Underground space pilot fire extinguishing system based on heterogeneous robot cooperation

Abstract

The invention provides an underground space pilot fire extinguishing system based on heterogeneous robot cooperation, and the system comprises a reconnaissance terminal which is a reconnaissance robot and is used for entering a fire scene, and generating a dynamic semantic map containing environment characteristics and fire source coordinates through a multi-modal sensing module; the execution end is an intelligent fire fighting truck, is provided with an auxiliary driving module and is used for receiving the navigation path and executing fire extinguishing operation according to self motion constraint; the cooperative control layer is configured to perform feasibility filtering on the environment characteristics collected by the reconnaissance end according to the physical size parameters of the execution end and generate a navigation path exclusive to the execution end; when the execution end is blocked in the advancing process or the sensor of the execution end loses efficacy due to environmental smoke, the execution end sends a support instruction to the reconnaissance end; and after receiving the support instruction, the reconnaissance end leaves the current position and actively returns to the position of the execution end for convergence guidance.

Description

Technical Field

[0001] This invention relates to a navigation firefighting system, specifically to an underground space navigation firefighting system based on heterogeneous robot collaboration. Background Technology

[0002] With the rapid popularization of new energy vehicles and the continuous expansion of urban underground spaces, the difficulty of fighting fires in underground parking garages (especially fires caused by thermal runaway of power batteries) is increasing. Underground spaces are typically characterized by complex terrain, narrow passages, and severe communication signal shielding. Once a fire breaks out, it can easily generate large amounts of dense smoke and high-temperature toxic gases in a short period of time, posing a significant threat to the lives of traditional manual firefighters. Therefore, using robots to replace firefighters in entering underground spaces to perform reconnaissance and firefighting tasks has become an important development trend in the field of smart firefighting.

[0003] Currently, existing firefighting robots and their navigation and coordination systems mainly suffer from the following technical solutions and limitations: 1. The contradiction between load capacity and passability is difficult to reconcile. Existing indoor firefighting robot solutions mostly employ a single reconnaissance robot (robot dog) to handle both reconnaissance and firefighting, or two identical reconnaissance robots working in a cohesive "reconnaissance end + execution end" configuration. While reconnaissance robots are highly flexible and have good maneuverability, their size and load-bearing capacity are limited, and the extinguishing agents they carry are insufficient to handle large-scale fires involving rapidly burning new energy vehicles in underground parking garages. On the other hand, using large wheeled or tracked intelligent fire trucks equipped with large-capacity extinguishing agents presents challenges in narrow or obstacle-filled underground parking garages, as they have poor maneuverability and are prone to getting stuck.

[0004] 2. Lack of "physical space constraint mapping" between heterogeneous robots

[0005] To combine the high mobility of robotic dogs with the powerful firefighting capabilities of large fire trucks, some cutting-edge solutions are exploring a heterogeneous collaborative model: "robotic dog reconnaissance and mapping + fire truck receiving the map and extinguishing the fire." However, existing collaborative navigation methods typically simply package and send the global environment map (such as a 3D point cloud map or a 2D raster map) generated by the robotic dog directly to the execution end. Due to the significant heterogeneous gap between robotic dogs and large fire trucks in terms of physical dimensions (length, width, height) and kinematic models (turning radius, crossing height), narrow gaps, steps, or low-lying pipes that robotic dogs can easily pass through often become impassable for large fire trucks. Existing systems lack map trimming and feasibility assessment logic based on the physical envelope of the execution end, making it extremely easy for fire trucks to get stuck in dead ends or collide with obstacles when following the robotic dog's map.

[0006] 3. Blindness caused by dense smoke and "navigation paralysis" due to communication disruptions.

[0007] When fires occur in underground parking garages, the scene is often accompanied by extremely high concentrations of smoke and water mist generated by firefighting. Existing fire-fighting robot navigation systems heavily rely on traditional LiDAR and visual SLAM algorithms. These sensors are prone to light scattering and signal attenuation in dense smoke environments, causing large fire trucks to become momentarily "blinded" and lose their location. In addition, underground spaces inherently lack GPS and network signals. Once a fire truck becomes lost or obstructed in dense smoke, its communication link with the reconnaissance and command centers is easily interrupted. The existing one-way map distribution mechanism cannot provide real-time dynamic guidance and escape support in extreme environments, ultimately leading to the failure of the entire fire-fighting and rescue mission.

