Building fire-fighting facility detection system based on intelligent robot autonomous navigation and position matching

Abstract

This invention discloses a building fire protection facility inspection system based on intelligent robot autonomous navigation and position matching, belonging to the field of building fire protection facility inspection technology. The system includes a robot body and integrated on it a multimodal perception module, a large-scale intelligent processing unit, an autonomous navigation and control module, and a power supply module. The autonomous navigation and control module generates path planning information and facility location information, and controls the platform to travel to the indicated location. The multimodal perception module collects multimodal data after the fire protection facilities' actions and transmits it to the large-scale intelligent processing unit. The large-scale intelligent processing unit processes and analyzes the multimodal data and generates an inspection report. This invention drives the inspection behavior and precisely corresponds it to the spatial location through facility location information. Combined with multimodal perception and large-scale intelligent analysis, it achieves fully automated inspection of fire protection facilities, improving inspection accuracy, comprehensiveness, and efficiency, and reducing human interference.

Description

Technical Field

[0001] This invention relates to the field of robot-based building fire protection facility inspection technology, and in particular to a building fire protection facility inspection system based on intelligent robot autonomous navigation and position matching. Background Technology

[0002] Fire protection facility inspection refers to the periodic inspection of the installation quality, functionality, and linkage performance of automatic sprinkler systems, smoke control and exhaust systems, and automatic fire alarm systems in buildings in accordance with national standards. The aim is to ensure that the facilities can be reliably activated in the event of a fire, and it is a core part of building fire safety management.

[0003] Currently, inspections primarily rely on manual methods. Inspectors, using standard equipment and carrying general testing tools, conduct visual inspections, functional tests, and linkage tests on the numerous and dispersed fire protection facilities within buildings. This approach suffers from the following problems: First, it is inefficient and time-consuming, with full-scale inspections of large buildings being time-consuming and limited in frequency. Second, the results are highly subjective and difficult to trace, with differing judgments among personnel and a lack of precise correlation between paper records and the spatial location of facilities. Third, the dispersed spatial distribution of fire protection facilities in the system makes it difficult for personnel to perform simultaneous multi-point operations. Fourth, the inspection data is disconnected from building information and cannot be integrated with maintenance records, making it difficult to consistently meet inspection requirements.

[0004] Although there have been attempts at robot inspection in recent years, most existing solutions use single-machine operation, which can only collect images according to preset paths. They cannot plan paths autonomously, accurately locate operation points, or perform contact operations such as valve rotation and safety pin removal. When multiple machines operate simultaneously, there is a lack of coordination mechanism, which can easily lead to path conflicts. The inspection data still relies on manual export and has not formed a closed loop.

[0005] Therefore, how to provide a fire protection facility detection system that can navigate autonomously, locate precisely, operate by touch, support multi-machine collaboration, and has full data traceability has become a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0006] To address the problems existing in the prior art, the main objective of this invention is to provide a building fire protection facility inspection system based on intelligent robot autonomous navigation and location matching, which can replace manual inspection of fire protection facilities based on autonomous navigation and precise positioning, and realize a fully automated closed loop for the entire process of fire protection facility inspection.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: A building fire protection facility detection system based on intelligent robot autonomous navigation and position matching includes: a robot body, and a multimodal perception module, a large model intelligent processing module, a fire protection facility triggering module, an autonomous navigation and control module, and a power supply module, all integrated on the robot body; The power supply module is used to provide operating power; The autonomous navigation and control module is used to generate path planning information for the preset detection area and matching facility location information, and to control the robot body to move according to the path planning information and to control the robot body to reach the position indicated by the facility location information. The multimodal sensing module is electrically connected to the large model intelligent processing module. It is used to collect multimodal data after the fire-fighting facilities are activated and transmit the multimodal data to the large model intelligent processing module. The large-scale intelligent processing module is used to process and analyze multimodal data to generate fire protection facility inspection reports.

[0008] As a preferred option, a multi-agent cooperative control layer is also included; The multi-agent collaborative control layer is integrated into each robot body, including a distributed region division and role negotiation module; The distributed region division and role negotiation module adopts a distributed consensus algorithm based on geohash. Each robot entity generates a hash value through Geohash encoding based on its own coordinates. Robot entities with the same hash value prefix are grouped into the same collaborative domain. A temporary coordinator is elected by comparing the size of the hash values. The temporary coordinator is responsible for task scheduling and conflict arbitration within the domain, and their term of office is tied to the task cycle within the domain.

[0009] The above scheme enables the robot to autonomously divide collaborative domains and elect temporary coordinators based on the spatial distribution of the building, without the need for intervention from the global central node, thereby improving the flexibility and self-organization capabilities of the system deployment.

[0010] As a preferred embodiment, the multi-agent collaborative control layer further includes a commitment-negotiation-monitoring protocol module; The commitment-negotiation-monitoring protocol module uses distributed ledger technology to record the soft commitments issued by the robot when it accepts a task. The soft commitments include the task completion time, resource usage period, and confidence level. The confidence level is calculated based on the robot's current battery level, health status, and historical fulfillment rate. The temporary coordinator determines resource reservations based on confidence level, and assigns main and backup robots to execute tasks with a confidence level below 70% in parallel. When the robot body is unable to fulfill its commitment, it sends a cancellation message, and the temporary coordinator initiates a compensation allocation algorithm to transfer the unfinished task to a nearby idle robot body.

[0011] The above scheme enables temporary coordinators to dynamically select and redundantly back up task execution entities based on confidence levels, and automatically transfer tasks when the robot entity is unable to perform its duties, thereby improving the reliability and fault tolerance of multi-robot collaborative detection tasks.

[0012] As a preferred embodiment, the multi-agent collaborative control layer further includes a two-level negotiation mechanism module; The two-level negotiation mechanism module includes a first-level negotiation subunit and a second-level negotiation subunit; The first-level negotiation subunit exchanges pose and driving intention through point-to-point communication between robot subjects, and uses a distributed coordination algorithm based on potential game to perform local path replanning. Each robot subject uses its own path cost function as its utility function, and converges to Nash equilibrium through iterative updates. When the first-level negotiation fails or critical resources are involved, the second-level negotiation subunit shall arbitrate based on distributed consensus by the temporary coordinator. The temporary coordinator maintains a local resource reservation table and prioritizes critical resources.

[0013] The above scheme enables multiple robots to resolve conflicts in their travel paths primarily through point-to-point negotiation, with a temporary coordinator arbitrating only when negotiation fails or when critical resources are involved. This reduces the scheduling overhead of the central node and improves the efficiency of intra-domain collaboration.

