505 also retracts or moves to its innermost position, so that the entire mechanical leg 5 presents the smallest spatial envelope, ensuring that it will not interfere with the ground or side objects when the robot is moving at high speed.

[0055] The mode switching process changes from wheeled to multi-legged: When the main controller issues a command to switch to multilegged mode, a coordinated sequence of actions is triggered: Lateral deployment: First, the lateral adjustment mechanism 505 of each leg starts to work under the drive of its actuator, pushing the entire mechanical leg 5 assembly to move radially outward along the chassis, making room for the leg to swing and land.

[0056] Joint extension and leg lowering: Immediately afterwards, the actuator of the second joint 503 is activated, driving the joint to rotate and swinging the first joint 502, leg support 501, and other components connected to it from their stored position to the outside and downward of the vehicle body. Subsequently, the actuator of the first joint 502 is activated to further adjust the height and posture of the leg support 501 relative to the vehicle body.

[0057] Longitudinal grounding adjustment: When the end of the leg 501 approaches the ground, the longitudinal adjustment mechanism 504 begins to work precisely. Its driver dynamically adjusts the extension and retraction of the adjustment shaft based on pre-loaded terrain data or real-time contact feedback, thereby precisely controlling the effective length of the leg 501, ensuring that all six legs can make smooth contact with the ground simultaneously or sequentially, and adapting to the initial undulations of the ground.

[0058] Body lifting and wheel lift-off: After all the mechanical legs 5 have established stable support with the ground, the longitudinal adjustment mechanism 504 can further coordinate its movements to lift the entire robot chassis frame 2 upwards until the drive wheels 4 are completely off the ground and locked. At this point, the switch is complete, and the robot is fully supported by the six mechanical legs 5.

[0059] Multi-legged walking and obstacle crossing process: In multi-legged mode, the robot achieves gait walking and obstacle crossing by coordinating and controlling the various joints and adjustment mechanisms of its six legs: Walking: By cyclically controlling the swing angle of the first joint 502 and the second joint 503 of each leg, the micro-motion of the lateral adjustment mechanism 505, and the extension and retraction of the longitudinal adjustment mechanism 504, the gait cycle of lifting the leg, stepping, and landing for support is realized, driving the robot to move forward, backward, or turn.

[0060] Terrain Adaptation: When encountering uneven ground, the longitudinal adjustment mechanism 504 of each leg can independently and in real time extend and retract its adjustment axis, much like an "adaptive suspension," adjusting the length of that leg individually to maintain fuselage stability. The lateral adjustment mechanism 505 can also be fine-tuned to optimize the support polygon and enhance lateral stability.

[0061] Obstacle crossing: When facing steps or ditches, the movement trajectory of the end of the leg frame 501 is controlled by the combined movement of the first joint 502 and the second joint 503. Combined with the large extension and retraction of the longitudinal adjustment mechanism 504, the leg can be lifted to step, climb or descend.

[0062] Switching back to wheel mode: The process is the reverse of the above switching. The longitudinal adjustment mechanisms 504 of each leg first retract in coordination, lowering the vehicle body so that the drive wheels 4 contact the ground. Subsequently, the joint actuators actuate, raising and retracting the leg supports 501. Finally, the lateral adjustment mechanism 505 retracts, bringing the entire leg to the driving position and locking it.

[0063] In summary, the mechanical leg 5 structure provides the main directional swing freedom through the first joint 502 and the second joint 503, provides key length adaptability through the longitudinal adjustment mechanism 504, and provides additional posture adjustment and storage space through the lateral adjustment mechanism 505.

[0064] In one embodiment of this application, the control platform further includes a fixed surveillance camera network; The fixed surveillance camera network includes multiple network cameras with pan-tilt-zoom (PTZ) switches. The network cameras are connected to the control platform. The environmental information includes wide-area visual information provided by the fixed surveillance camera network. The fixed surveillance camera network includes multiple network cameras with pan-tilt-zoom (PTZ) switches. The network cameras are connected to the control platform. A path planning engine is provided. The spatial information of the work area includes a pre-built 3D map of the park. The path planning engine is connected to the pre-built 3D map of the park. The path planning engine is configured to: integrate the static spatial information of the 3D map of the park with the real-time wide-area visual information provided by the fixed monitoring camera network, and plan a pre-planned path for the fire-fighting robot with at least one mode switching point according to the decision result of the linkage decision engine. The location of the mode switching point is determined based on terrain feature information and / or material property information. The control instruction set includes pre-planned paths.