[0008] In summary, how to resolve path size mapping conflicts during heterogeneous robot collaboration, and how to ensure the accurate navigation and safe movement of large fire trucks in extreme and harsh environments such as blinding due to dense smoke and signal jamming, are the core technical challenges that urgently need to be addressed in this field. Summary of the Invention

[0009] In view of the above-mentioned technical defects and pain points in the existing technology, the purpose of this invention is to provide an underground space navigation and fire extinguishing system and method based on heterogeneous robot collaboration.

[0010] This invention aims to solve the problem of path mapping failure caused by differences in physical size and motion model when heterogeneous robots (small quadruped robot dogs and large intelligent fire trucks) work together. At the same time, it overcomes the fatal defects of large fire trucks' own sensors being blinded and navigation paralyzed due to dense smoke, water mist and network signal blocking when fires occur in underground spaces, thereby achieving accurate, efficient and safe fire fighting in complex underground spaces.

[0011] To achieve the above objectives, the present invention provides an underground space navigation and firefighting system based on heterogeneous robot collaboration, comprising: Reconnaissance end: This is a reconnaissance robot used to enter the fire scene and generate a dynamic semantic map containing environmental features and fire source coordinates through a multimodal perception module; The execution end: It is an intelligent fire truck equipped with an auxiliary driving module, which is used to receive navigation paths and perform fire extinguishing operations according to its own motion constraints; Collaborative control layer: It is configured to perform feasibility filtering on the environmental features collected by the reconnaissance terminal based on the physical size parameters of the execution terminal, and generate a navigation path exclusive to the execution terminal; The system is also equipped with a turnaround and support module: when the execution end is obstructed during its journey or its own sensors fail due to environmental smoke, the execution end sends a support command to the reconnaissance end; after receiving the support command, the reconnaissance end leaves its current position and actively turns back to the location of the execution end for rendezvous and guidance.

[0012] Furthermore, when generating the navigation path, the collaborative control layer automatically marks the fire truck's "no-entry zone" and "suggested passage zone" in the generated dynamic semantic map based on the length, width, height, and turning radius parameters of the execution end.

[0013] Furthermore, a dual fire alarm confirmation mechanism is configured. The reconnaissance terminal goes to the scene after sensing smoke through its own camera or receiving fire alarm information from the fire control system. Only after the reconnaissance terminal or firefighters confirm that it is a real fire alarm will the execution terminal go to the scene to extinguish the fire according to the navigation path, and will not respond to unconfirmed single false fire alarms.

[0014] Furthermore, after the reconnaissance terminal and the execution terminal converge, they establish a local communication handshake protocol wirelessly to synchronize data; the wireless communication method includes at least one of WiFi4, WiFi5, WiFi6 or microwave communication.

[0015] Furthermore, after establishing a communication handshake, the reconnaissance terminal probes in front of the execution terminal and determines whether the roads on both sides are passable based on the size parameters of the execution terminal; if it is determined that the roads are not passable, the reconnaissance terminal guides the execution terminal to take a reverse or detour alternative route to the fire point.

[0016] This invention also provides a method for navigating and extinguishing fires in underground spaces based on heterogeneous robot collaboration, comprising the following steps: S1) The reconnaissance terminal detects the fire scene and, after confirming the fire, sends a dispatch command to the execution terminal; S2) The reconnaissance terminal generates a dynamic semantic map, and the collaborative control layer executes a map trimming algorithm adapted to the envelope of the execution terminal based on the physical size parameters of the execution terminal to generate a navigation path with "passable" determination logic specific to the execution terminal. S3) The execution terminal receives the navigation path and starts the assisted driving module to move towards the fire source; S4) When the execution terminal is obstructed by smoke or encounters an obstacle, it sends a support command to the reconnaissance terminal; after receiving the command, the reconnaissance terminal actively returns to the location of the execution terminal. S5) After the reconnaissance terminal and the execution terminal meet, they establish a communication handshake and dynamically guide the execution terminal by replanning an alternative route until they reach the fire extinguishing point.

[0017] Furthermore, in step S2, the map trimming algorithm specifically involves the following: when the robot dog is building the map, it automatically marks the "no-passage zone" of the execution end in the point cloud map; in addition to the route, the data transmitted to the execution end also includes the judgment logic for whether the fire truck can pass.