[0014] As a preferred embodiment, the multi-agent collaborative control layer further includes a central-edge hybrid scheduling module; When the large model intelligent processing module is working normally, the central-edge hybrid scheduling module is responsible for cross-domain task allocation and long-term plan optimization, while the temporary coordinator is responsible for intra-domain task scheduling. When the communication of the large model intelligent processing module is interrupted, the temporary coordinator automatically switches to offline autonomous mode, continues to process tasks within the domain through distributed consensus, and synchronizes data with the large model intelligent processing module after communication is restored.

[0015] The above scheme enables robots to autonomously elect a replacement coordinator in the event of a temporary coordinator failure, and enables robots to maintain collaborative operation through point-to-point communication and Byzantine fault-tolerant consensus algorithm when global communication is interrupted, thereby improving the system's anti-interference capability and robustness.

[0016] As a preferred embodiment, the multi-agent collaborative control layer further includes a distributed collaborative fault-tolerant module; When the temporary coordinator fails, the distributed collaborative fault-tolerant module enables the robot subjects within the domain to elect a new temporary coordinator through a lightweight consensus algorithm. The lightweight consensus algorithm adopts a variant of the Raft consensus algorithm and achieves master node election through election timeout and heartbeat mechanisms. When the communication of the large model intelligent processing module is interrupted, all robot bodies automatically switch to distributed collaborative mode, exchange poses and task progress through point-to-point communication, and maintain collaborative operation using a practical Byzantine fault-tolerant algorithm. When the number of nodes does not exceed 7, PBFT is used, and when the number of nodes exceeds 7, a simplified version of SBFT algorithm is used. After communication is restored, the data is synchronized with the large model intelligent processing module.

[0017] The above scheme enables robots to autonomously elect a replacement coordinator in the event of a temporary coordinator failure, and enables robots to maintain collaborative operation through point-to-point communication and Byzantine fault-tolerant consensus algorithm when global communication is interrupted, thereby improving the system's anti-interference capability and robustness.

[0018] As a preferred embodiment, the robot body includes a multi-degree-of-freedom manipulator, a force-controlled actuator, a tool library, and a vision guidance unit. The tool library includes a variety of end-effector tools, which are selected from at least one of valve operating tools, safety pin removal tools, detection sensors, and image acquisition devices. The multi-degree-of-freedom manipulator automatically selects and replaces the corresponding end effector based on the detection task. Force-controlled actuators are used to operate valve handles with a set torque and monitor torque and displacement feedback, or to pull out safety pins and monitor pull-out force feedback, and automatically reset after operation.

[0019] The above solution enables the robot to automatically change its end effector according to the inspection task, and to complete valve operation and safety pin removal by force control actuator with set torque or pulling force, thereby improving the automation level and operational safety of contact inspection operations.

[0020] As a preferred embodiment, the multimodal sensing module 20 includes a dynamic sensing node; the dynamic sensing node is configured to be temporarily installed on the fire protection facility to be detected or on an adjacent structure by a multi-degree-of-freedom manipulator; the dynamic sensing node includes one or more of a pressure sensor, a flow sensor, an image acquisition unit, a temperature and humidity sensor, a smoke concentration sensor, and an electrical parameter sensor; the dynamic sensing node is equipped with a microcontroller, a wireless communication unit, and a power supply unit.

[0021] The above solution enables the robot to temporarily deploy wireless sensing nodes at the inspection site, allowing for multi-point, long-term, or parallel data collection of designated fire protection facilities, thereby expanding the detection and sensing range and parallel data collection capabilities.

[0022] As a preferred option, the autonomous navigation and control module includes a mobile chassis, an environmental perception unit, and a positioning unit; The mobile chassis adopts an omnidirectional moving structure; Both the environmental sensing unit and the positioning unit are fixedly connected to the mobile chassis; The environmental perception unit is used to collect environmental data around the robot body, and the localization unit is used to collect the pose data of the robot body. The environmental perception unit and the positioning unit are electrically connected to the autonomous navigation and control module, and transmit the collected environmental data and pose data to the autonomous navigation and control module.

[0023] The above solution enables the robot body to integrate environmental perception and positioning units on an omnidirectional mobile chassis, providing real-time environmental and pose data for the autonomous navigation and control module, thereby improving the robot's positioning accuracy and mobility in complex building environments.

[0024] As a preferred option, the large model intelligent processing module also includes a data association subunit, a hierarchical detection and judgment subunit, and a report generation subunit; The data association subunit is used to associate detection data with facility identifiers, timestamps, and robot identifiers, and to form an immutable detection chain through distributed ledger technology; The hierarchical detection and judgment subunit is used to sequentially perform existence detection, compliance detection, responsiveness detection and linkage detection, and to re-examine the detection data according to the three-level judgment logic; The report generation subunit is used to summarize the test data and generate a test report.

[0025] The above solution enables the immutable association of test data with facility identification, timestamps, and robot identification. It also improves the accuracy and traceability of test results through hierarchical test judgment and three-level re-inspection logic. The report generation subunit automatically outputs test reports that meet the specifications.

[0026] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) In this invention, by integrating the autonomous navigation and control module, the multimodal perception module, the large model intelligent processing module and the fire protection facility triggering module set on the robot body, a complete automated closed loop from autonomous navigation, precise positioning, contact operation, multimodal data acquisition to intelligent analysis and report generation is constructed, realizing unmanned detection of building fire protection facilities throughout the entire process, and significantly improving detection efficiency and data reliability.

[0027] (2) In this invention, through the distributed area division and role negotiation module and the commitment-negotiation-monitoring protocol module in the multi-agent collaborative control layer, multiple robot subjects can autonomously divide the collaborative domain according to the building fire protection zone boundary and elect a temporary coordinator. The soft commitment confidence is calculated based on the current power, health status and historical performance rate of the robot subject. The temporary coordinator reserves resources according to the confidence and assigns the main and backup robot subjects to perform low confidence tasks in parallel. When the robot subject cannot perform the task, the task is transferred to the nearby idle robot subject through the compensation allocation algorithm. This solves the problems of difficulty in multi-point synchronous linkage, uneven task allocation and task interruption caused by single-machine failure in the traditional manual and single-machine detection mode, and realizes the reliability of task scheduling and execution continuity of multi-robot collaborative operation.