[0065] The following is a description of the specific implementation process of the control platform and path planning system in actual fire emergency response: In actual park fire emergency scenarios, the control platform and its integrated fixed surveillance camera network and path planning engine together constitute a system that begins working from the very first moment a fire occurs.

[0066] Phase 1: Fire detection and initial overall situation mapping.

[0067] When a fire breaks out in a certain part of the park, the network of fixed surveillance cameras with pan-tilt units, which are widely deployed on the roofs of factories, corridors, and warehouses, takes immediate action.

[0068] These cameras typically operate 24 / 7, transmitting their video streams in real-time to the control platform via the campus network. The control platform's built-in intelligent video analytics module, acting as part of a coordinated decision-making engine, continuously scans the video streams. Once algorithms identify pre-set alarm signals such as flame characteristics, unusual smoke, or people running, a fire alarm is immediately triggered. The platform automatically controls the nearest rotatable camera's pan-tilt-zoom (PTZ) to focus on the alarm area, zooming and tracking to obtain initial wide-area visual information, such as the location of the flames, their approximate direction of spread, visible obstacles (e.g., collapsed shelves blocking aisles), and whether there are signs of trapped personnel.

[0069] Phase Two: Information Fusion and Intelligent Path Generation.

[0070] Upon triggering the alarm, the control platform automatically activates the emergency response procedures related to the incident area. The route planning engine then starts and performs the following key steps: Static map loading: The engine immediately retrieves a pre-built, high-precision 3D map of the campus. This digital map contains all static spatial information, including the exterior and internal structure of all buildings, the precise width of each passageway, the location and step height of stairs, the type of doors, and the coordinates of important facilities such as power switches and chemical storage points.

[0071] Dynamic information overlay: The engine overlays the real-time wide-area visual information obtained from the fixed camera in the first stage onto the static 3D map as a transparent dynamic layer.

[0072] For example, a main passageway that was originally clear on the map appears to be shrouded in thick smoke and has fallen debris in the real-time footage; or, a normal room marked on the map shows an abnormally high temperature in its doors and windows according to real-time thermal imaging data.

[0073] "Predictive" path planning: Based on preliminary fire information, such as "packaging materials in warehouse 2 of area B are on fire", the linkage decision engine will make preliminary tactical decisions based on its built-in fire protection knowledge base, such as "water-based fire extinguishing agent needs to be used" and "approach from the east wind direction".

[0074] After receiving these decision intentions, the path planning engine begins to plan a pre-planned path for the firefighting robot from the assembly point to the fire scene.

[0075] Path generation: Based on the fused information of "static map + dynamic vision", the engine calculates a feasible route in three-dimensional space that balances the shortest time and the highest safety. It can not only avoid permanent walls on the map, but also avoid temporary congestion points found in real-time footage.

[0076] Mode switching point pre-embedding: The engine analyzes each segment of terrain along the path: for long, straight, flat asphalt roads confirmed by both the map and the video, it marks them as "wheeled high-speed passage sections"; for internal warehouse areas that are about to be entered, where the map shows that the ground may have tracks or potholes, and the real-time video shows that there are scattered goods on the ground, the engine will precisely set a mode switching point at the entrance and pre-instruct the robot to switch to multi-legged obstacle-crossing mode at this point; for fire escape stairs that must be passed, the engine will set another switching point at the stairwell. The location and timing of these mode switching points are precisely determined based on the prediction of the terrain features ahead, such as stairs and uneven ground.

[0077] Phase 3: Command Issuance and Execution Blueprint.

[0078] Once the planning is complete, this intelligent pre-planned path, including coordinate sequences, speed suggestions, and most importantly, mode switching point instructions, is encapsulated in a control instruction set and immediately sent to the standby firefighting robots via wireless network.

[0079] Furthermore, the control platform also includes a personnel detection module and an escape route planning module; the personnel detection module is used to analyze video information provided by the fixed surveillance camera network to identify and locate trapped personnel; The escape route planning module is configured to plan a dynamic escape route from the location of the trapped personnel to the safe area based on the 3D map of the park, real-time fire environment information, and the location of the trapped personnel. The collaborative decision-making engine is also configured to generate personnel guidance task instructions when trapped personnel are identified. The firefighting robot also includes a guidance device, which is connected to the main controller and is used to guide trapped personnel to evacuate along a dynamic escape route according to personnel guidance task instructions.

