A data transmission and storage method for an autonomous detection process in a closed space

Abstract

The application provides a data transmission and storage method for autonomous detection in a closed space, and relates to the technical field of emergency rescue, which comprises the following steps: moving and collecting multi-source perception data in the closed space by using an autonomous detection device, and generating space point cloud data by using a simultaneous localization and mapping (SLAM) module carried by the autonomous detection device; performing real-time analysis on the multi-source perception data by using a front-end computing unit of the autonomous detection device to obtain an analysis result, and dynamically generating a priority label representing rescue efficiency of the data according to the analysis result, so as to encapsulate the original data into a data packet of the priority label. The application realizes intelligent perception, elastic transmission and integrated fusion storage of data in an extreme environment.

Description

Technical Field

[0001] This invention relates to the field of emergency rescue technology, and in particular to a data transmission and storage method for autonomous detection in confined spaces. Background Technology

[0002] A metal mine 300 meters underground collapsed suddenly due to a geological disaster, forming multiple blocked sections and irregular tunnels. The rescue team adopted a multi-hop transmission scheme based on radio relay to carry out the search and rescue: three repeaters were set up at the entrance of the mine and outside the collapsed section to establish a communication link, and an autonomous detection robot equipped with a thermal imaging camera and a toxic gas sensor was sent to explore deep into the mine.

[0003] However, due to the complex structure of the tunnel and the obstruction of dust, the wireless signal attenuated drastically, and the link frequently experienced packet loss and interruption. Furthermore, the robot did not intelligently filter the collected data, and transmitted a large number of videos of tunnel walls without signs of life, normal concentration air data, critical data on excessive levels of toxic gas (carbon monoxide), and blurry thermal images of suspected trapped personnel indiscriminately. Valuable narrowband bandwidth was occupied by invalid data, resulting in a delay in transmitting critical information about excessive levels of toxic gas, which is related to the safety of rescuers, back to the command center. The location data of suspected trapped personnel was also lost multiple times due to the unstable link, which restricted the timeliness and accuracy of rescue decisions and could not meet the needs of rapid location and precise rescue after an accident in a confined space. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a data transmission and storage method for autonomous detection in confined spaces, so as to realize intelligent sensing, flexible transmission and integrated storage of data in extreme environments.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0006] Firstly, a data transmission and storage method for autonomous exploration in confined spaces, the method comprising:

[0007] Step 1: The autonomous detection device moves within a confined space and collects multi-source sensing data. It then uses the real-time positioning and mapping SLAM module on the autonomous detection device to generate spatial point cloud data. The front-end computing unit of the autonomous detection device performs real-time analysis on the multi-source sensing data to obtain analysis results. Based on the analysis results, it dynamically generates priority tags that characterize the data rescue effectiveness, and encapsulates the raw data into data packets with priority tags.

[0008] Step 2: The autonomous detection device assesses the communication link status with the backend network, obtains the assessment results, and performs adaptive hierarchical transmission based on the assessment results and data packets with priority tags: When the link is connected, data packets are scheduled for back transmission according to the order of priority tags, with high-priority data packets enjoying a higher transmission guarantee level; when the link is interrupted, the autonomous detection device caches data packets in local storage units according to priority order and continues to perform the detection task. After the communication link is restored, the high-priority data packets in the cache are retransmitted first.

[0009] Step 3: The edge command unit receives high-priority data packets and spatial point cloud data retransmitted by the autonomous detection equipment. The edge command unit uses the spatial point cloud data as a unified spatial reference, assigns spatial coordinates to each data packet, and associates and fuses different types and time-related sensing data with the same or adjacent spatial coordinates to construct a digital map based on a unified spatiotemporal index.

[0010] Furthermore, the autonomous detection device moves within a confined space and collects multi-source sensing data. It then utilizes the real-time localization and mapping (SLAM) module onboard the device to generate spatial point cloud data, including:

[0011] Step 1.1: The front-end computing unit processes multi-source perception data in parallel, including at least thermal imaging video stream, visible light video stream and environmental audio stream, and performs human target detection, posture recognition and abnormal sound recognition respectively to obtain recognition results;

[0012] Step 1.2: Based on the identification results, a comprehensive judgment is made. When a confirmed vital sign is identified, a first priority label is generated; when a suspicious vital sign is identified, a second priority label is generated; when environmental hazard data is identified, a third priority label is generated; and when a scene without abnormalities is identified, a fourth priority label is generated.

[0013] Step 1.3: Differentiate the encapsulation of the raw sensing data according to the priority labels: encapsulate the data carrying the first priority label into a data packet containing high-fidelity compressed sensing data; encapsulate the data carrying the second and third priority labels into a compressed data packet containing only key feature information; and encapsulate the data carrying the fourth priority label into a data packet containing only status reports.

[0014] Furthermore, the front-end computing unit of the autonomous detection equipment performs real-time analysis on multi-source sensing data to obtain analysis results, and dynamically generates priority tags characterizing data rescue effectiveness based on the analysis results, so as to encapsulate the raw data into data packets with priority tags, including:

[0015] Step 1.4: Receive the recognition results, including human target detection boxes and pose classification results;

[0016] Step 1.5: Confidence assessment and spatiotemporal alignment of each recognition result: Match and fuse detection boxes with spatial location correlation within the same time window with the recognition results to obtain comprehensive judgment data;

[0017] Step 1.6: Based on the comprehensive analysis data and according to the predefined analysis rules, perform logical association judgment: if the stable human target in the thermal imaging video stream and the active human posture in the visible light video stream are successfully matched in time and space, then the first priority label is obtained; if only the stable human target in the thermal imaging video stream exists, then the second priority label is generated.

[0018] Step 1.7: Based on the judgment conclusion, obtain the final priority label.

[0019] Furthermore, the autonomous probing device assesses the communication link status with the backend network, obtains the assessment results, and performs adaptive hierarchical transmission based on the assessment results and data packets with priority tags: When the link is connected, data packets are scheduled for backhaul according to the order of priority tags, with higher priority data packets enjoying a higher transmission guarantee level; when the link is interrupted, the autonomous probing device caches data packets in local storage units according to priority order and continues to perform the probing task. After the communication link is restored, the high-priority data packets in the cache are retransmitted first, including:

[0020] Step 2.1: The communication module of the autonomous detection device periodically detects and evaluates the quality of the communication link with the backend network to obtain the link status evaluation result;

[0021] Step 2.2: Execute a branch decision based on the link status assessment result. If the link status assessment result is connected and stable, proceed to step 2.3; if the link status assessment result is interrupted, proceed to step 2.5.

[0022] Step 2.3: Establish a real-time transmission queue and insert the data packets with priority tags into the real-time transmission queue in descending order of priority tags. Among them, data packets with the first priority tag are assigned the highest transmission priority and are scheduled for transmission with the highest transmission guarantee level, including dedicated channel, enhanced forward error correction and shortened retransmission interval.

[0023] Step 2.4: During network transmission, monitor the load status of the real-time transmission queue in real time. If the load is lower than the preset threshold, schedule and send data packets with second, third and fourth priority tags in sequence.