[0028] (3) In this invention, through the two-level negotiation mechanism module, the central-edge hybrid scheduling module and the distributed collaborative fault-tolerant module in the multi-agent collaborative control layer, the robot subjects exchange poses and driving intentions through point-to-point communication and use a game theory-based distributed coordination algorithm for local path replanning. When the first-level negotiation fails or involves key resources, the temporary coordinator arbitrates based on distributed consensus and maintains a local resource reservation table to prioritize key resources. When the central platform communication is interrupted, the temporary coordinator automatically switches to offline autonomous mode and continues to process tasks within the domain through distributed consensus. When the temporary coordinator fails, the robot subjects within the domain elect a new temporary coordinator through a lightweight consensus algorithm. When the central platform communication is interrupted, all robot subjects switch to distributed collaborative mode and maintain collaborative operation through a Byzantine fault-tolerant consensus algorithm. This solves the problems of path conflict, resource contention, single-point dependence of the central platform and system failure when communication is interrupted in the operation of multiple robot subjects, and realizes the high robustness and high availability of the multi-robot subject collaborative system. Attached Figure Description

[0029] Figure 1 This is a functional block diagram of a building fire protection facility detection system based on intelligent robot autonomous navigation and location matching according to an embodiment of the present invention.

[0030] Figure 2 This is a schematic diagram of multi-robot collaboration in a building fire protection facility detection system based on intelligent robot autonomous navigation and location matching according to an embodiment of the present invention.

[0031] Reference numerals: 10, Robot body; 20, Multimodal perception module; 30, Large model intelligent processing module; 40, Fire protection facility triggering module; 50, Autonomous navigation and control module; 60, Power supply module; 70, Report generation and output module; 80, Multi-robot body collaborative architecture. Detailed Implementation

[0032] To better illustrate the objectives, technical solutions, and advantages of the present invention, the specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0033] like Figures 1 to 2 As shown, an embodiment of this application relates to a building fire protection facility detection system based on intelligent robot autonomous navigation and location matching, comprising: a robot body 10, and a multimodal perception module 20, a large-scale intelligent processing module 30, a fire protection facility triggering module 40, an autonomous navigation and control module 50, and a power supply module 60, all integrated on the robot body 10. The power supply module 60 provides operating power. The autonomous navigation and control module 50 generates path planning information for a preset detection area and matching facility location information, controls the robot body 10 to move according to the path planning information, and controls the robot body 10 to reach the position indicated by the facility location information. The multimodal perception module 20, electrically connected to the large-scale intelligent processing module 30, collects multimodal data after the fire protection facility's actions and transmits the multimodal data to the large-scale intelligent processing module 30. The large-scale intelligent processing module 30 processes and analyzes the multimodal data to generate a fire protection facility detection report.

[0034] Based on the building fire protection facility inspection system of this embodiment, during the inspection process, the autonomous navigation and control module 50 generates path planning information and matching facility location information according to the preset inspection area, and controls the robot body 10 to autonomously travel to each fire protection facility location according to the planned path; after arriving at the location, in response to the inspection command, the fire protection facility triggering module 40 triggers the corresponding fire protection facility action according to the positioning information; the multimodal perception module 20 synchronously collects multimodal data such as images, sounds, and pressure after the fire protection facility action, and transmits the data to the large model intelligent processing module 30; the large model intelligent processing module 30 processes and analyzes the received multimodal data, performs existence detection, compliance detection, responsiveness detection, and linkage integrity detection, and finally generates a fire protection facility inspection report.

[0035] Therefore, the building fire protection facility inspection system of this embodiment controls the precise positioning and movement of the robot body 10 in the building, and integrates the automated triggering, full-dimensional data collection, intelligent judgment and report generation of building fire protection facilities in the robot body 10, thus constructing a complete automated closed loop, realizing unmanned inspection of building fire protection facilities throughout the entire process, and significantly improving inspection efficiency and data reliability.

[0036] In this embodiment, the robot body 10 serves as the mobile execution carrier of the system, including the robot body 10 mobile platform and the multimodal perception module 20, the large model intelligent processing module 30, the fire-fighting facility triggering module 40, the autonomous navigation and control module 50, and the power supply module 60 integrated on the platform.

[0037] The robot body 10's mobile platform includes an omnidirectional mobile chassis, an environmental perception unit, and a positioning unit. The omnidirectional mobile chassis integrates a LiDAR, a depth camera, and an inertial measurement unit (IMU) to achieve centimeter-level autonomous navigation and pose estimation. The environmental perception unit, including a LiDAR and a depth camera, is used to collect environmental data around the robot body 10; the positioning unit, including an IMU and an odometry, is used to collect pose data of the robot body 10. The environmental perception unit and the positioning unit are electrically connected to the autonomous navigation and control module 50, and transmit the collected environmental data and pose data to the autonomous navigation and control module 50.

[0038] The power supply module 60 uses a lithium battery pack with a built-in battery management system for real-time monitoring of battery voltage, current, and temperature, and features low-voltage protection and thermal protection. The power supply module 60 is electrically connected to the multimodal sensing module 20, the large-scale intelligent processing module 30, the fire-fighting facility triggering module 40, and the autonomous navigation and control module 50, providing each module with the necessary operating voltage and current.

[0039] In this embodiment, the autonomous navigation and control module 50 is integrated on the robot body 10. It is used to generate path planning information for a preset detection area and matching facility location information, and control the robot body 10 to move according to the path planning information and reach the position indicated by the facility location information. The autonomous navigation and control module 50 includes a global navigation unit and a local avoidance unit.

[0040] The global navigation unit uses laser-synchronized localization and mapping (SLAM) technology combined with wheeled odometers to achieve real-time pose estimation. It combines a fire-fighting facility pose database with a global spatiotemporal occupancy grid and employs the A* algorithm to generate the globally optimal path. During the robot's movement, pre-embedded ultra-wideband positioning base stations or QR codes are identified as absolute anchor points to correct accumulated positioning errors and achieve centimeter-level positioning.

[0041] The local avoidance unit exchanges pose and driving intention in real time through point-to-point communication between the robot bodies 10, and performs local path replanning in combination with the dynamic window method (DWA) to achieve multi-robot cooperative avoidance; it automatically decelerates and issues a voice prompt when it detects people approaching, and triggers an emergency stop in case of emergency.

[0042] In this embodiment, the fire-fighting facility triggering module 40 is integrated into the robot body 10 and electrically connected to the large model intelligent processing module 30. It is used to generate detection commands based on positioning information and trigger the action of fire-fighting facilities, and also has the ability to temporarily deploy sensing nodes. The fire-fighting facility triggering module 40 can be configured as one or more combinations of mechanical triggering unit, electrical signal triggering unit, simulated heat source generation unit, and simulated smoke generation unit.

[0043] In this embodiment, the robot body 10 includes a multi-degree-of-freedom manipulator, a force control actuator, a tool library, and a vision guidance unit.