[0080] The guiding device includes at least one of the following: LED displays or laser projectors used to display directional information; A directional speaker used to broadcast voice guidance information.

[0081] In actual fire rescue operations, the system's personnel guidance and escape functions operate in parallel with firefighting missions. Specifically, Trapped personnel discovery and location: A network of fixed surveillance cameras deployed throughout the park continuously transmits video streams back to the control platform. The platform's built-in personnel detection module analyzes these images in real time, enabling it to identify not only flames and smoke but also stationary or moving human figures in complex environments such as dense smoke and low light.

[0082] Once a trapped person is detected in an unsafe area, the module immediately marks their precise 3D coordinates by fusing multiple camera views with a 3D map and assesses their status, such as mobility and whether they have fallen.

[0083] Dynamic escape route planning: The escape route planning module of the control platform can be integrated with the fire extinguishing route planning engine or activated independently. This module starts from the current location of the trapped personnel and ends at the nearest safe exit or assembly point.

[0084] During planning, the module deeply integrates the following information to generate a real-time optimal channel: Static information: The location of building structures, doors, stairs, and safety exits in the 3D map of the park.

[0085] Dynamic hazard information: Real-time data transmitted from fixed cameras and robots that have entered the fire area, including the direction of fire spread, high-temperature areas, dense smoke diffusion paths, and passage blockage status.

[0086] Resource information: Location and status of nearby available firefighting robots.

[0087] The planned escape routes are dynamic and personalized. For multiple trapped individuals in the same area, the system may plan different diversion routes to avoid congestion in the passageways; for those with mobility impairments, the routes will prioritize avoiding obstacles such as stairs.

[0088] Task generation and robot scheduling: After receiving the location of personnel and the planned escape routes, the collaborative decision-making engine will conduct multi-task priority arbitration. Usually, life-saving tasks will be given the highest priority.

[0089] The engine will assign the most suitable firefighting robot to perform the guidance task, either nearby or based on its capabilities. In addition to escape routes, the instruction set also includes specific guidance strategies, such as "accompanying guidance," "forward navigation," and "remote instruction."

[0090] The robot performs a guidance task: The designated firefighting robot may suspend or adjust its firefighting mission and quickly move to the vicinity of the trapped personnel. The robot uses its visual perception modules, such as the forward-facing main camera 10 and thermal imager, to make a final identification and tracking of the personnel.

[0091] The robot performs guidance through its guidance device: Visual guidance: High-brightness, multi-directional LED arrow matrix display screens and laser-projected walking arrows on the ground clearly indicate the direction of movement for people.

[0092] Auditory guidance: Voice commands, such as "Please follow the direction of the green arrow" and "Please move towards me," are played through directional speakers and can be switched to multiple languages.

[0093] Environmental perception and interaction: The robot uses side cameras and sensors to ensure that the escape route is unobstructed as it moves. It can interact with people simply through voice or light signals, such as "Please follow" or "Please crouch down to pass."

[0094] Closed-loop escort and status reporting: The robot moves along the planned escape route, occasionally turning back to check if the people are keeping up, forming an escort mode.

[0095] The robot will transmit real-time data to the control platform regarding the status of personnel during the guidance process, including whether they are following, their current location, and their initial physiological state. If the original route becomes dangerous due to a sudden change in the fire situation, the platform will immediately plan and issue a new alternative escape route for both the robot and the trapped personnel.

[0096] Once personnel are guided to a safe area, the robot marks the task as complete and can return to the fire scene to continue firefighting operations or receive new instructions, depending on the situation.

[0097] In one embodiment of this application, the main controller is further configured to: perform real-time environmental perception through a visual perception module while the fire-fighting robot moves along a pre-planned path; If the deviation between the perceived real-time terrain and material information and the predicted information on which the pre-planned path is based exceeds the safety threshold, the current path will be interrupted and a new path will be replanned locally based on the real-time perceived information to bypass unpredicted obstacles or dangerous areas. At the same time, the deviation information and the replanned path segment are transmitted back to the control platform.

[0098] In actual firefighting operations, the local replanning function of the main controller is a key autonomous capability that ensures the robot can respond to emergencies at the fire scene and guarantees the ultimate success of the mission. Its implementation process is a precise closed loop based on real-time perception, intelligent judgment, and rapid calculation. Dynamic monitoring and deviation identification, perception, and judgment: When the robot follows the "pre-planned path" issued by the control platform, its visual perception module is always in a high-frequency working state.