[0024] Step 2.5: When the link is interrupted, the autonomous detection device switches to offline working mode and starts the data caching mechanism: the data packets to be transmitted are stored in the local shockproof and waterproof storage unit in the order of their priority tags, and the detection task continues to be executed.

[0025] Step 2.6: In offline working mode, the communication module continuously attempts to restore the communication link; when the link is detected to be restored, it prioritizes extracting cached data packets with the first priority tag from the shockproof and waterproof storage unit for retransmission. After the high priority data packets are retransmitted or the network is idle, it retransmits the cached data of other priorities in sequence.

[0026] Furthermore, the edge command unit receives high-priority data packets and spatial point cloud data retransmitted by autonomous detection devices. Using the spatial point cloud data as a unified spatial reference, the edge command unit assigns spatial coordinates to each data packet and correlates and fuses sensing data of different types and times with the same or adjacent spatial coordinates to construct a digital map based on a unified spatiotemporal index, including:

[0027] Step 3.1: The edge command unit receives and parses the data transmitted back from the autonomous detection device, and obtains the perception data, corresponding priority tags, and data acquisition timestamps encapsulated in the data packet; at the same time, it receives and processes the spatial point cloud data stream generated by the SLAM module.

[0028] Step 3.2, assign spatiotemporal coordinates to each received data packet: based on the data packet's acquisition timestamp, match the corresponding acquisition time and device pose in the spatial point cloud data stream, and calculate the three-dimensional spatial coordinates of the data packet's perceived content within the enclosed space;

[0029] Step 3.3: Using the three-dimensional spatial coordinates and the acquisition timestamp as the joint index key, store the data packet with the completed coordinate assignment into the spatiotemporal fusion database; the spatiotemporal fusion database automatically associates sensing data with adjacent three-dimensional spatial coordinates, different acquisition timestamps, different types and different priorities to the same data record;

[0030] Step 3.4: Based on the spatiotemporal fusion database and spatial point cloud data, construct a digital map of the enclosed space, and dynamically overlay the perceived data as an interactive layer onto the corresponding coordinate position of the digital map to obtain a digital map based on a unified spatiotemporal index.

[0031] Furthermore, spatiotemporal coordinates are assigned to each received data packet: based on the data packet's acquisition timestamp, the corresponding acquisition time and device pose are matched in the spatial point cloud data stream to calculate the three-dimensional spatial coordinates of the data packet's perceived content within the enclosed space, including:

[0032] Step 3.21: Extract the data collection timestamp carried in the data packet;

[0033] Step 3.22: Based on the data acquisition timestamp, search for the device pose data synchronized with the timestamp in the spatial point cloud data stream. The device pose data includes the three-dimensional position and attitude of the autonomous detection device in the global coordinate system.

[0034] Step 3.23: Obtain the fixed installation parameters of the sensing sensor corresponding to the data packet on the autonomous detection device. The installation parameters define the transformation relationship between the sensor coordinate system and the device body coordinate system.

[0035] Step 3.24: Combining the device pose data and installation parameters, calculate the three-dimensional spatial coordinates of the data packet sensing content in the global coordinate system through coordinate transformation.

[0036] Furthermore, based on the spatiotemporal fusion database and spatial point cloud data, a digital map of a confined space is constructed, and the perceived data is dynamically overlaid as an interactive layer onto the corresponding coordinate positions of the digital map, resulting in a digital map based on a unified spatiotemporal index, including:

[0037] Step 3.41: Process the spatial point cloud data, including noise reduction, registration and reconstruction, to obtain a three-dimensional mesh model of the closed spatial structure, which serves as the static base of the digital map;

[0038] Step 3.42: On the static base of the digital map, establish a map coordinate system consistent with the global coordinate system;

[0039] Step 3.43: Determine the corresponding geographical location in the map coordinate system based on the three-dimensional spatial coordinates of each data packet;

[0040] Step 3.44: Define different visualization layers for different types of sensing data, and dynamically render the determined geographical locations, along with the associated sensing data content, priority labels, and collection timestamps, as data markers onto the corresponding visualization layers.

[0041] Step 3.45: Integrate the visualization layer and digital map base, and provide a map interaction interface; receive query commands for specific locations on the map through the map interaction interface, aggregate and return all multi-source historical sensing data related to the location from the spatiotemporal fusion database, and realize query and aggregate display by map location.

[0042] Secondly, a data transmission and storage system for autonomous exploration in confined spaces includes:

[0043] The acquisition module is used for the autonomous detection device to move in a confined space and collect multi-source sensing data, and to generate spatial point cloud data using the real-time positioning and mapping SLAM module on the autonomous detection device; the front-end computing unit of the autonomous detection device performs real-time analysis on the multi-source sensing data to obtain analysis results, and dynamically generates priority tags that characterize the data rescue effectiveness based on the analysis results, so as to encapsulate the raw data into data packets with priority tags.

[0044] The execution module is used to autonomously probe the status of the communication link with the backend network, obtain the assessment results, and perform adaptive hierarchical transmission based on the assessment results and data packets with priority tags: when the link is connected, data packets are scheduled for back transmission according to the order of priority tags, with high-priority data packets enjoying a higher transmission guarantee level; when the link is interrupted, the autonomous probe caches data packets in the local storage unit according to priority order and continues to execute the probe task. After the communication link is restored, the high-priority data packets in the cache are retransmitted first.

[0045] The processing module is used by the edge command unit to receive high-priority data packets and spatial point cloud data retransmitted by autonomous detection devices. The edge command unit uses the spatial point cloud data as a unified spatial reference, assigns spatial coordinates to each data packet, and associates and fuses different types and time-related sensing data with the same or adjacent spatial coordinates to construct a digital map based on a unified spatiotemporal index.

[0046] Thirdly, a computing device, comprising:

[0047] One or more processors;

[0048] A storage device for storing one or more programs that, when executed by one or more processors, cause the one or more processors to implement the method.

[0049] Fourthly, a computer-readable storage medium storing a program that, when executed by a processor, implements the method.

[0050] The above-described solution of the present invention has at least the following beneficial effects:

[0051] This invention overcomes the technical problems of existing technologies in autonomous detection in confined spaces, such as the transmission delay of critical life rescue information under extremely unstable network conditions, the contradiction between limited communication resources and massive data, link interruption leading to detection work interruption or data loss, and low efficiency of multi-source heterogeneous rescue data fusion. It employs a front-end computing unit to perform real-time analysis of multi-source sensing data and dynamically generate priority labels representing rescue effectiveness, and encapsulates data in a differentiated manner according to priority. This is achieved by combining the real-time evaluation and adaptive hierarchical transmission mechanism of communication link status by autonomous detection equipment, and the edge command unit's use of SLAM point cloud data as a unified spatial reference to assign spatiotemporal coordinates to data packets, as well as the association and fusion of sensing data of different types and times to construct a digital map based on a unified spatiotemporal index. This results in ensuring reliable, prioritized, and low-latency return of high-priority life rescue information, improving network bandwidth utilization efficiency, ensuring the continuity of the search and rescue process, and providing commanders with a unified and aggregated situational awareness view to improve decision-making efficiency. Attached Figure Description

[0052] Figure 1 This is a flowchart illustrating a data transmission and storage method for autonomous exploration in confined spaces, provided by an embodiment of the present invention.