[0044] The multi-degree-of-freedom (DOF) manipulator is used to perform contact operations such as valve rotation and safety pin removal, as well as the installation and retrieval of dynamic sensing nodes. The tool library includes various end-effectors, selected from at least one of valve operating tools, safety pin removal tools, detection sensors, image acquisition devices, and dynamic sensing nodes. The multi-DOF manipulator automatically selects and replaces the corresponding end-effector based on the detection task. Force-controlled actuators are used to operate valve handles with a set torque and monitor torque and displacement feedback, or to remove safety pins and monitor removal force feedback, or to control clamping force during the installation and retrieval of dynamic sensing nodes, and automatically reset after operation. The vision guidance unit includes an end-effector vision camera or a gimbal camera, used to acquire real-time images of the target fire-fighting facility after the robot body 10 reaches its vicinity. Target detection algorithms are used to identify detection points and installation positions, providing visual feedback to the robotic arm.

[0045] The robot body 10 also includes a visual servo alignment unit. This unit is used to activate a visual guidance unit to acquire real-time images of the target fire-fighting facility after the robot body 10 reaches its vicinity, and to identify the detection point and installation location using a target detection algorithm. Based on visual feedback, the visual servo alignment unit drives a multi-degree-of-freedom manipulator to align the force-controlled actuator or end effector with the detection point or installation location through inverse kinematics calculations. If visual recognition fails, the visual servo alignment unit adjusts the robot body 10's pose or gimbal angle.

[0046] In this embodiment, the multimodal perception module 20 is integrated into the robot body 10 and electrically connected to the large-scale intelligent processing module 30. It is used to collect multimodal data after the fire-fighting facilities have moved and to transmit the collected response status data to the large-scale intelligent processing module 30. The multimodal perception module 20 is also electrically connected to the autonomous navigation and control module 50 and the large-scale intelligent processing module 30, respectively, to receive positioning information to synchronously collect data and transmit the collected data to the large-scale intelligent processing module 30.

[0047] The multimodal sensing module 20 includes a non-contact detection subunit, a contact operation subunit, an environmental parameter acquisition subunit, and a dynamic sensing node subunit. The non-contact detection subunit uses a pan-tilt camera to collect data on the facility's appearance, pressure gauge readings, and indicator light status, automatically identifying these using optical character recognition (OCR) and image classification algorithms; it also uses an infrared thermal imager to detect abnormal surface temperatures and a laser rangefinder to detect the distance between sprinkler heads and obstacles. The contact operation subunit uses a robotic arm gripper to hold a valve handle, employing force control to rotate it with a set torque, monitoring the rotation angle and force feedback curve to determine valve flexibility and whether it is stuck; it automatically resets after operation; it also attempts to remove the safety pin from a fire extinguisher using the robotic arm gripper, determining that the safety pin is not removed or is corroded if the force feedback exceeds a threshold; it returns to its original position after testing. The environmental parameter acquisition subunit uses onboard environmental sensors to collect temperature, humidity, and smoke concentration data near the detection point.

[0048] In this embodiment, the dynamic sensing node is configured to be temporarily mounted on the fire protection facility to be inspected or on an adjacent structure by a multi-degree-of-freedom manipulator. The dynamic sensing node includes one or more of the following: pressure sensor, flow sensor, image acquisition unit, temperature and humidity sensor, smoke concentration sensor, and electrical parameter sensor. The dynamic sensing node incorporates a microcontroller, a wireless communication unit, and a power supply unit. The power supply unit provides an independent power source and can use a high-capacity disposable battery or a rechargeable battery pack. The dynamic sensing node is located at a preset position in the tool bay and is collaboratively managed by the robot body and the backend large model module to assist in completing remote inspection tasks locally on the current robot body.

[0049] The main robot is responsible for node selection, deployment, data aggregation, and retrieval. The large-scale intelligent processing module 30, acting as the intelligent decision-making unit, is responsible for determining the node configuration parameters, acquisition strategies, and data analysis logic based on the requirements of the detection task. Upon receiving a fire protection facility detection task, the large-scale intelligent processing module 30, combining the fire protection facility detection standard library and the system linkage logic library, analyzes and decomposes the detection task, identifying sub-tasks that require simultaneous data acquisition from multiple points, long-term continuous monitoring, or parallel data acquisition while the main robot is performing other tasks. These sub-tasks are marked as remote tasks suitable for dynamic sensing node-assisted execution. Based on the characteristics of the sub-tasks, the large-scale intelligent processing module 30 selects suitable sensor combinations from the node resource pool to form dynamic sensing nodes, generates a node configuration parameter file, and sends it to the main robot via a wireless communication network. The configuration parameter file includes node identifiers, sensor type combinations, sampling frequency, reporting strategies, wake-up conditions, and task execution cycles.

[0050] After receiving the node configuration parameters from the large-scale intelligent processing module 30, the robot body retrieves the selected dynamic sensing nodes from its tool library and installs them at designated locations based on the structural characteristics of the facility to be inspected and the requirements of the inspection items. The robot body performs the installation work through a multi-degree-of-freedom manipulator. During the installation process, the robot body uses its onboard vision sensors to collect images of the installation points in real time and uploads these images to the large-scale intelligent processing module 30 for installation position verification. The large-scale intelligent processing module 30 uses a target detection algorithm to identify the installation reference features on the fire protection facility body, compares the deviation between the node installation position and the preset installation pose, and issues adjustment commands to the robot body when the deviation exceeds the allowable range.

[0051] After being activated, the dynamic sensing node collects data according to the sampling frequency and acquisition strategy set by the large-scale model intelligent processing module 30. The collected pressure values, flow rates, temperature and humidity values, smoke concentration values, electrical parameter values, and image frame data are transmitted back to the robot body in real time via a wireless communication link. The robot body then aggregates data from other multimodal sensing modules, performs integrity verification and timestamp alignment on the data packets, and transmits them to the large-scale model intelligent processing module 30. Upon receiving the detection dataset uploaded by the robot body, the large-scale model intelligent processing module 30 performs fusion analysis on the multimodal data.

[0052] Once all the detection tasks are completed, the robot body, following the retrieval instructions from the large model module in the background, will disassemble the dynamic sensing nodes from their installation locations and move them to the designated storage location in the tool library.

[0053] In this embodiment, the large model intelligent processing module 30 is integrated on the robot body 10 to process and analyze the received data in order to generate a fire protection facility inspection report.

[0054] The large model intelligent processing module 30 includes a data association subunit, a hierarchical detection and judgment subunit, and a report generation subunit.