[0099] The main controller continuously compares the real-time point cloud data, images, and temperature information captured by the sensors with the prediction information carried in the instruction set at the millisecond level.

[0100] This predictive information is an environmental snapshot that the path planning engine used when it generated the path. It might be described as "a 2-meter-wide flat passage at the 35-meter mark, with a concrete surface".

[0101] Example of deviation: When the robot travels to the predetermined position, the forward-facing camera identifies that the passage is partially blocked by a collapsed metal shelf, leaving an effective width of only 0.8 meters, which is a terrain feature deviation; at the same time, the infrared thermal imager 9 shows an additional dangerous area with extremely high local temperature on the shelf; the multispectral imager 12 may also detect unknown reflective liquid material on the ground, which is a material property deviation.

[0102] Threshold Judgment: The algorithm built into the main controller quantifies these deviations. For example, it compares parameters such as "channel width reduced by 1.2 meters," "high temperature source >150 degrees Celsius ahead," and "unpredictable liquid substance on the ground" with preset safety thresholds. These thresholds are set based on the robot's size, maneuverability, heat resistance, and anti-slip capabilities. Once any deviation exceeds the threshold, the system determines that "the original path is not safe to traverse under the current conditions."

[0103] Autonomous Disruption and Local Replanning Decision-Making and Execution: Upon determining that the deviation exceeds the limit, the main controller immediately interrupts the tracking of the currently pre-planned path segment. The robot enters a paused or low-speed alert state. Immediately afterwards, without waiting for new instructions from the remote control platform (because communication at the fire scene may be unstable and remote decision-making may be delayed), it initiates local replanning.

[0104] The replanning engine starts from the robot's current position and aims to reconnect to the original planned path or reach the next key node after bypassing obstacles, with the next mode switching point as the temporary target.

[0105] The planning is based entirely on real-time perception information from the robot's local environment. Utilizing instant-generated 3D obstacle contours and safety boundaries, it quickly calculates one or more local alternative path segments within its computational capabilities. This new path prioritizes safety by avoiding high temperatures and leaks, while also considering efficiency and minimizing the detour distance.

[0106] Execution of new path and information feedback: Once a feasible local path is generated, the main controller immediately directs the robot to execute it. This may involve fine-tuning the movement mode; for example, in very narrow maneuvering spaces, even if the original instruction was wheeled mode, the robot may automatically switch to multi-legged mode to achieve higher control precision.

[0107] Simultaneously, the communication module is triggered, packaging and transmitting critical field data back to the control platform. The transmitted information includes at least: Deviation information: The specific discrepancy between the prediction and the actual situation, such as "the passageway at coordinates X, Y is blocked by a metal shelf, the actual width is 0.8 meters, and the shelf temperature is about 200 degrees Celsius".

[0108] The replanned path segment: the specific trajectory coordinate sequence that the robot autonomously selects to detour.

[0109] Environmental snapshot: A summary of relevant sensor data, such as images and heat maps.

[0110] Platform synchronization and global optimization closed loop: After receiving the feedback information, the control platform will immediately take action based on it: Map Update: The dynamic information of "X,Y channel blocked" is used as a layer and updated in real time on the public combat map shared by all robots and commanders to prevent other robots from planning to this route.

[0111] Strategy Assessment: Assess whether the local obstacle affects the overall mission. If the impact is significant, such as if it is the only passage to the main fire area, the platform may replan the global mission and path for all relevant robots and issue new planning instructions.

[0112] Knowledge base learning: This event can be stored as a case in the system's knowledge base. In the future, when planning paths in similar environments, the algorithm may introduce a more conservative safety margin.

[0113] In one embodiment of this application, the multispectral imager 12 is configured to acquire spectral images of the target in the visible light and at least one non-visible light band, and the linkage decision engine identifies material properties by analyzing the characteristic spectral lines of the spectral images. The multispectral imager 12 and the linked decision engine are configured to identify material properties including at least one of the following categories by analyzing spectral images: The properties of liquid contaminants on the ground are identified, and the results are used to trigger a switch to a multi-legged movement mode. The chemical properties of the combustible material are identified, and the results are used to trigger a selection command for a specific extinguishing agent. The identification results of the building structure material properties are used to trigger passability decision instructions for impact or detour.