[0053] Figure 2 This is a schematic diagram of a data transmission and storage system for autonomous exploration in a confined space, provided by an embodiment of the present invention. Detailed Implementation

[0054] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0055] like Figure 1 As shown, an embodiment of the present invention proposes a data transmission and storage method for autonomous exploration in confined spaces, the method comprising the following steps:

[0056] Step 1: The autonomous detection device moves within a confined space and collects multi-source sensing data. It then uses the real-time positioning and mapping SLAM module on the autonomous detection device to generate spatial point cloud data. The front-end computing unit of the autonomous detection device performs real-time analysis on the multi-source sensing data to obtain analysis results. Based on the analysis results, it dynamically generates priority tags that characterize the data rescue effectiveness, and encapsulates the raw data into data packets with priority tags.

[0057] Step 2: The autonomous detection device assesses the communication link status with the backend network, obtains the assessment results, and performs adaptive hierarchical transmission based on the assessment results and data packets with priority tags: When the link is connected, data packets are scheduled for back transmission according to the order of priority tags, with high-priority data packets enjoying a higher transmission guarantee level; when the link is interrupted, the autonomous detection device caches data packets in local storage units according to priority order and continues to perform the detection task. After the communication link is restored, the high-priority data packets in the cache are retransmitted first.

[0058] Step 3: The edge command unit receives high-priority data packets and spatial point cloud data retransmitted by the autonomous detection equipment. The edge command unit uses the spatial point cloud data as a unified spatial reference, assigns spatial coordinates to each data packet, and associates and fuses different types and time-related sensing data with the same or adjacent spatial coordinates to construct a digital map based on a unified spatiotemporal index.

[0059] In this embodiment of the invention, the invention overcomes the technical problems of existing technologies in autonomous detection in confined spaces, such as the transmission delay of critical life rescue information in extremely unstable network environments, the contradiction between limited communication resources and massive data, the interruption of detection work or data loss due to link interruption, and the low efficiency of multi-source heterogeneous data fusion. This is achieved by using an autonomous detection device to collect multi-source sensing data and generate spatial point cloud data through a SLAM module while collecting multi-source sensing data. The front-end computing unit analyzes the multi-source sensing data in real time and dynamically generates priority labels representing rescue effectiveness to encapsulate data packets. Combined with the link status assessment results, adaptive hierarchical transmission of data packets is performed. Furthermore, the edge command unit assigns spatial coordinates to each data packet using SLAM point cloud data as a unified spatial reference. Different types and time sensing data under the same or adjacent spatial coordinates are associated and fused to construct a digital map based on a unified spatiotemporal index.

[0060] In a preferred embodiment of the present invention, step 1 above may include:

[0061] Step 1.1: The front-end computing unit processes multi-source sensing data in parallel, including at least thermal imaging video stream, visible light video stream, and ambient audio stream. It performs human target detection, pose recognition, and abnormal sound recognition to obtain recognition results. Specifically, the front-end computing unit simultaneously initiates processing of the thermal imaging video stream, visible light video stream, and ambient audio stream. The processing of these three data streams is carried out in parallel without interference. For the thermal imaging video stream, the front-end computing unit performs frame-by-frame continuous analysis, focusing on capturing areas in the image that conform to human contour features and human temperature range. By continuously tracking the stability and morphological integrity of these areas, it confirms whether a human target exists. The target detection operation focuses on various targets appearing in the visible light video stream. The front-end computing unit analyzes the target's limb movements, continuity, and posture characteristics to determine whether there are active movements such as waving or turning the head, thus completing the posture recognition operation. The front-end computing unit continuously receives and analyzes the audio signals of the environmental audio stream, compares them with the background sound features of a normal enclosed space environment, and identifies abnormal sound signals such as cries for help, knocking sounds, or groans that are different from the normal environmental sounds, thus completing the abnormal sound recognition operation. The three data streams perform corresponding recognition operations respectively. The front-end computing unit collects and organizes the human target detection results, posture recognition results, and abnormal sound recognition results.

[0062] Step 1.2: Based on the identification results, a comprehensive judgment is made. When a definite life sign is identified, a first priority label is generated; when a suspicious life sign is identified, a second priority label is generated; when environmental hazard data is identified, a third priority label is generated; and when no abnormal scene is identified, a fourth priority label is generated. Specifically, this includes: the front-end computing unit summarizes all identification results, performs comprehensive integration, correlation, and in-depth analysis on the results, and checks whether stable human target detection results and active human posture recognition results are obtained simultaneously. If both results are satisfied and are synchronous in time and corresponding in space, it is determined that a definite life sign has been identified, i.e., a first priority label is generated. If only a stable human target is detected but no active posture is identified, or only a clear abnormal sound is identified but no human target is detected, it is determined that a suspicious life sign has been identified, i.e., a second priority label is generated. If the identification results include gas concentration exceeding the standard, abnormal temperature, structural stability risks, etc., which may threaten the life safety of trapped personnel, a third priority label is generated. If the human target detection results show no human presence, the posture recognition results show no active posture, the abnormal sound recognition results show no abnormal sound, and all environmental data are within a safe and normal range, it is determined that no abnormal scene has been identified, i.e., a fourth priority label is generated.

[0063] Step 1.3: Differentiated encapsulation of raw sensing data based on priority tags: Data carrying the first priority tag is encapsulated into a data packet containing high-fidelity compressed sensing data; data carrying the second and third priority tags is encapsulated into a compressed data packet containing only key feature information; data carrying the fourth priority tag is encapsulated into a data packet containing only status reports. Specifically, the front-end computing unit performs targeted encapsulation processing on the raw sensing data according to different priority tags. For the raw sensing data carrying the first priority tag, in order to retain as much detailed information related to determining vital signs as possible while taking into account transmission efficiency, a high-fidelity compression algorithm is used to process the data, reducing the data size while ensuring that the core features and details of the data are not lost. For the original sensing data carrying the second and third priority labels, key feature information that reflects the core situation is extracted from the data. If the key feature is the human outline corresponding to the suspicious life signs, the specific parameters exceeding the standard and the numerical range corresponding to the environmental hazard data, the extracted key feature information is compressed to remove redundant data and packaged into a compressed data package containing only key feature information. For the original sensing data carrying the fourth priority label, since there is no abnormal information and key rescue-related content, only the core status information such as the equipment working status, data acquisition time, and approximate detection location in the data is extracted and organized into a concise and clear status report, which is then packaged into a data package containing only the status report.

[0064] In a preferred embodiment of the present invention, the front-end computing unit processes multi-source sensing data such as thermal imaging video stream, visible light video stream, and environmental audio stream in parallel, performs human target detection, posture recognition, and abnormal sound recognition respectively, and generates first to fourth level priority labels based on the recognition results. The original sensing data is then differentiated and encapsulated according to these priority labels. This overcomes the technical problems in the prior art, such as the indiscriminate transmission of multi-source sensing data leading to the occupation of valuable narrowband bandwidth by invalid data, the transmission delay of critical life information and environmental hazards, and the contradiction between massive data and limited communication resources. Thus, it achieves the technical effect of improving information density from the source, filtering invalid information, optimizing bandwidth utilization, and providing precise priority criteria for adaptive hierarchical transmission, ensuring priority transmission of critical data.