[0055] The data association subunit is used to automatically associate each piece of detection data with facility identifier, detection timestamp, robot body 10 identifier, robot body 10 pose and sensor calibration status, forming an immutable detection chain through distributed ledger technology.

[0056] The hierarchical testing and judgment subunit executes four levels of judgment in sequence—existence testing, compliance testing, responsiveness testing, and linkage testing—according to the business logic of building fire protection facility testing. Within each level, the test data is re-examined according to a three-level judgment logic. (1) Existence detection is used to verify whether the fire protection subsystem or its components are set up as required by the design. It outputs the result of existence or absence by visual recognition and spatial coordinate matching. (2) Compliance testing is used to determine whether the installation location, specifications, installation method, quantity and configuration of the fire protection subsystem or its components meet the requirements of the fire protection technical standards. The results of the test are output as qualified or unqualified by numerical comparison, location comparison or status comparison. (3) The responsiveness test is used to verify whether the fire-fighting equipment can make the expected response action when it receives the trigger signal. By calculating the time difference between the trigger signal and the feedback signal and comparing it with the allowable response time, the result of the judgment of whether the response is qualified or unqualified is output. (4) Linkage detection is used to verify whether all devices that should be linked are started after the trigger signal is issued. By comparing the actual feedback device set with the linkage device set, the result of the linkage integrity is output as qualified or unqualified. In each of the above levels of detection, the detection data is considered normal when it is within the preset standard threshold; when the detection data exceeds the threshold, a re-inspection is automatically triggered, and the robot body's pose or lighting conditions are adjusted before re-collection; if the re-inspection is still abnormal, it is marked as abnormal, and multi-angle images and force feedback curves are retained as evidence; if the detection point cannot be identified due to environmental interference, it is marked as identification failure and included in the list of items to be re-inspected.

[0057] The report generation subunit is used to summarize all test data and hierarchical judgment results to generate a test report. Based on the testing standard system, the report generation subunit determines the report template, categorizes and summarizes the judgment results of existence testing, compliance testing, responsiveness testing, and linkage testing by subsystem, and fills in four sections: basic information, sub-item judgment results, key data records, and comprehensive conclusions and recommendations. The report content includes statistics on the completion of 10 tasks for each robot entity, a facility list and test item judgment results, an anomaly location distribution map, and a historical trend comparison chart. The report supports manual review of anomalies and exports electronic records that comply with fire protection testing specifications. Example 1

[0058] The following describes the typical workflow of the building fire protection facility detection system in this embodiment, based on the modules described above.

[0059] The robot receives detection tasks from the fire protection facility management platform via wireless communication, specifically for wet automatic sprinkler systems. The autonomous navigation and control module generates a globally optimal path based on a built-in fire protection facility pose database and controls the robot to autonomously travel to the preset position of the end-point test device corresponding to the target alarm valve group. During travel, the global navigation unit integrates laser SLAM and wheeled odometers to achieve real-time pose estimation, while the local obstacle avoidance unit uses a dynamic window method for dynamic obstacle avoidance.

[0060] Before performing the coordinated testing, the robot system needs to verify the existence and compliance of system components. The robot body sequentially completes the existence and compliance testing of key components such as fire pumps, elevated fire water tanks, water flow indicators, and signal valves. All test results are uploaded to the large model module to determine if the preconditions are met. The synchronous multimodal perception module and the temporarily deployed dynamic perception nodes work together to collect a series of data: water flow indicators, alarm valve pressure switches, hydraulic alarm bells, elevated fire water tank flow switches, fire pumps, temporary pressure nodes, etc. The robot body returns to the end-point test device. The fire facility triggering module starts to simulate the water discharge process of a sprinkler head, realizing synchronous detection of responsiveness and coordination. The robot body aggregates, verifies, and timestamps all data packets from its various perception units and dynamic perception nodes. The data association subunit of the large model intelligent processing module associates all data with the current testing task, facility identification, and timestamp. The hierarchical detection and judgment subunit initiates the linkage detection logic. The set of linked devices is derived from the fire protection facility detection standard library and the system linkage logic library. The set of linked devices corresponding to alarm valve group W-1 is preset, including water flow indicator, alarm valve pressure switch, hydraulic alarm bell, high-level fire water tank flow switch, and fire pump. After completion, the robot body resets the automatic sprinkler system. The robot body then moves back to the fire pump outlet main pipe, and the robotic arm disassembles the temporarily deployed dynamic sensing nodes, returning them to the designated compartment in the tool library for charging or standby. The report generation subunit summarizes the judgment results, key data, and multi-angle images of all levels in this detection task.

[0061] Therefore, it can be seen that, according to the building fire protection facility inspection system of this embodiment, the robot body utilizes its autonomous navigation, multimodal perception, intelligent decision-making and precise operation capabilities to complete the full-process, automated inspection of a complex fire protection system in accordance with national standards.

[0062] In this embodiment, the building fire protection facility detection system also includes a collaborative domain scheduling and task management layer. The collaborative domain scheduling and task management layer is used to divide the building space into multiple collaborative domains and set up a regional coordinator in each collaborative domain to realize distributed task scheduling and resource management.

[0063] The collaborative domain scheduling and task management layer includes a collaborative domain partitioning module, a regional coordinator, and a cross-domain handover module.

[0064] The collaborative domain division module divides the building into multiple collaborative domains based on the building's fire compartments, floors, and equipment layers' natural boundaries. Each collaborative domain acts as an independent scheduling unit, with tasks within the domain autonomously managed by the regional coordinator and cross-domain tasks coordinated by the global platform. A virtual buffer zone is set at the regional boundary, which is a spatial area ranging from 1.5 m to 3 m on both sides of the regional boundary, used to trigger the cross-domain handover process when the robot body 10 enters the zone.

[0065] A regional coordinator is deployed within each collaborative domain to perform task decomposition, resource reservation, conflict resolution, and state synchronization within the domain. When two or more robot entities 10 simultaneously reserve the same fire protection facility inspection task, the regional coordinator allocates the execution order based on task priority and the battery status of the robot entities 10. The regional coordinator can be either an edge server deployed within the domain or a single robot entity 10 within the domain. The regional coordinator adopts a primary-backup redundancy architecture: the primary coordinator is the edge server, and the backup coordinator is a single robot entity 10 within the domain with a battery level above 80% of its rated capacity and a CPU clock speed of at least 2.5 GHz; the primary and backup coordinators synchronize their status via heartbeats with an interval of no more than 1 second; if the primary coordinator fails, the backup coordinator automatically becomes the primary coordinator within 3 seconds and broadcasts its new address to all robot entities 10 within the domain. For regions without deployed edge servers, the primary coordinator is elected by the robot entities 10 within the domain using a lightweight consensus algorithm, and the backup coordinator is the robot entity 10 with the second-highest number of votes; the election process only requires the participation of three or more robot entities 10 within the domain.