[0114] In the actual operation of firefighting robots, the collaborative material identification function of the multispectral imager 12 and the linkage decision engine is implemented in a continuous chain from physical signal acquisition to chemical property determination, and finally drives precise action: Spectral data acquisition: obtaining the "optical fingerprint" of substances. As the robot approaches the target area, such as a patch of ground, a flame, or a wall, the multispectral imager 12 begins operation. It doesn't just take pictures; it simultaneously performs "spectral splitting" imaging of the target scene. Within microseconds, it collects information on the intensity of reflected light from the target in the visible light bands, such as red, green, and blue, as well as at least one non-visible light band, such as near-infrared or short-wave infrared. Substances with different chemical compositions possess unique absorption and reflection properties for specific wavelengths of light due to their molecular structures.

[0115] Characteristic spectral line analysis and material identification: comparing against a "fingerprint database" The acquired raw spectral image data is transmitted in real time to a dedicated analysis module in the linked decision engine.

[0116] Feature extraction: The algorithm extracts the location, intensity, and width of key reflectance curves or characteristic absorption peaks from multi-band data. For example, hydrocarbons, such as oils, have typical absorption characteristics in the near-infrared band; certain plastics have unique reflection peaks in specific short-wave infrared bands.

[0117] Database comparison: The extracted feature spectral lines will be matched and compared at high speed with the pre-stored, calibrated material spectral database in the engine.

[0118] Attribute determination: Through matching, the engine can identify the properties of materials with a high degree of confidence. This is not just about identifying "liquid" or "solid", but can be specific to "water-based liquids", "hydrocarbon oils", "thermoplastics containing phosphorus flame retardants", "silicate building materials such as gypsum board" or "reinforced concrete".

[0119] 3. Identification Results Drive Precise Decision-Making: From "What is it?" to "How to do it?" The identified material properties are not the ultimate goal; their value lies in their immediate transformation into actionable rescue decisions. Based on the identified property categories, the system automatically triggers preset, optimized response strategies: Scenario 1: Identify the properties of liquid contaminants on the ground, such as hydraulic oil and solvents.

[0120] Process: As the robot moves forward, the multispectral imager 12 scans the ground ahead and finds that the spectral characteristics of a region highly match the "lubricating oil" in the database, and the infrared thermal imager 9 does not show high temperature in that area.

[0121] Decision-making and Action: The collaborative decision-making engine determined that the area posed a risk of "loss of traction and pollution spread." This attribute recognition result served as direct evidence, immediately triggering the generation of the "first instruction": before reaching the area, switch the movement mode from wheeled to hexapod. The robot then raised its tires in advance and switched to six-legged walking, avoiding tire slippage, contamination, and potential secondary disasters.

[0122] Scenario 2: Identify the chemical properties of combustible materials, such as lithium batteries, magnesium metal, and wood.

[0123] Process: The robot is aimed at the fire source, and the multispectral imager 12 analyzes the spectral characteristics of the burning material and volatile gases through the flames and smoke. The identification results show that the characteristic spectral lines match the entries in the "metallic magnesium" database.

[0124] Decision and Action: The linkage decision engine immediately invokes the fire protection rule base to confirm that magnesium metal fires are Class D and water-sensitive. This "combustible material chemical property" identification result directly triggers the generation of a precise "second instruction": 1. Fire extinguishing strategy instruction: Select Class D dry powder extinguishing agent and lock the corresponding agent compartment solenoid valve in preparation for opening; 2. Safety Policy Instruction: Simultaneously generate a "third instruction" that commands the path planning engine to plan a path from the current upwind position to the fire source, and strictly prohibits approaching from the downwind direction.

[0125] Scenario 3: Identify the properties of building structural materials, such as lightweight partition walls and load-bearing brick walls.

[0126] Process: When it is necessary to break down or cross obstacles, the robot performs a multispectral scan of the wall in front. Spectral feature analysis shows that its main materials are "plasterboard" and "wooden framing", identifying it as a typical non-load-bearing lightweight partition wall.

[0127] Decision-making and Action: Based on the identification result of "building structure material properties - low strength, non-load-bearing", the linkage decision engine triggers a passability decision command. It may decide "can be broken through by impact" and simultaneously generate auxiliary commands: to obtain the maximum impact force, instruct the robot to switch to wheel mode and accelerate before impact. Conversely, if it identifies a "reinforced concrete" load-bearing wall, the engine will decide "cannot be impacted, need to find a door or window" or "need to use demolition tools for delicate work".