[0065] In a preferred embodiment of the present invention, step 1 above may include:

[0066] Step 1.4 involves receiving recognition results, including human target detection boxes and pose classification results. Specifically, the front-end computing unit initiates a dedicated data receiving process to synchronously receive various recognition results generated by parallel processing. These recognition results specifically include human target detection boxes from the thermal imaging video stream human target detection operation and pose classification results from the visible light video stream pose recognition operation. During the receiving process, each type of recognition result is verified one by one to ensure the integrity of the original data acquisition time without any missing data, thus preparing for the spatiotemporal alignment operation. After the reception is completed, all recognition results are classified and organized, and a temporary data index is established according to data type and acquisition time order for easy and quick retrieval.

[0067] Step 1.5: Confidence assessment and spatiotemporal alignment of each recognition result: Matching and fusing detection boxes and recognition results with spatial location correlation within the same time window to obtain comprehensive judgment data. Specifically, this includes: a front-end computing unit conducting a confidence assessment on each received and processed recognition result. During the assessment, the target clarity, feature matching degree, and signal stability of the recognition result are referenced. Multiple indicators are used to quantify the reliability of each recognition result. Invalid recognition results with confidence levels below the set standard are eliminated, while valid recognition results with high confidence levels are retained. Subsequently, spatiotemporal alignment is performed. A fixed-length time window is set to ensure that all recognition results participating in the alignment are within the same time interval. Then, the spatial location correlation between different recognition results is determined. By comparing the coordinate range of the human target detection box with the target position corresponding to the pose classification result, it is confirmed whether there is a positional overlap or proximity relationship between the two. For detection boxes and recognition results that meet the same time window requirements and are spatially correlated, matching and fusing are performed. The effective information is integrated, and duplicate or contradictory content is eliminated. Finally, comprehensive judgment data that can fully reflect the actual situation of the detection scene is obtained.

[0068] Step 1.6: Based on comprehensive analysis data and according to predefined analysis rules, perform logical association judgment: If the stable human target in the thermal imaging video stream and the active human pose in the visible light video stream are successfully matched in time and space, a first priority label is obtained; if only the stable human target in the thermal imaging video stream exists, a second priority label is generated. Specifically, the front-end computing unit retrieves the predefined analysis rules, which clarify the judgment conditions corresponding to different priority labels, extracts human target information related to the thermal imaging video stream from the comprehensive analysis data, and judges whether the human target is a stable human target. The judgment criterion is that the human target can be detected in multiple consecutive time windows and the target outline remains intact. Within a reasonable range, extract posture information related to the visible light video stream to determine whether there is active human posture. The judgment criteria are that the target limb has obvious voluntary movements, such as raising a hand, turning the head, bending over, etc. Then, perform spatiotemporal correlation verification to check whether the stable human target and the active human posture are synchronized in time and whether they appear in the same or adjacent locations in space. If both meet the spatiotemporal matching conditions, it is determined that a definite life sign has been identified and a first priority label is obtained. If, in the comprehensive analysis of the data, only a stable human target of thermal imaging video stream can be extracted, and no active human posture of visible light video stream is found, and there is no other effective information to support the determination of life signs, it is determined that a suspicious life sign has been identified and a second priority label is generated.

[0069] Step 1.7: Based on the judgment conclusion, obtain the final priority label. Specifically, this includes: performing a final verification of the logical association judgment conclusion in Step 1.6. The judgment process strictly follows the predefined judgment rules, with no logical loopholes or judgment errors. If the judgment conclusion is that the stable human target in the thermal imaging video stream and the active human posture in the visible light video stream are successfully matched in time and space, then the priority label corresponding to the recognition result is finally determined as the first priority label. If the judgment conclusion is that only the stable human target in the thermal imaging video stream exists, then the priority label corresponding to the recognition result is finally determined as the second priority label. After the verification is completed, the final determined priority label is stored and synchronously transmitted to the differentiated encapsulation. Based on the priority, the original perception data is processed in a targeted manner to provide clear and accurate basis.

[0070] In this embodiment of the invention, by first receiving recognition results such as human target detection boxes and posture classification results, and then performing confidence assessment and spatiotemporal alignment on each recognition result, and matching and fusing detection boxes and recognition results with spatial location correlation within the same time window to obtain comprehensive judgment data, and performing logical association judgment based on the comprehensive judgment data and predefined judgment rules, a first priority label is obtained if the stable human target in the thermal imaging video stream and the active human posture in the visible light video stream are successfully matched spatiotemporally; if only the stable human target in the thermal imaging video stream exists, a second priority label is generated, and the final priority label is obtained based on the judgment conclusion. Therefore, this technical means overcomes the technical problems in the prior art, such as the lack of effective integration and intelligent association analysis of multi-source perception recognition results, which leads to inaccurate priority label generation, easy misjudgment or omission of key life information, and invalid data occupying bandwidth and delaying the transmission of key information due to the lack of precise priority guidance. Thus, it achieves the technical effect of making the generation of priority labels more accurate and reliable, providing accurate basis for differentiated data encapsulation and adaptive hierarchical transmission, optimizing bandwidth utilization efficiency, and ensuring priority and low-latency transmission of key life information.

[0071] In a preferred embodiment of the present invention, step 2 above may include:

[0072] Step 2.1: The communication module of the autonomous detection device periodically detects and evaluates the quality of the communication link with the back-end network to obtain the link status evaluation results. Specifically, the communication module of the autonomous detection robot pre-sets a reasonable fixed time interval based on the depth of the 300-meter underground metal mine, the multiple blocked sections, and the characteristics of the irregular tunnel environment. The fixed time interval promptly captures rapid changes in the link status caused by dust flow and tunnel structure obstruction, ensuring that no critical link quality fluctuations are missed. The communication module continuously sends detection signals to the back-end network built by three repeaters according to the set time interval, and simultaneously receives feedback signals from the back-end network in real time. During signal reception, the communication module focuses on monitoring four core indicators: signal strength, to reflect the attenuation of the wireless signal in the mine dust obstruction and complex tunnel structure, and to determine whether the signal can meet basic transmission requirements; signal stability, to observe whether the feedback signal fluctuates frequently or is sometimes strong and sometimes weak, and to evaluate whether the link is in a stable state; data packet loss rate, to count the proportion of no corresponding feedback received after each detection signal is sent, and to quantify the reliability of link data transmission; and transmission delay, to record the time difference between the detection signal being sent and the feedback being received, and to determine the real-time performance of data transmission. The communication module collects, records, and analyzes these four indicators one by one. When the signal strength remains within the preset effective range, the signal does not fluctuate frequently, the data packet loss rate is lower than the preset threshold, and the transmission delay is controlled within a reasonable range, the link status is determined to be connected and stable. When no feedback signal is received from the back network and there is no response after multiple probes, the link status is determined to be interrupted. Finally, a clear link status assessment result is obtained, providing an accurate basis for branch decision-making.