[0018] Furthermore, a power self-check step is included before step S1: the reconnaissance terminal and the execution terminal are set with a power threshold. When the power is low, the reconnaissance terminal returns to the execution terminal or a nearby charging dock for charging, and the execution terminal sends a notification requesting the administrator to connect to a power source for charging.

[0019] Furthermore, in step S5: if the reconnaissance terminal determines that the road ahead is impassable and cannot be bypassed, the reconnaissance terminal guides the execution terminal to perform a reverse action and reach the fire extinguishing point along the alternative route of the new instruction to carry out fire extinguishing.

[0020] The present invention has the following beneficial effects: 1. Pioneering "Envelope Map Pruning" Algorithm to Completely Solve the Path Gap Between Heterogeneous Robots: This invention breaks away from the traditional crude "reconnaissance-map transmission-execution-follow-the-path" model. When the reconnaissance end (robot dog) is building the map, it fully considers the length, width, height, and turning radius of the execution end (fire truck), eliminating narrow areas (such as steps, fallen ventilation ducts, etc.) that the robot dog can pass through but the fire truck would get stuck in advance. The map transmitted to the fire truck has built-in "passability" judgment logic, greatly improving the safety rate of large fire-fighting equipment in complex terrain.

[0021] 2. A unique "return and dynamic navigation" mechanism endows fire trucks with "guide dog" functionality: In the extreme environment of dense smoke and high water mist during a fire in an underground parking garage, conventional large vehicle lidar and vision can easily become momentarily blinded. This invention innovatively introduces a "request for help - active return - local handshake" mechanism. After receiving a lost navigation command from the fire truck, the robot dog actively returns to provide assistance and acts as a "guide dog" ahead, using its multimodal perception to conduct close-range pathfinding and dynamic guidance (even guiding the vehicle to reverse and escape from difficulties), perfectly solving the navigation paralysis problem of heavy fire trucks in harsh environments.

[0022] 3. Physical collaboration and local handshake communication enhance practical application: It enables integrated carrying and automatic power replenishment of the reconnaissance and execution ends; and when meeting in an underground network-free environment, it establishes a local communication handshake through a dedicated microwave or WiFi protocol, ensuring high bandwidth and low latency for image transmission and point cloud data transmission.

[0023] 4. Dual confirmation mechanism for real and false fire alarms to avoid resource waste: By assigning confirmation weights to the robot dog "sentinel" and "firefighter" in the preceding reconnaissance, false alarms and mis-reports of fire systems commonly found in underground parking garages are effectively filtered out, ensuring that intelligent fire trucks "do not deploy unless there is a real fire", thus optimizing the system's energy management and deployment efficiency. Attached Figure Description

[0024] Figure 1 This is the flowchart of this method.

[0025] Figure 2 This is a schematic diagram of a module in one embodiment of the present invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention.

[0027] This invention provides a heterogeneous robot collaborative underground space navigation and firefighting system, comprising: Reconnaissance end: This is a reconnaissance robot used to enter the fire scene and generate a dynamic semantic map containing environmental features and fire source coordinates through a multimodal perception module; The execution end: It is an intelligent fire truck equipped with an auxiliary driving module, which is used to receive navigation paths and perform fire extinguishing operations according to its own motion constraints; Collaborative control layer: It is configured to perform feasibility filtering on the environmental features collected by the reconnaissance terminal based on the physical size parameters of the execution terminal, and generate a navigation path exclusive to the execution terminal; The system is also equipped with a turnaround and support module: when the execution end is obstructed during its journey or its own sensors fail due to environmental smoke, the execution end sends a support command to the reconnaissance end; after receiving the support command, the reconnaissance end leaves its current position and actively turns back to the location of the execution end for rendezvous and guidance.

[0028] Furthermore, when generating the navigation path, the collaborative control layer automatically marks the fire truck's "no-entry zone" and "suggested passage zone" in the generated dynamic semantic map based on the length, width, height, and turning radius parameters of the execution end.

[0029] Furthermore, a dual fire alarm confirmation mechanism is configured. The reconnaissance terminal goes to the scene after sensing smoke through its own camera or receiving fire alarm information from the fire control system. Only after the reconnaissance terminal or firefighters confirm that it is a real fire alarm will the execution terminal go to the scene to extinguish the fire according to the navigation path, and will not respond to unconfirmed single false fire alarms.