[0128] This application proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any step of the control method for the fire-fighting robot described above.

[0129] Specifically, this application is used in practice.

[0130] Fire monitoring and command generation: The park's fixed surveillance camera network identifies fires and issues alarms. The control platform retrieves a 3D map and integrates real-time video information. After analysis by the linkage decision engine, a set of control commands is generated, including pre-planned paths, embedded mode switching points, and fire extinguishing strategies, and then sent to the fire-fighting robot.

[0131] Robot execution and mode switching: After receiving instructions, the robot moves at high speed in wheeled mode along the path. At preset switching points, such as uneven ground or stairs, the main controller controls the lifting device and mechanical legs 5 to work together to switch the movement mode from wheeled to hexapod mode for stable obstacle crossing.

[0132] Firefighting and Personnel Guidance: Upon arrival at the fire scene, the robot uses a multispectral imager 12 and an infrared thermal imager 9 to identify the fire source properties, such as a metal fire, and automatically selects the corresponding extinguishing agent, such as Class D dry powder, and sprays it from upwind. If trapped personnel are detected, the robot guides them to evacuate along a dynamically planned escape route via LED screens, lasers, or voice commands.

[0133] Autonomous adaptation and closed-loop feedback: During movement, the robot verifies the environment in real time through visual perception. If the actual terrain / hazard deviates significantly from the prediction, it replans its path locally to bypass obstacles and sends the information back to the platform to update the global map. Multispectral material recognition results directly drive decisions such as switching movement modes and selecting fire extinguishing agents, enabling precise rescue.

[0134] In summary, the intelligent firefighting robot system and control method described in this application enable the firefighting robot to autonomously adapt to complex terrains such as oil spills and staircases, and accurately identify the nature of the fire source, thereby automatically selecting the optimal movement method and extinguishing agent. The system integrates fixed monitoring data with the robot's own perception data to achieve dynamic path planning and personnel guidance, significantly improving its mobility in complex fire scenes, firefighting accuracy, and personnel rescue efficiency, while effectively reducing the risks to rescue personnel.

[0135] In the description of this specification, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0136] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0137] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A control method of a fire fighting robot, characterized by, The method comprises the following steps: Obtaining wide-area environment image information through a fixed monitoring camera network deployed in the work area, and obtaining close-range environment image information through a mobile perception system carried by the fire-fighting robot; Identifying the wide-area environment image information and the close-range environment image information, and extracting fire scene environment information including at least terrain feature information and material attribute information; Based on the terrain feature information and the material attribute information, making a linkage decision to generate a control instruction set, the control instruction set including a first instruction for controlling robot movement mode switching and a second instruction for controlling a fire extinguishing strategy; The control instruction set is sent to the fire-fighting robot to control it to perform corresponding actions.

2. The method of claim 1, wherein, The linkage decision step specifically includes: When it is identified from the image information that there is liquid oil or chemical leakage on the ground, generating a first instruction to switch the robot movement mode from wheeled mode to multi-legged mode; When it is identified from the image information that the fire source is a metal fire, generating a third instruction to select a D-class dry powder extinguishing agent as the second instruction, and generating a third instruction to plan a path approaching the fire source from the upwind direction.

3. The method of claim 2, wherein, The linkage decision is also based on pre-acquired work area space information to generate a pre-planned path containing at least one mode switching point as part of the control instruction set; The position of the mode switching point is determined according to the terrain feature information and / or the material attribute information.