[0073] Step 2.2: Execute branch decisions based on the link status assessment results. If the link status assessment result is connected and stable, proceed to step 2.3; if the link status assessment result is interrupted, proceed to step 2.5. Specifically, this includes: After receiving the link status assessment results, the control unit of the autonomous exploration robot immediately initiates the branch decision process. The decision process focuses on the reliability requirements of data transmission in the mine rescue scenario and establishes clear judgment criteria: First, check whether the signal strength in the assessment results remains within a preset effective range. This range is pre-calibrated based on the signal attenuation law at a depth of 300 meters in the mine to ensure effective data transmission. Next, judge the signal stability index to confirm that the feedback signal does not fluctuate frequently or recover briefly after a sudden interruption. Check whether the data packet loss rate is lower than a preset threshold. This threshold is set in conjunction with the characteristics of the radio relay multi-hop transmission scheme to avoid misjudging the link status due to normal minor packet loss. Finally, confirm whether the transmission delay is controlled within a reasonable range required for rescue to ensure the real-time performance of data transmission. If all four conditions above are met, it indicates that the link can stably support data transmission. The control unit determines that the link status is connected and stable, and proceeds to step 2.3 to execute the real-time transmission process. If the evaluation results show that the signal has completely disappeared, and no feedback is received from the back network after multiple consecutive transmissions of detection signals, and no data can be transmitted, it indicates that the link is completely interrupted due to structural obstruction caused by mine collapse or repeater failure. The control unit determines that the link status is interrupted, and proceeds to step 2.5 to execute the relevant operations of the offline working mode. The entire decision-making process responds quickly to ensure that the data processing strategy can be adjusted in a timely manner after the link status changes, and to avoid the loss of key data due to decision-making delays.

[0074] Step 2.3: Establish a real-time transmission queue and insert the data packets with priority tags into the real-time transmission queue in descending order of priority tags. Among them, the data packets with the first priority tag are assigned the highest transmission priority and are scheduled for transmission with the highest transmission guarantee level, including dedicated channels, enhanced forward error correction and shortened retransmission intervals. Specifically, when the real-time transmission process is initiated, the control unit of the autonomous exploration robot immediately creates a dedicated real-time transmission queue. This queue is used to orderly schedule all data packets with priority tags to be transmitted, avoiding chaotic transmission of data with different priorities. The control unit identifies each data packet to be transmitted, extracts the priority tag corresponding to each data packet, and inserts the data packets into the real-time transmission queue in descending order of priority tags. The first priority tag corresponds to data related to trapped personnel with confirmed vital signs in the mine, the second priority tag corresponds to data on suspected signs of life, the third priority tag corresponds to environmental hazard data such as excessive levels of toxic gases, and the fourth priority tag corresponds to status report data in normal scenarios. Data packets with higher priority are ranked higher in the queue to ensure priority access to transmission resources. Data packets with the first priority tag in the queue are assigned the highest transmission priority because they are directly related to the life safety of the trapped personnel and are the core basis for rescue decisions. To address link instability caused by the complex structure of the mine and dust obstruction, the communication module employs the highest transmission guarantee level: a dedicated channel is allocated to first-priority data packets, independent of the transmission channels for other priority data packets, preventing low-priority data from preempting critical transmission resources and ensuring that channel bandwidth is dedicated to the transmission of vital signs data; enhanced forward error correction technology is adopted, adding redundant check information to the data, so that when data is partially lost due to signal attenuation or interference, the redundant information can be used to recover the complete data, reducing the impact of dust obstruction on data transmission; at the same time, the retransmission interval is shortened, and the communication module monitors the transmission status of first-priority data packets in real time. Once a data transmission failure or packet loss is detected, the retransmission mechanism is immediately initiated without waiting for the regular retransmission cycle, reducing the risk of data loss. After completing the above configuration, the communication module prioritizes and sends data packets with first-priority tags according to the queue order, ensuring that vital signs data is transmitted back to the command center at the fastest speed and with the highest reliability.

[0075] Step 2.4: During network transmission, the load status of the real-time transmission queue is monitored in real time. If the load is below a preset threshold, data packets with second, third, and fourth priority tags are scheduled and sent in sequence. Specifically, during the continuous transmission of data packets with the first priority tag, the communication module of the autonomous detection robot continuously monitors the load status of the real-time transmission queue in real time to avoid affecting the transmission efficiency of high-priority data due to excessive queue load. The load status is comprehensively measured by two core indicators: first, the cumulative number of data packets to be transmitted in the queue, reflecting the degree of queue congestion; second, the proportion of bandwidth currently occupied by transmission, reflecting the utilization of network resources. The preset threshold is set according to the narrowband bandwidth characteristics of the mine radio relay multi-hop transmission scheme to ensure that a load status below the threshold will neither cause queue congestion nor occupy too much bandwidth to affect the transmission stability of the first priority data packets. The communication module continuously compares the real-time collected load indicators with the preset threshold. If the load status is detected to be below the preset threshold, it indicates that there are remaining network bandwidth resources and that it will not interfere with the transmission of the first priority data packets. At this point, the control unit sequentially retrieves the corresponding data packets from the real-time transmission queue according to the order of second priority tags, third priority tags, and fourth priority tags, and schedules their transmission. Specifically, the suspected life signs data (second priority tag) can provide the rescue team with potential search directions; the environmental hazard data (third priority tag) such as excessive toxic gases directly relates to the operational safety of rescue personnel; and the anomaly-free scenario status report (fourth priority tag) allows the command center to monitor the detection progress and avoid repeated searches in ineffective areas. This on-demand scheduling method fully utilizes remaining bandwidth resources while ensuring that low-priority data does not preempt the transmission channels of high-priority data, solving the problem of indiscriminate transmission in existing technologies that leads to narrowband bandwidth being occupied by invalid data.

[0076] Step 2.5: When the link is interrupted, the autonomous detection device switches to offline working mode and activates the data caching mechanism. It stores the data packets to be transmitted in the local shockproof and waterproof storage unit according to their priority tags, and continues to execute the detection task. Specifically, when the control unit determines that the link is interrupted, the autonomous detection robot immediately and automatically switches to offline working mode to ensure that the detection task is not halted due to the link interruption. Simultaneously, the control unit quickly activates the data caching mechanism, first sorting all the data packets to be transmitted one by one, extracting the priority tag of each data packet, and sorting them according to the first priority, second priority, third priority, and fourth priority order to ensure that critical data can be retrieved first from the cache. After sorting, the control unit sequentially stores these data packets into the robot's built-in shockproof and waterproof storage unit. This specially designed storage unit can withstand the harsh environment after a mine collapse, effectively resisting the effects of collisions, vibrations, and moisture, preventing data corruption or loss due to environmental conditions. After data caching, the autonomous detection robot does not stop its current detection task and continues to move within the mine according to the preset search and rescue route. It continuously captures human targets in the tunnels using its onboard thermal imaging camera and monitors changes in the concentration of gases such as carbon monoxide in real time using a toxic gas sensor, continuously collecting multi-source sensing data. For newly collected raw data, the robot's front-end computing unit still performs real-time analysis, priority tag generation, and differentiated encapsulation according to the established process. The newly generated data packets with priority tags are also stored in the shockproof and waterproof storage unit in priority order, ensuring the continuity of the entire search and rescue process and preventing detection stagnation and data loss due to link interruption. This solves the problem of severe information lag after link interruption in existing technologies.