[0030] Furthermore, after the reconnaissance terminal and the execution terminal converge, they establish a local communication handshake protocol wirelessly to synchronize data; the wireless communication method includes at least one of WiFi4, WiFi5, WiFi6 or microwave communication, and may also include methods such as StarFlash and Bluetooth.

[0031] Furthermore, after establishing a communication handshake, the reconnaissance terminal probes in front of the execution terminal and determines whether the roads on both sides are passable based on the size parameters of the execution terminal; if it is determined that the roads are not passable, the reconnaissance terminal guides the execution terminal to take a reverse or detour alternative route to the fire point.

[0032] like Figure 1 As shown, the present invention also provides a method for navigating and extinguishing fires in underground spaces based on heterogeneous robot collaboration, comprising the following steps: S1) The reconnaissance terminal detects the fire scene and, after confirming the fire, sends a dispatch command to the execution terminal; S2) The reconnaissance terminal generates a dynamic semantic map, and the collaborative control layer executes a map trimming algorithm adapted to the envelope of the execution terminal based on the physical size parameters of the execution terminal to generate a navigation path with "passable" determination logic specific to the execution terminal. S3) The execution terminal receives the navigation path and starts the assisted driving module to move towards the fire source; S4) When the execution terminal is obstructed by smoke or encounters an obstacle, it sends a support command to the reconnaissance terminal; after receiving the command, the reconnaissance terminal actively returns to the location of the execution terminal. S5) After the reconnaissance terminal and the execution terminal meet, they establish a communication handshake and dynamically guide the execution terminal by replanning an alternative route until they reach the fire extinguishing point.

[0033] Furthermore, in step S2, the map trimming algorithm specifically involves the following: when the robot dog is building the map, it automatically marks the "no-passage zone" of the execution end in the point cloud map; in addition to the route, the data transmitted to the execution end also includes the judgment logic for whether the fire truck can pass.

[0034] Furthermore, a power self-check step is included before step S1: the reconnaissance terminal and the execution terminal are set with a power threshold. When the power is low, the reconnaissance terminal returns to the execution terminal or a nearby charging dock for charging, and the execution terminal sends a notification requesting the administrator to connect to a power source for charging.

[0035] Furthermore, in step S5: if the reconnaissance terminal determines that the road ahead is impassable and cannot be bypassed, the reconnaissance terminal guides the execution terminal to perform a reverse action and reach the fire extinguishing point along the alternative route of the new instruction to carry out fire extinguishing.

[0036] like Figure 2 As shown, the present invention will now be described in detail with reference to embodiments.

[0037] Example 1: Overall System Architecture and Physical Co-design

[0038] This embodiment provides a heterogeneous robot collaborative underground space navigation and firefighting system, mainly applicable to environments such as underground parking garages that are prone to generating dense smoke, signal jamming, and complex terrain. The system mainly includes a reconnaissance terminal, an execution terminal, a collaborative control layer, and a turnaround and response module.

[0039] 1. Physical integration of the reconnaissance end (quadruped robot dog) and the execution end (intelligent fire truck)

[0040] The reconnaissance unit is a quadruped robot dog equipped with a multimodal perception module (including a camera, infrared thermal imaging, lidar, or millimeter-wave radar). In non-fire alarm situations, the reconnaissance unit serves as a cooperative accessory to the execution unit, and can be attached to or mounted on the execution unit's vehicle body for hibernation, vehicle-mounted operation, and automatic power replenishment. When the battery is low and the unit is not in the vehicle, it can also find the nearest independent charging station for charging.

[0041] 2. Heterogeneous power collaborative management mechanism

[0042] Reconnaissance Terminal Power Management: The robot dog can be programmed to patrol at a frequency of once every hour, achieving blind-spot-free coverage of the underground parking garage. During patrols, when the system detects that the robot dog's battery level is below 40%, the robot dog will automatically return to the fire truck or find the nearest independent charging station to recharge.

[0043] Battery management at the execution end: The intelligent fire truck performs a daily self-check of its battery level. When the battery level drops below 40%, the system sends a notification to the administrator, requesting manual charging. In principle, fire trucks are not allowed to remain in a charging state for extended periods to prevent the risk of explosion due to overcharging.

[0044] Example 2: Fire Alarm Dual Authority Confirmation and Envelope Mapping Method

[0045] Based on the above system, this embodiment elaborates on the specific implementation process of the system from detecting a fire to planning a path.