4. An intelligent firefighting robot system characterized by, It comprises: A control platform and at least one fire-fighting robot; The control platform comprises a communication module and a linkage decision engine for generating a control instruction set according to received environment information containing work area space information; The fire-fighting robot comprises: A mobile chassis (1); A chassis frame (2) installed on the bottom surface of the chassis frame (2) through a lifting drive device (3), and four sets of drive wheels (4) arranged on the chassis frame (2); Six mechanical legs (5) arranged around the hexagonal mobile chassis (1), each mechanical leg (5) comprising a leg frame (501), a first joint (502), a second joint (503), a longitudinal adjustment mechanism (504), a transverse adjustment mechanism (505), and multiple sets of folding and unfolding drives (506); A visual perception module for real-time environment verification of the fire-fighting robot; A fire extinguishing execution module comprising a fire spraying assembly (6), a mixing device (7), multiple sets of agent warehouses, multiple connection pipes (8), and multiple electromagnetic valves, wherein the input end of the mixing device (7) is in communication with the multiple sets of agent warehouses through the multiple connection pipes (8), the input end of the mixing device (7) is connected with a water source through a connection pipe (8), the output of the mixing device (7) is connected with the input of the fire spraying assembly (6), and the input port of each connection pipe (8) is provided with a corresponding electromagnetic valve; A main controller in communication connection with the linkage decision engine, the lifting drive device (3), the drive wheels (4), the folding and unfolding drives (506) of the mechanical legs (5), the visual perception module, and the electromagnetic valves. The main controller is configured to: in response to the first instruction from the linkage decision engine, control the lifting driving device (3) to act in cooperation with the retracting and driving device (506) of the mechanical leg (5), so that the driving wheel (4) contacts the ground with one of the mechanical legs (5) to switch the moving mode; in response to the second instruction, control the corresponding electromagnetic valve to open.

5. The system of claim 3, wherein, The visual perception module comprises: a multispectral imager (12) for collecting spectral image data of a target; an infrared thermal imager (9) for collecting infrared radiation image data of a scene, the multispectral imager (12) and the infrared thermal imager (9) being installed side by side on a pan-adjustable gimbal; a forward main camera (10) installed at a central axis position on the front part of the fire-fighting robot vehicle body for acquiring the environment in front of the robot; at least two lateral auxiliary cameras (11) symmetrically installed at positions close to the front and rear of the vehicle body for acquiring images of the environment on the sides of the robot.

6. The system of claim 3, wherein, Each of the mechanical legs (5) comprises: a leg support (501) for supporting the robot; a first joint (502) having one end hinged to the proximal end of the leg support (501); a second joint (503) having one end hinged to the other end of the first joint (502); a longitudinal adjustment mechanism (504) having an adjustment shaft hinged to the other end of the second joint (503); a transverse adjustment mechanism (505) having an adjustment shaft connected to the adjustment seat of the longitudinal adjustment mechanism (504) a plurality of retracting and driving devices (506) provided on the rotation shafts of the first joint (502), the second joint (503), the longitudinal adjustment mechanism (504), and the transverse adjustment mechanism (505).

7. The system of claim 3, wherein, The control platform further comprises a fixed monitoring camera network; The fixed monitoring camera network comprises a plurality of network cameras with gimbals, the network cameras being in communication connection with the control platform, the environmental information comprising wide-area visual information provided by the fixed monitoring camera network, the fixed monitoring camera network comprising a plurality of network cameras with gimbals, the network cameras being in communication connection with the control platform; a path planning engine, the work area spatial information comprising a pre-constructed park three-dimensional map, the path planning engine accessing the pre-constructed park three-dimensional map, wherein the path planning engine is configured to: fuse the static spatial information of the park three-dimensional map and the real-time wide-area visual information provided by the fixed monitoring camera network, and plan a pre-planned path containing at least one mode switching point for the fire-fighting robot according to the decision result of the linkage decision engine, the position of the mode switching point being determined based on the terrain feature information and / or the material attribute information; the control instruction set comprising the pre-planned path.

8. The system of claim 6, wherein, The main controller is further configured to perform real-time environment sensing through the visual sensing module during movement of the firefighting robot along the pre-planned path; If a deviation between the sensed real-time terrain and material information and pre-judgment information on which the pre-planned path is generated exceeds a safety threshold, the current path is interrupted, and a segment of path is re-planned locally based on the real-time sensing information to bypass an unpre-judged obstacle or dangerous area; Meanwhile, the deviation information and the re-planned path segment are returned to the control platform.

9. The system of claim 3, wherein, The multispectral imager (12) is configured to acquire spectral images of a target in a visible light and at least one non-visible light band, and the linked decision engine identifies material attributes by analyzing characteristic spectral lines of the spectral images; The multispectral imager (12) and the linked decision engine are cooperatively configured to identify material attributes including at least one of the following types by analyzing spectral images: Ground liquid pollutant attributes, whose identification results are used to trigger a switching instruction to a multi-legged movement mode; Combustion chemical attributes, whose identification results are used to trigger a selection instruction of a specific fire extinguishing agent; Building structure material attributes, whose identification results are used to trigger a passability decision instruction of impact or bypass.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The program, when executed by a processor, implements the method of any one of claims 1-3.

Images