[0077] Step 2.6: In offline working mode, the communication module continuously attempts to restore the communication link. When link restoration is detected, it prioritizes retransmitting cached data packets with the first priority tag from the shockproof and waterproof storage unit. After the high-priority data packets are retransmitted or the network is idle, it retransmits cached data of other priorities in sequence. Specifically, in offline working mode, the communication module of the autonomous exploration robot does not stop working but continuously sends link restoration detection signals to the backend network at fixed time intervals to attempt to re-establish the communication connection. After each detection signal is sent, the communication module continuously listens to the feedback information from the backend network and verifies the validity of the feedback signals one by one, including whether the signal strength meets the transmission requirements, whether the signal is stable, and whether a complete communication link can be formed. If the communication module receives a valid response from the backend network and confirms through multiple consecutive probes that the link can stably transmit data without frequent packet loss or interruption, the link is deemed to have been restored. The communication module immediately initiates the cached data retransmission process, prioritizing the extraction of cached data packets with the first priority tag from the shockproof and waterproof storage unit. These data packets correspond to confirmed vital signs data, which are the most critical basis for rescue decisions. During retransmission, the highest transmission guarantee level in step 2.3 is still used, including dedicated channels, enhanced forward error correction, and shortened retransmission intervals, to ensure that core data can be quickly and reliably transmitted back to the command center. Once all data packets with the first priority tag have been retransmitted, or the communication module detects that the network is idle, i.e., the load of the real-time transmission queue is below the preset threshold and there is no longer any high-priority data to be transmitted, other priority cached data are extracted from the storage unit in the order of second priority tag, third priority tag, and fourth priority tag for retransmission. By prioritizing the transmission of critical data, this method ensures that core data such as information on trapped personnel and excessive levels of toxic gases are not delayed for too long due to link interruption. It effectively solves the problem of delay and loss of critical information in existing technologies, providing timely and complete data support for rescue decisions.

[0078] In this embodiment of the invention, because the communication module of the autonomous detection device periodically detects and evaluates the communication link quality with the backend network, and performs branch decisions based on the evaluation results, a real-time transmission queue is established in the real-time transmission process, and data packets with priority tags to be transmitted are inserted into the queue in descending order of priority tags. Data packets with the first priority tag are assigned the highest transmission priority and the highest transmission guarantee level, including dedicated channels, enhanced forward error correction, and shortened retransmission intervals, is enabled for scheduling and transmission. At the same time, the load status of the real-time transmission queue is monitored in real time during network transmission. When the load is lower than a preset threshold, data packets with the second, third, and fourth priority tags are scheduled and transmitted in sequence. In offline working mode, the autonomous detection device switches to this mode and starts a data caching mechanism, storing the data packets to be transmitted in the order of their priority tags into a local shockproof and waterproof storage unit and continuing to perform detection tasks. The communication module continues to operate. The technology attempts to restore the communication link. When the link is restored, it prioritizes retransmitting cached data packets with the highest priority tag from the shockproof and waterproof storage unit. Other priority cached data packets are retransmitted in sequence after the high-priority data packets are retransmitted or the network is idle. This overcomes the technical problems in existing technologies, such as link instability in confined spaces leading to delays in critical data transmission, limited communication bandwidth being occupied by low-priority data, detection work stagnation or data loss after link interruption, and the inability to obtain dedicated transmission guarantees for critical data. As a result, it achieves the following technical effects: ensuring priority, low latency, and reliable transmission of high-priority core rescue data; optimizing the utilization efficiency of limited bandwidth resources; preventing low-priority data from occupying critical transmission resources; ensuring the continuity of detection work and preventing data loss when the link is interrupted; and quickly retransmitting critical data after the link is restored. This improves the elasticity and stability of data transmission and provides timely core support for rescue decision-making.

[0079] In a preferred embodiment of the present invention, step 3 above may include:

[0080] Step 3.1: The edge command unit receives and parses the data transmitted back from the autonomous detection device, obtaining the perception data, corresponding priority tags, and data acquisition timestamps encapsulated in the data packets. Simultaneously, it receives and processes the spatial point cloud data stream generated by the SLAM module. Specifically, the edge command unit continuously monitors the rear network built by three repeaters, synchronously receiving real-time data packets transmitted back from the autonomous detection robot 300 meters underground in a metal mine, as well as cached data packets retransmitted after the link is restored. For each received data packet, the edge command unit initiates a parsing process, extracting various types of perception data encapsulated within the data packet, including images of suspected trapped personnel captured by thermal imaging cameras, carbon monoxide concentration data collected by toxic gas sensors, and tunnel environment video data. It also extracts the priority tags corresponding to each data packet to determine the importance of the data for rescue, and accurately obtains the data acquisition timestamp to record the specific time the data was generated. While receiving and parsing the data packets, the edge command unit simultaneously receives the spatial point cloud data stream generated by the SLAM module on the autonomous detection robot. This data stream contains spatial structure information of multiple blocked sections and irregular tunnels inside the mine. The edge command unit performs preliminary processing on the spatial point cloud data stream, removing noise points caused by dust obscuring the mine shaft, and correcting point cloud deviations caused by robot movement and shaking. This ensures that the point cloud data can accurately reflect the three-dimensional spatial structure of the mine shaft, laying the foundation for spatiotemporal coordinate assignment and digital map construction.

[0081] Step 3.2, assign spatiotemporal coordinates to each received data packet: Based on the data packet's acquisition timestamp, match the corresponding acquisition time and device pose in the spatial point cloud data stream, and calculate the three-dimensional spatial coordinates of the data packet's perceived content in the enclosed space. Specifically, the edge command unit initiates the spatiotemporal coordinate assignment process for each completed parsing data packet. First, the acquisition timestamp of the data packet is extracted to determine the precise time of data generation. Then, in the pre-processed spatial point cloud data stream, the device pose data that is synchronized with the timestamp or has the smallest time difference is retrieved in chronological order. This data records in detail the three-dimensional position of the autonomous exploration robot in the global coordinate system of the mine at that moment, as well as key information such as the robot's orientation and posture, ensuring the time accuracy of coordinate matching. The edge command machine retrieves the fixed installation parameters of the sensing sensors corresponding to the data packet on the autonomous exploration robot. These parameters clarify the transformation relationship between the coordinate systems of different sensors such as thermal imaging cameras and toxic gas sensors and the robot's body coordinate system, avoiding coordinate calculation deviations caused by differences in sensor installation positions. The edge command machine combines the retrieved device pose data and sensor installation parameters, and through coordinate transformation calculations, transforms the sensing content corresponding to the data packet from the sensor's local coordinate system to the mine's global coordinate system, accurately calculating the three-dimensional spatial coordinates of the sensing content in the confined space of a 300-meter underground metal mine.