[0066] The cross-domain handover module handles task handover when the robot body 10 crosses the boundary of the collaborative domain. When the robot body 10 enters the virtual buffer, the cross-domain handover process is automatically triggered. The cross-domain handover process includes: the robot body 10 sending a ready-to-handover signal to the source region coordinator; the source region coordinator encapsulating the task context of the robot body 10 into a handover packet, which includes the currently executing detection task identifier, the remaining path node sequence, the list of occupied resources, and the pose data of the robot body 10; the source region coordinator sending a handover request to the target region coordinator; the target region coordinator receiving and verifying the handover packet and updating its local resource occupancy table; the target region coordinator sending a handover confirmation to the robot body 10, and the robot body 10 continuing to execute the detection operation; the robot body 10 does not stop moving during the handover process.

[0067] In this embodiment, the building fire protection facility detection system also includes a multi-agent collaborative control layer. This layer is integrated into each robot body 10 and is used to achieve collaborative scheduling, conflict resolution, and fault tolerance recovery among multiple robot bodies 10 within the same collaborative domain. The multi-agent collaborative control layer includes a distributed region partitioning and role negotiation module, a commitment-negotiation-monitoring protocol module, a two-level negotiation mechanism module, a central-edge hybrid scheduling module, and a distributed collaborative fault tolerance module.

[0068] The distributed region partitioning and role negotiation module adopts a distributed consensus algorithm based on geohash. Each robot entity 10 generates a hash value through Geohash encoding based on its own coordinates. Robot entities 10 with the same hash value prefix are grouped into the same collaborative domain, and a temporary coordinator is elected by comparing the size of the hash values. The temporary coordinator is authorized by the regional coordinator or elected through consensus among the robot entities 10 within the domain. It is responsible for the real-time scheduling of tasks and conflict arbitration within the domain, and its term of office is bound to the task cycle within the domain.

[0069] The commitment-negotiation-monitoring protocol module uses distributed ledger technology to record the soft commitments issued by robot body 10 when accepting a task. These soft commitments include the task completion time, resource usage period, and confidence level. The confidence level is calculated based on the current battery level, health status, and historical fulfillment rate of robot body 10. The temporary coordinator determines resource reservations based on the confidence level, allocating tasks with a confidence level below 70% to primary and backup robot bodies 10 for parallel execution. If a robot body 10 commits to performing a linkage detection task for an alarm valve group, and is unable to continue due to a battery level below 30% during its journey, it sends a cancellation message to the temporary coordinator. The temporary coordinator then initiates a compensation allocation algorithm, transferring the incomplete detection task to another robot body 10 within the same collaborative domain that has a battery level above 60% and is in an idle state.

[0070] The two-level negotiation mechanism module includes a first-level negotiation subunit and a second-level negotiation subunit. The first-level negotiation subunit exchanges poses and driving intentions through point-to-point communication between the robot bodies 10. It employs a distributed coordination algorithm based on potential game theory for local path replanning. Each robot body 10 uses its own path cost function as its utility function, and converges to a Nash equilibrium through iterative updates. When two robot bodies 10 are moving towards each other in a 1.8 m wide corridor, they exchange their poses, speeds, and remaining task time through point-to-point communication. The robot with the earlier system timestamp has priority, while the other avoids it. In the second-level negotiation subunit, if the first-level negotiation fails or involves critical resources, a temporary coordinator arbitrates based on distributed consensus. These critical resources include the end-of-line test device, the operating position of the alarm valve group, and the operating area in front of the fire pump control cabinet. The temporary coordinator maintains a local resource reservation table and prioritizes critical resources, with the fire pump control cabinet testing task having a higher priority than the end-of-line test device testing task.

[0071] When the central-edge hybrid scheduling module is operating normally, the large-scale intelligent processing module 30 is responsible for cross-domain task allocation and long-term plan optimization, while the temporary coordinator is responsible for intra-domain task scheduling. When communication with the large-scale intelligent processing module 30 is interrupted, the temporary coordinator automatically switches to offline autonomous mode, continues to process intra-domain tasks through distributed consensus, and synchronizes data with the large-scale intelligent processing module 30 after communication is restored. For example, when communication between the global scheduling platform of a ten-story building and the regional coordinator of the collaborative domain from the first basement level to the third floor is interrupted, the robot body 10 in that collaborative domain continues to perform pump room inspection and alarm valve group detection tasks under the scheduling of the temporary coordinator. After communication is restored, it automatically uploads the detection data and execution logs of the completed tasks.

[0072] When the temporary coordinator fails, the distributed collaborative fault-tolerant module enables the robot entities 10 within the domain to elect a new temporary coordinator through a lightweight consensus algorithm. This lightweight consensus algorithm uses a variant of the Raft consensus algorithm, and master node election is achieved through election timeout and heartbeat mechanisms. When the temporary coordinator of a collaborative domain goes offline due to hardware failure, the remaining six robot entities 10 within the domain complete the election of a new coordinator within 2 seconds. The newly elected coordinator takes over the scheduling of unfinished tasks within the domain. When communication with the large model intelligent processing module 30 is interrupted, all robot entities 10 automatically switch to distributed collaborative mode, exchanging poses and task progress through point-to-point communication. A practical Byzantine fault-tolerant algorithm is used to maintain collaborative operation; PBFT is used when the number of nodes does not exceed 7, and a simplified version of the SBFT algorithm is used when the number of nodes exceeds 7. Data is synchronized with the large model intelligent processing module 30 after communication is restored.

[0073] In this embodiment, the building fire protection facility detection system also includes a fire protection facility pose database, which provides facility spatial information to the autonomous navigation and control module 50. The fire protection facility pose database pre-stores the type, three-dimensional spatial pose, standard values ​​of detection items, and pose offset of each fire protection facility within the building. The three-dimensional spatial pose includes the X, Y, and Z coordinates of the facility's center point and the rotation angles Roll, Pitch, and Yaw around the three coordinate axes, in meters (m) and degrees (deg). The pose offset of the detection point includes the offset of the operating point coordinates relative to the facility body and the approach direction vector of the end effector. For example, for a wet alarm valve assembly, the operating point coordinate offset is set to 0.2 m along the positive X-axis and 0.5 m along the positive Z-axis relative to the valve assembly's center point, and the approach direction vector is a horizontal approach along the positive X-axis.