[0046] Step S1: Fire Detection and Dual Rights Confirmation

[0047] The fire control system in the underground parking garage frequently experiences false alarms. This system is designed so that fire trucks do not blindly respond to a single system fire alarm signal.

[0048] When the fire alarm sounds, the robot dog is dispatched to the scene to investigate; alternatively, it may detect smoke during routine autonomous patrols using its visual camera. Upon arrival, the robot dog uses multimodal sensors to determine if it is a real fire. If confirmed as a real fire (or verified by firefighters at the control center via footage transmitted from the robot dog), the robot dog sends the confirmed fire alarm command and coordinates to the fire truck and control center, at which point the fire truck is activated and dispatched.

[0049] Step S2: Envelope Mapping Based on Heterogeneous Sizes

[0050] At the fire scene, the robot dog used a multimodal perception module to perform SLAM mapping, generating a dynamic semantic map that includes the location of the fire source, obstacles, and terrain features.

[0051] Because the robot dog is small, it can pass through narrow gaps or steps, while the fire truck is large and has a long turning radius. At this point, the collaborative control layer intervenes: The collaborative control layer obtains the physical dimensions (length, width, height) and kinematic constraints (turning radius, chassis height, etc.) of the fire truck.

[0052] The system inputs these parameters into the dynamic semantic map and executes the "envelope map pruning algorithm". That is, in the point cloud map generated by the robot dog, it automatically filters out areas that fire trucks cannot pass through and marks them as "no passage zones"; and marks areas that meet the conditions for fire truck passage as "suggested passage zones".

[0053] The data transmitted from the robot dog to the fire truck includes not only the route, but also the logic for determining whether the road is passable based on the size of the fire truck.

[0054] Specifically, the underlying execution logic of the map trimming algorithm is as follows: First, the reconnaissance unit (robot dog) acquires 3D point cloud data of the environment using LiDAR and converts it into a 3D voxel map. The collaborative control layer extracts the core physical dimension parameters of the execution unit (intelligent fire truck), including vehicle width (W), vehicle height (H), vehicle length (L), and minimum turning radius (R).

[0055] Secondly, the system performs height collision detection: in the 3D voxel map, all point cloud obstacles (such as suspended ventilation ducts, low beams, etc.) with height in the range of 0 to H are extracted with the ground as the reference, and they are projected downwards onto the 2D grid map to form the basic obstacle boundary.

[0056] Next, the system executes the two-dimensional dilation algorithm: using (W / 2 + safety margin) as the dilation radius, it expands the basic obstacle boundaries in the two-dimensional grid map outward; for the corner areas in the map, the system further introduces the minimum turning radius (R) and vehicle length (L) to calculate the scanning envelope and remove the overlapping areas of the inner contour of the turn.

[0057] Finally, after the above-mentioned "3D height check + 2D boundary expansion + turning envelope calculation", the continuous blank areas in the grid map that are not covered by obstacles and expansion layers are marked as absolutely safe "recommended passage zones". The data transmitted from the reconnaissance end to the execution end is this processed safe grid map, thus fundamentally avoiding the jamming phenomenon caused by size blind spots at the execution end.

[0058] Step S3: Fire truck dispatch and assisted driving

[0059] After receiving the authorization command and exclusive envelope navigation path sent by the robot dog, the fire truck activates the assisted driving module and autonomously moves towards the fire source along the marked "recommended passageway".

[0060] Example 3: Turnaround and Dynamic Navigation in Extreme Environments

[0061] In real-world fire scenarios in underground parking garages, as the fire spreads, the scene is often filled with thick smoke and water vapor, and the underground environment lacks network signals. This embodiment specifically illustrates the escape and navigation mechanism in situations where fire trucks are "blinded."

[0062] Step S4: Fire truck seeks assistance despite being blind and robot dog actively returns.

[0063] When a fire truck encounters dense smoke that causes its lidar and vision cameras to malfunction and make it unable to detect road conditions, or when it encounters unexpected obstacles (such as other escape vehicles or fallen pipes) that cause it to get stuck, the fire truck sends a "lost" or "request for support" command to the robot dog.

[0064] Upon receiving the instruction, the robot dog (return and rendezvous module) immediately leaves the fire site it is currently exploring, actively turns back along the original route, and rushes to the specific location where the fire truck is currently stationed to rendezvous.