[0082] Step 3.3: Using the 3D spatial coordinates and acquisition timestamp as a joint index key, the data packets with completed coordinate assignments are stored in the spatiotemporal fusion database. The spatiotemporal fusion database automatically associates sensing data with adjacent 3D spatial coordinates, different acquisition timestamps, different types, and different priorities to the same data record. Specifically, the edge command unit uses the calculated 3D spatial coordinates and the acquisition timestamp of the data packet as a joint index key to sequentially store each data packet with completed spatiotemporal coordinate assignments into the spatiotemporal fusion database. This database, designed for the complex environment of a 300-meter underground metal mine, presets a reasonable spatial association threshold. When the 3D spatial coordinates of different data packets fall within this threshold range, they are considered adjacent 3D spatial coordinates. The database automatically associates sensing data with these adjacent 3D spatial coordinates and different acquisition timestamps. Simultaneously, it integrates sensing data of different types and priority labels, such as thermal imaging video data, toxic gas concentration data, and environmental audio data, into the same data record. For example, it associates and stores carbon monoxide concentration data, corresponding thermal imaging data, and environmental sound data collected at different times at a certain tunnel location, forming a complete data archive for that location. This associative storage method effectively solves the data dispersion problem caused by traditional independent storage. Even if the cached data is retransmitted by the robot, it can be quickly integrated into the corresponding data record through the composite index key. This ensures that key information such as the location data of suspected trapped personnel and the data on excessive toxic gas will not be separated from the associated context due to the interruption of the link during retransmission, providing complete data support for subsequent rapid querying and decision-making.

[0083] Step 3.4: Based on the spatiotemporal fusion database and spatial point cloud data, a digital map of the confined space is constructed. The perceived data is dynamically overlaid as an interactive layer onto the corresponding coordinates of the digital map, resulting in a digital map based on a unified spatiotemporal index. Specifically, the edge command unit initiates the construction process of the confined space digital map based on the associated data in the spatiotemporal fusion database and the deeply processed spatial point cloud data. First, the spatial point cloud data undergoes further noise reduction, registration, and reconstruction processing to remove redundant points and deviations caused by mine dust and shadows from sealed sections. This generates a 3D mesh model that accurately reflects the internal structure of the mine, the location of sealed sections, and the orientation of irregular tunnels. If a 2D view is required for the rescue scenario, a 2D planar projection map can be generated simultaneously as the static base of the digital map. Next, a map coordinate system completely consistent with the global coordinate system of the mine is established on this static base to ensure accurate correspondence between data coordinates and map locations. Subsequently, based on the three-dimensional spatial coordinates of each data packet, its corresponding geographical location is accurately located in the map coordinate system. Simultaneously, independent visualization layers are defined for different types of sensing data, such as thermal imaging video data, toxic gas concentration data, and environmental audio data. The geographical location, along with its associated sensing data content, priority labels, and collection timestamps, is dynamically rendered onto the corresponding visualization layer as data markers. For example, data indicating excessive carbon monoxide is marked as a red warning point, and the location of suspected trapped personnel is marked as a yellow priority attention marker. All visualization layers are integrated with the static digital map base to build an intuitive map interaction interface. Rescue commanders can click on any location on the mine's digital map through this interface. The system immediately aggregates and returns all multi-source historical sensing data for that location and adjacent locations from the spatiotemporal fusion database, including gas concentration changes at different times, corresponding thermal imaging images, and environmental sounds. This eliminates the need for manual comparison of scattered data, allowing for a rapid understanding of the complete situation in the area. This effectively solves the problem of delayed decision-making caused by critical information delays and lost location data, providing strong technical support for rapid location and precise rescue after the collapse of a 300-meter-deep underground metal mine.

[0084] In a preferred embodiment of the present invention, steps 3.1 to 3.4 involve an edge command unit receiving and parsing data packets transmitted or supplemented by the autonomous detection device to obtain sensing data, priority tags, and data acquisition timestamps. Simultaneously, it receives and processes the spatial point cloud data stream generated by the SLAM module. Then, based on the acquisition timestamp of each data packet, it matches the corresponding acquisition time and device pose in the spatial point cloud data stream to calculate the three-dimensional spatial coordinates of the sensed content within the enclosed space. Subsequently, using the three-dimensional spatial coordinates and acquisition timestamps as a joint index key, the data packets are stored in a spatiotemporal fusion database. This allows the database to automatically associate sensing data with different acquisition timestamps, types, and priorities under adjacent three-dimensional spatial coordinates to the same data record. Finally, a digital map of the enclosed space is constructed based on the spatiotemporal fusion database and spatial point cloud data. The perceived data is dynamically overlaid as an interactive layer onto the corresponding coordinate position of the digital map. This overcomes the technical problems of existing technologies, such as the spatiotemporal asynchrony and independent storage of multi-source heterogeneous rescue data leading to low fusion efficiency, the need for commanders to manually compare multi-source information such as video, gas, and location, which delays decision-making, and the inability to obtain a unified aggregated situational awareness view. This achieves spatiotemporal synchronization and integrated fusion of multi-source rescue data, providing commanders with an intuitive, unified, and interactive digital map situational view. It also enables the rapid acquisition of full-dimensional historical perception data of a specific location without cumbersome manual comparison, thereby improving the efficiency and accuracy of command and decision-making.

[0085] In a preferred embodiment of the present invention, step 3 above may include:

[0086] Step 3.21 involves extracting the data acquisition timestamps carried in the data packets. Specifically, after parsing the data packets transmitted or retransmitted by the autonomous exploration robot, the edge command unit initiates the data acquisition timestamp extraction process for each data packet. These data packets include thermal imaging data and toxic gas concentration data transmitted by the robot during real-time exploration in a 300-meter-deep underground metal mine, as well as cached data retransmitted after the link is restored. The edge command unit uses a dedicated timestamp extraction module to accurately read the timestamp information carried in the header of the data packets. This timestamp is the precise moment recorded synchronously by the built-in clock module when the robot collects data at a specific location in the mine, reflecting the true time of data generation. During the extraction process, the edge command unit verifies the format and integrity of the timestamps, eliminating timestamp anomalies caused by link interruptions or data transmission delays, ensuring the accuracy of the acquisition timestamp for each data packet. For retransmitted data on the suspected location of trapped personnel or excessive toxic gas levels, emphasis is placed on accurate timestamp extraction to provide a reliable foundation for the time synchronization and matching of spatial point cloud data, avoiding subsequent coordinate calculation errors due to time information deviations.

[0087] Step 3.22: Based on the data acquisition timestamp, search for device pose data synchronized with the timestamp in the spatial point cloud data stream. This device pose data includes the three-dimensional position and attitude of the autonomous exploration device in the global coordinate system. Specifically, after the edge command unit extracts the acquisition timestamp of the data packet, it initiates a retrieval and matching process in the pre-processed spatial point cloud data stream. This spatial point cloud data stream is generated by the SLAM module mounted on the autonomous exploration robot and contains spatial structure information of multiple blocked sections and irregular tunnels inside the mine, as well as device pose data at different times during the robot's exploration. The edge command unit performs a frame-by-frame retrieval of the spatial point cloud data stream in chronological order, searching for device pose data that is completely synchronized with the current data packet acquisition timestamp or has the smallest time difference. The device pose data records in detail the three-dimensional position information of the autonomous exploration robot in the mine's global coordinate system at that precise moment, including the robot's specific coordinates in the mine's depth, lateral, and longitudinal dimensions, as well as key attitude information such as the robot's orientation, pitch angle, and roll posture. During the retrieval process, the edge command unit takes into account the frequent packet loss and interruption of the mine link, allowing for a very small time difference threshold. This ensures that even data packets with slightly different timestamps due to link delays can be matched with the closest device pose data. At the same time, the matching results are verified a second time to eliminate pose data anomalies caused by robot movement jitter or dust obstruction, ensuring the accurate correspondence between device pose data and data packet acquisition time.