[0074] The fire protection facility pose database is established in the following way: When the first robot body 10 is deployed, a high-precision 3D point cloud map is generated by fusing LiDAR and depth cameras. The robot body 10 is then manually taught to sequentially reach the locations of each fire protection facility, record the facility pose, and enter it into the database. Alternatively, a visual marker-assisted method is used, where the pose is automatically calculated and entered by recognizing QR codes or AprilTags pre-attached to the facility body. Subsequent robot bodies 10 can directly load the map through map sharing without repeated calibration. During movement, the pose is corrected by recognizing the same visual markers or natural features. Example 2

[0075] The following describes the typical workflow of the building fire protection facility detection system in this embodiment, based on the modules described above.

[0076] At the start of the detection task, the large model task planning engine in the large model intelligent processing module 30 receives natural language detection instructions, automatically breaks them down into facility-level task packages by combining the building information model and the fire protection facility pose database, and optimizes the task granularity based on historical data. The collaborative domain division module divides the building into multiple collaborative domains according to the building's fire protection zones, and the task packages are assigned to the corresponding area coordinators according to the domain.

[0077] During the task allocation phase within the domain, each robot entity 10 issues a soft commitment to the regional coordinator through the commitment-negotiation-monitoring protocol module in the multi-agent collaborative control layer. The regional coordinator calculates the confidence level based on the current battery level, operating status, and historical fulfillment rate of the robot entity 10, and assigns tasks with confidence levels below a preset threshold to primary and backup robot entities 10 for parallel execution. After accepting the task, the robot entity 10 moves to the target detection point according to the path planned by the autonomous navigation and control module 50.

[0078] For facilities requiring continuous monitoring, after arriving at the detection point, the robot body 10 uses the multi-degree-of-freedom manipulator in the fire protection facility trigger module 40 to grab a dynamic sensing node from the tool library. The visual servo alignment unit guides the robotic arm to temporarily install the sensing node on the facility to be monitored or a nearby structure. Once deployed, the dynamic sensing node begins continuous monitoring, using an intermittent working mode to collect parameters such as pressure, flow rate, images, temperature, humidity, and smoke. When the monitored parameters continuously drift beyond a preset trend threshold, the robot automatically generates a trend verification task and sends it to the regional coordinator.

[0079] During navigation and operation, when multiple robot bodies 10 have path conflicts or need to compete for key resources such as elevators and fire doors, coordination is carried out through a two-level negotiation mechanism module: the robot bodies 10 first exchange poses and driving intentions through point-to-point communication, and use the dynamic window method to replan local paths; if the first-level negotiation fails, the regional coordinator will arbitrate based on distributed consensus and allocate resources according to the principle of priority for urgent tasks and low-battery robot bodies 10.

[0080] For inspection items requiring contact operation, the robot body 10 uses the multi-degree-of-freedom manipulator in the fire protection facility trigger module 40. The visual servo alignment unit guides the end effector to align with the valve handle or safety pin, setting the torque to perform rotation or removal operations, and monitoring torque and displacement feedback. It automatically resets after the operation is complete. For locations with deployed dynamic sensing nodes, the robot body 10 sends a node retrieval request to the area coordinator. The area coordinator confirms the subsequent task schedule for that location and replies with a retrieval permission command. The robot body 10 then uses the multi-degree-of-freedom manipulator to retrieve the dynamic sensing node and store it in the tool library.

[0081] During the detection process, the multimodal sensing module 20 collects image data such as the appearance of the facility, pressure gauge readings, and indicator light status through the non-contact detection subunit; it detects the surface temperature of the equipment using an infrared thermal imager; and it measures spatial dimensions such as nozzle spacing using a laser rangefinder. Simultaneously, it collects temperature, humidity, and smoke concentration data through the environmental parameter acquisition subunit. The monitoring data continuously transmitted back by the dynamic sensing nodes, together with the data collected by the robot body 10, constitute a multimodal data stream. All multimodal data, after being associated with the facility identifier, timestamp, and robot body 10 identifier by the data association subunit, forms an immutable detection chain.

[0082] During the data judgment phase, the hierarchical detection and judgment subunit executes a three-level judgment logic: when the detection data is within the preset standard threshold, it is judged as normal; when the data exceeds the threshold, a re-inspection is automatically triggered, and the robot body's pose or lighting conditions are adjusted before re-collecting data; if the re-inspection is still abnormal, it is marked as abnormal, and multi-angle images and force feedback curves are retained as evidence; if the detection point cannot be identified due to environmental interference, it is marked as identification failure and included in the list of items to be re-inspected.

[0083] When multiple robot bodies 10 are deployed to work collaboratively, the detection results of each robot body 10 are aggregated to the regional coordinator through a central-edge hybrid scheduling module. The regional coordinator uses distributed ledger technology to deduplicate, align, and merge the detection results from each platform, forming a unified structured data field across platforms. If communication with the central platform is interrupted, the regional coordinator automatically switches to offline autonomous mode, continues to process tasks within the domain through distributed consensus, and synchronizes data with the global platform after communication is restored. If the temporary coordinator fails, the robot bodies 10 within the domain elect a new temporary coordinator within 3 seconds using a lightweight consensus algorithm. During the election, the robot bodies 10 suspend accepting new tasks, and completed tasks are mutually confirmed through point-to-point communication.

[0084] Finally, the report generation subunit determines the report template based on the testing standard system. It then fills in the summarized results of existence testing, compliance testing, responsiveness testing, and linkage integrity testing into four sections of the report template: basic information, sub-item judgment results, key data records, and comprehensive conclusions and recommendations, automatically generating the testing report. The report generation and output module 70 formats the testing report output by the report generation subunit and supports exporting electronic records that conform to fire protection testing specifications.

[0085] Therefore, this embodiment integrates the multimodal perception module 20, the large-scale intelligent processing module 30, the fire protection facility triggering module 40, the autonomous navigation and control module 50, and the power supply module 60 onto the robot body 10, constructing a complete automated closed loop from autonomous navigation, precise positioning, contact operation, multimodal data acquisition to intelligent analysis and report generation. The autonomous navigation and control module 50 drives the robot body 10 to move precisely to the preset detection point, the fire protection facility triggering module 40 automatically triggers facility actions based on the positioning information, the multimodal perception module 20 synchronously collects multidimensional response data, and the large-scale intelligent processing module 30 correlates, judges, and generates reports on the data, realizing fully unmanned detection of building fire protection facilities and significantly improving detection efficiency and data reliability.