[0065] To address the issue of the reconnaissance and execution ends being unable to find each other in dense underground smoke environments, in this embodiment, when the execution end is blinded and rendered immobile by its own vision and lidar, it will immediately package its last accurate odometer coordinates and IMU (Inertial Measurement Unit) pose data and transmit them to the reconnaissance end ahead via low-frequency radio or microwave communication with strong penetration capabilities.

[0066] Meanwhile, the execution unit's front end integrates a UWB (Ultra-Wideband) positioning beacon module, which continuously broadcasts signals when stationary. Upon receiving a distress call, the reconnaissance unit switches to "Homing Mode": it uses its onboard millimeter-wave radar (which has strong penetrating power in dense smoke and dust) for dynamic obstacle avoidance and uses its UWB receiving antenna to measure the time difference of arrival (TDOA) and signal strength indication (RSSI) of the signal emitted by the execution unit's beacon in real time. By calculating the distance and azimuth of the UWB signal, the reconnaissance unit can accurately locate the execution unit's front end in a zero-visibility, dense smoke environment and complete a close-range physical rendezvous.

[0067] Step S5: Local Area Communication Handshake and "Guide Dog"-Style Dynamic Navigation

[0068] Local handshake: After the robot dog and the fire truck meet, in order to overcome the signal shielding in the underground garage, the two establish a local communication handshake protocol through wireless means (such as WiFi 4, WiFi 5, WiFi 6 protocols or microwave communication) to ensure that the pathfinding data (video stream, radar point cloud) collected by the robot dog can be synchronized to the fire truck with millisecond-level low latency.

[0069] Dynamic Pathfinding and Guidance: After establishing a connection, the robot dog drives in front of the fire truck, acting as a "guide dog." The robot dog uses its infrared thermal imaging and millimeter-wave radar to penetrate dense smoke, detect the road ahead in real time, and combine this with the envelope size of the fire truck to determine whether there is enough space on both sides for the fire truck to pass.

[0070] Execution of Escape and Dynamic Navigation: If the path ahead is deemed passable, the robot dog maintains a safe handshake distance with the fire truck and guides it forward. If the path ahead is deemed impassable due to an obstacle, the robot dog will plan an alternative route for the fire truck (reversing, turning, or detouring). During this process, a dynamic navigation mechanism based on an "electronic virtual traction rope" is established between the reconnaissance and execution ends. The specific implementation method is as follows: The reconnaissance terminal maintains a safe forward distance of 3 to 5 meters from the execution terminal. During this process, the reconnaissance terminal continuously fuses infrared thermal imaging and millimeter-wave radar to perceive the road conditions ahead and generates locally unobstructed "dynamic virtual waypoints" at extremely high frequencies (such as 10Hz).

[0071] The reconnaissance terminal uses a high-bandwidth, low-latency local area network such as WiFi 6 to send the aforementioned dynamic virtual waypoint coordinate sequence, along with the unobstructed infrared thermal imaging video stream from the front, to the execution terminal in real time.

[0072] After receiving the virtual waypoint sequence, the assisted driving module of the execution end uses either the PurePursuit algorithm or the Model Predictive Control (MPC) algorithm to control the throttle and steering of the underlying drive-by-wire chassis in real time, using the waypoints issued by the reconnaissance end as the target to follow. When the reconnaissance end determines that there is a dead end ahead, it will stop generating forward waypoints and instead generate a smooth "reverse guide trajectory" for the execution end based on the rear space. The execution end will automatically reverse along the guide trajectory to get out of trouble until a valid path to the fire source is re-planned, and finally successfully guided to the best fire extinguishing point to carry out firefighting operations.

[0073] Post-fire maintenance: After the firefighting operation was completed, the site manager notified the manufacturer to dispatch a professional cleaning vehicle to the underground garage. The waste liquid (including hazardous waste liquids such as hydrogen sulfide) recovered from the fire truck was intercepted and treated, new fire extinguishing liquid was added, and the pipelines, truck body, and chassis were cleaned and maintained on-site to quickly restore the system to standby status.

[0074] The embodiments described above are merely further illustrations of the present invention and are not intended to limit the present invention in any other way. The present invention may have many other embodiments. Without departing from the spirit and essence of the present invention, those skilled in the art can make various corresponding modifications and changes based on the present invention, but all such modifications and changes should fall within the protection scope of the present invention.

Claims

1. A heterogeneous robot collaborative underground space navigation and firefighting system, characterized in that, include: Reconnaissance end: This is a reconnaissance robot used to enter the fire scene and generate a dynamic semantic map containing environmental features and fire source coordinates through a multimodal perception module; The execution end: It is an intelligent fire truck equipped with an auxiliary driving module, which is used to receive navigation paths and perform fire extinguishing operations according to its own motion constraints; Collaborative control layer: It is configured to perform feasibility filtering on the environmental features collected by the reconnaissance terminal based on the physical size parameters of the execution terminal, and generate a navigation path exclusive to the execution terminal; The system is also equipped with a turnaround and support module: when the execution end is obstructed during its journey or its own sensors fail due to environmental smoke, the execution end sends a support command to the reconnaissance end; after receiving the support command, the reconnaissance end leaves its current position and actively turns back to the location of the execution end for rendezvous and guidance.

2. The underground space navigation and firefighting system based on heterogeneous robot collaboration according to claim 1, characterized in that: When generating the navigation path, the collaborative control layer automatically marks the fire truck's "no-entry zone" and "suggested passage zone" in the generated dynamic semantic map based on the length, width, height, and turning radius parameters of the execution end.

3. The underground space navigation and firefighting system based on heterogeneous robot collaboration according to claim 1, characterized in that: Equipped with a dual fire alarm confirmation mechanism, the reconnaissance terminal goes to the scene after sensing smoke through its own camera or receiving fire alarm information from the fire control system; only after the reconnaissance terminal or firefighters confirm that it is a real fire alarm will the execution terminal go to the scene to extinguish the fire according to the navigation path, and will not respond to unconfirmed single false fire alarms.

4. The underground space navigation and firefighting system based on heterogeneous robot collaboration according to claim 1, characterized in that: After the reconnaissance terminal and the execution terminal meet, they establish a local communication handshake protocol wirelessly to synchronize data; the wireless communication method includes at least one of WiFi4, WiFi5, WiFi6 or microwave communication.

5. The underground space navigation and firefighting system based on heterogeneous robot collaboration according to claim 4, characterized in that: After establishing a communication handshake, the reconnaissance terminal probes in front of the execution terminal and determines whether both sides of the road are passable based on the size parameters of the execution terminal; If the route is deemed impassable, the reconnaissance terminal guides the execution terminal to take an alternative route, either reversing or taking a detour, to the fire location.

6. A method for navigating and extinguishing fires in underground spaces based on heterogeneous robot collaboration, using the system described in any one of claims 1-5, characterized in that, Includes the following steps: S1) The reconnaissance terminal detects the fire scene and, after confirming the fire, sends a dispatch command to the execution terminal; S2) The reconnaissance terminal generates a dynamic semantic map, and the collaborative control layer executes a map trimming algorithm adapted to the envelope of the execution terminal based on the physical size parameters of the execution terminal to generate a navigation path with "passable" determination logic specific to the execution terminal. S3) The execution terminal receives the navigation path and starts the assisted driving module to move towards the fire source; S4) When the execution terminal is obstructed by smoke or encounters an obstacle, it sends a support command to the reconnaissance terminal; after receiving the command, the reconnaissance terminal actively returns to the location of the execution terminal. S5) After the reconnaissance terminal and the execution terminal meet, they establish a communication handshake and dynamically guide the execution terminal by replanning an alternative route until they reach the fire extinguishing point.

7. The underground space navigation and firefighting method based on heterogeneous robot collaboration according to claim 6, characterized in that, In step S2, the map trimming algorithm specifically involves the following steps: when the robot dog is building the map, it automatically marks the "no-passage zone" of the execution end in the point cloud map; in addition to the route, the data transmitted to the execution end also includes the judgment logic for whether the fire truck can pass.

8. The underground space navigation and firefighting method based on heterogeneous robot collaboration according to claim 6, characterized in that, Before step S1, there is also a power self-check step: the reconnaissance terminal and the execution terminal are set with a power threshold. When the power is low, the reconnaissance terminal returns to the execution terminal or a nearby charging dock for charging, and the execution terminal sends a notification to the administrator requesting the administrator to connect to a power source for charging.

9. The underground space navigation and firefighting method based on heterogeneous robot collaboration according to claim 8, characterized in that, In step S5: if the reconnaissance terminal determines that the road ahead is impassable and cannot be bypassed, the reconnaissance terminal guides the execution terminal to perform a reverse action and reach the fire extinguishing point along the alternative route of the new instruction to carry out fire extinguishing.

Images