[0086] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0087] The above embodiments are merely illustrative of several implementation methods of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A building fire protection facility detection system based on intelligent robot autonomous navigation and location matching, characterized in that, include: The robot body, and the multimodal perception module, large model intelligent processing module, fire-fighting facility triggering module, autonomous navigation and control module and power supply module, all integrated on the robot body; The power supply module is used to provide operating power; The autonomous navigation and control module is used to generate path planning information for the preset detection area and matching facility location information, and to control the robot body to move according to the path planning information and to control the robot body to reach the position indicated by the facility location information. The multimodal sensing module is electrically connected to the large model intelligent processing module. It is used to collect multimodal data after the fire-fighting facilities are activated and transmit the multimodal data to the large model intelligent processing module. The large-scale intelligent processing module is used to process and analyze multimodal data to generate fire protection facility inspection reports.

2. The building fire protection facility detection system according to claim 1, characterized in that, It also includes a multi-agent collaborative control layer; The multi-agent collaborative control layer is integrated into each robot body, including a distributed region division and role negotiation module; The distributed region division and role negotiation module adopts a distributed consensus algorithm based on geohash. Each robot entity generates a hash value through Geohash encoding based on its own coordinates. Robot entities with the same hash value prefix are grouped into the same collaborative domain. A temporary coordinator is elected by comparing the size of the hash values. The temporary coordinator is responsible for task scheduling and conflict arbitration within the domain, and their term of office is tied to the task cycle within the domain.

3. The building fire protection facility detection system according to claim 2, characterized in that, The multi-agent collaborative control layer also includes a commitment-negotiation-monitoring protocol module; The commitment-negotiation-monitoring protocol module uses distributed ledger technology to record the soft commitments issued by the robot when it accepts a task. The soft commitments include the task completion time, resource usage period, and confidence level. The confidence level is calculated based on the robot's current battery level, health status, and historical fulfillment rate. The temporary coordinator determines resource reservations based on confidence level, and assigns main and backup robots to execute tasks with a confidence level below 70% in parallel. When the robot body is unable to fulfill its commitment, it sends a cancellation message, and the temporary coordinator initiates a compensation allocation algorithm to transfer the unfinished task to a nearby idle robot body.

4. The building fire protection facility detection system according to claim 2, characterized in that, The multi-agent collaborative control layer also includes a two-level negotiation mechanism module; The two-level negotiation mechanism module includes a first-level negotiation subunit and a second-level negotiation subunit; The first-level negotiation subunit exchanges pose and driving intention through point-to-point communication between robot subjects, and uses a distributed coordination algorithm based on potential game to perform local path replanning. Each robot subject uses its own path cost function as its utility function, and converges to Nash equilibrium through iterative updates. When the first-level negotiation fails or critical resources are involved, the second-level negotiation subunit shall arbitrate based on distributed consensus by the temporary coordinator. The temporary coordinator maintains a local resource reservation table and prioritizes critical resources.

5. The building fire protection facility detection system according to claim 2, characterized in that, The multi-agent collaborative control layer also includes a central-edge hybrid scheduling module; When the large model intelligent processing module is working normally, the central-edge hybrid scheduling module is responsible for cross-domain task allocation and long-term plan optimization, while the temporary coordinator is responsible for intra-domain task scheduling. When the communication of the large model intelligent processing module is interrupted, the temporary coordinator automatically switches to offline autonomous mode, continues to process tasks within the domain through distributed consensus, and synchronizes data with the large model intelligent processing module after communication is restored.

6. The building fire protection facility detection system according to claim 2, characterized in that, The multi-agent collaborative control layer also includes a distributed collaborative fault-tolerant module; When the temporary coordinator fails, the distributed collaborative fault-tolerant module enables the robot subjects within the domain to elect a new temporary coordinator through a lightweight consensus algorithm. The lightweight consensus algorithm adopts a variant of the Raft consensus algorithm and achieves master node election through election timeout and heartbeat mechanisms. When the communication of the large model intelligent processing module is interrupted, all robot bodies automatically switch to distributed collaborative mode, exchange poses and task progress through point-to-point communication, and maintain collaborative operation using a practical Byzantine fault-tolerant algorithm. When the number of nodes does not exceed 7, PBFT is used, and when the number of nodes exceeds 7, a simplified version of SBFT algorithm is used. After communication is restored, the data is synchronized with the large model intelligent processing module.

7. The building fire protection facility detection system according to claim 1, characterized in that, The main body of the robot includes a multi-degree-of-freedom manipulator, a force-controlled actuator, a tool library, and a vision guidance unit; The tool library includes a variety of end-effector tools, which are selected from at least one of valve operating tools, safety pin removal tools, detection sensors, and image acquisition devices. The multi-degree-of-freedom manipulator automatically selects and replaces the corresponding end effector based on the detection task. Force-controlled actuators are used to operate valve handles with a set torque and monitor torque and displacement feedback, or to pull out safety pins and monitor pull-out force feedback, and automatically reset after operation.

8. The building fire protection facility detection system according to claim 7, characterized in that, The multimodal sensing module 20 includes a dynamic sensing node; the dynamic sensing node is configured to be temporarily installed on the fire protection facility to be detected or on an adjacent structure by a multi-degree-of-freedom manipulator; the dynamic sensing node includes one or more of the following: pressure sensor, flow sensor, image acquisition unit, temperature and humidity sensor, smoke concentration sensor and electrical parameter sensor; the dynamic sensing node is equipped with a microcontroller, a wireless communication unit and a power supply unit.

9. The building fire protection facility detection system according to claim 1, characterized in that, The autonomous navigation and control module includes a mobile chassis, an environmental perception unit, and a positioning unit; The mobile chassis adopts an omnidirectional moving structure; Both the environmental sensing unit and the positioning unit are fixedly connected to the mobile chassis; The environmental perception unit is used to collect environmental data around the robot body, and the localization unit is used to collect the pose data of the robot body. The environmental perception unit and the positioning unit are electrically connected to the autonomous navigation and control module, and transmit the collected environmental data and pose data to the autonomous navigation and control module.

10. The building fire protection facility detection system according to claim 1, characterized in that, The large model intelligent processing module also includes a data association subunit, a hierarchical detection and judgment subunit, and a report generation subunit; The data association subunit is used to associate detection data with facility identifiers, timestamps, and robot identifiers, and to form an immutable detection chain through distributed ledger technology; The hierarchical detection and judgment subunit is used to sequentially perform existence detection, compliance detection, responsiveness detection and linkage detection, and to re-examine the detection data according to the three-level judgment logic; The report generation subunit is used to summarize the test data and generate a test report.

Images