[0088] Step 3.23: Obtain the fixed installation parameters of the sensing sensor corresponding to the data packet on the autonomous detection device. These installation parameters define the transformation relationship between the sensor coordinate system and the device's body coordinate system. Specifically, the edge controller identifies the type of sensing sensor corresponding to the current data packet. This data packet may originate from a thermal imaging camera mounted on the autonomous detection robot, or it may come from different devices such as a toxic gas sensor. Subsequently, the edge controller retrieves the fixed installation parameters of this type of sensing sensor on the autonomous detection robot from a pre-set sensor parameter database. These installation parameters are pre-calibrated and stored before the robot leaves the factory or before the detection mission begins. They define in detail the translation, rotation, and other transformation relationships between the sensor coordinate system and the robot's body coordinate system, including key information such as the sensor's installation position coordinates on the robot, installation angle, and offset from the robot's central axis. For example, parameters such as the installation height and horizontal orientation angle of the thermal imaging camera, and the distance between the probe installation position of the toxic gas sensor and the center of the robot body, will be accurately retrieved. The edge command machine will check the completeness and accuracy of the retrieved installation parameters to ensure that the parameters are not deviated due to collisions or vibrations of the robot in the mine, and to avoid the accuracy of subsequent coordinate transformations being affected by incorrect sensor installation parameters. This will provide an accurate basis for the conversion of perception data from the sensor's local coordinate system to the mine's global coordinate system.

[0089] Step 3.24: Combining equipment pose data and installation parameters, the three-dimensional spatial coordinates of the data packet's perceived content in the global coordinate system are calculated through coordinate transformation. Specifically, this includes: the edge controller integrating the matched equipment pose data with the retrieved sensor installation parameters, initiating the coordinate transformation process, using the sensor coordinate system as a reference, and transforming the coordinates of the perceived content corresponding to the data packet from the sensor local coordinate system to the robot's body coordinate system according to the translation and rotation relationships in the sensor installation parameters, eliminating coordinate deviations caused by differences in sensor installation positions. Combining the robot's three-dimensional position and attitude information in the mine's global coordinate system from the equipment pose data, the coordinates of the perceived content in the robot's body coordinate system are further transformed to the mine's global coordinate system. During the process, the edge command system fully considers the global coordinate system calibration rules of the 300-meter underground metal mine, and combines the coordinate information of fixed reference points such as the mine entrance and the blocked section to make real-time corrections to the coordinate parameters during the transformation process. This compensates for errors caused by changes in robot posture and irregularities in the mine's spatial structure. Through this series of coordinate transformation calculations, the system accurately calculates the three-dimensional spatial coordinates of the data packet's perceived content within the confined space of the 300-meter underground metal mine. Whether it is the real-time transmission of data on excessive toxic gases or the supplementary transmission of thermal imaging data of suspected trapped personnel, this process can accurately bind the data to the specific physical location of the mine, achieving a precise correspondence between the data and the actual tunnel location. This provides precise location support for subsequent multi-source data association and fusion and rescue decision-making.

[0090] In this embodiment of the invention, a technique is employed that first, the data acquisition timestamp carried in the data packet is extracted; then, based on the timestamp, the device pose data synchronized with it is searched in the spatial point cloud data stream. This device pose data includes the three-dimensional position and attitude of the autonomous detection device in the global coordinate system. The fixed installation parameters of the sensing sensor corresponding to the data packet on the autonomous detection device are obtained. These installation parameters clarify the transformation relationship between the sensor coordinate system and the device body coordinate system. Finally, by combining the searched device pose data and the obtained installation parameters, coordinate transformation calculations are performed to calculate the three-dimensional spatial coordinates of the data packet's sensing content in the global coordinate system. Therefore, it overcomes the technical problem in existing technologies where the binding between multi-source sensing data and the actual physical location of the confined space is inaccurate and the time and space are not synchronized, resulting in the data being disconnected from the specific tunnel location. This affects the effectiveness of multi-source heterogeneous data fusion and makes it impossible for commanders to accurately match the data with the actual location. Thus, it achieves the precise binding of the sensing content of each data packet with the physical location of the confined space, ensuring the consistency of multi-source data in time and space. This provides a reliable coordinate foundation for subsequent multi-source data association and fusion, unified spatiotemporal index construction, and accurate digital map overlay, enabling commanders to clearly identify the specific location corresponding to key data and improve the accuracy of rescue decisions.

[0091] In a preferred embodiment of the present invention, step 3 above may include:

[0092] Step 3.41 involves processing the spatial point cloud data, including noise reduction, registration, and reconstruction, to obtain a three-dimensional mesh model of the closed spatial structure, serving as the static base for the digital map. Specifically, the edge command unit first receives the spatial point cloud data stream generated by the SLAM module mounted on the autonomous exploration robot. This data stream contains spatial structural information of the irregular tunnels within the sealed section of a 300-meter-deep underground metal mine, along with noise points caused by dust obstruction, deviations caused by sensor jitter during robot movement, and measurement errors from the equipment itself. The edge command unit initiates a three-step processing procedure. The first step is noise reduction. A filtering algorithm analyzes each point in the spatial point cloud data, eliminating dust noise points with signal strength below a preset threshold and isolated abnormal deviation points, retaining only valid point cloud data that accurately reflects the mine structure to ensure the accuracy of the base data for subsequent processing. To ensure accuracy, the second step is registration. Because the autonomous exploration robot moves and probes within the mine, the point cloud data collected at different times exhibits positional and angular deviations. The edge command unit extracts feature markers from each segment of the point cloud data and aligns multiple segments of point cloud data to the same spatial reference system through feature point matching, correcting the misalignment of point cloud stitching caused by changes in the robot's movement posture. The third step is reconstruction. Based on the noise-reduced and registered effective point cloud data, a 3D modeling algorithm is used to construct a 3D mesh model that can completely restore the internal structure of the mine. The model will accurately present the location and shape of the blocked section, the direction and width of irregular tunnels, and the spatial relationships of each area. Simultaneously, according to the actual needs of rescue command, a 2D planar projection map can be generated, clearly marking the mine's planar layout and key area distribution. Finally, the 3D mesh model or 2D planar projection map serves as the static base for the digital map.

[0093] Step 3.42: On the static base of the digital map, establish a map coordinate system consistent with the global coordinate system. Specifically, the edge command machine initiates the map coordinate system establishment process on the static base of the digital map. First, the calibration benchmark of the mine's global coordinate system is clearly defined. Using a fixed reference point at the mine entrance as the origin, the positive X-axis is defined as horizontally forward, the positive Z-axis as vertically upward, and the positive Y-axis as horizontally perpendicular to the X-axis. This constructs a global coordinate system that conforms to the three-dimensional spatial characteristics of the mine. Based on the parameter rules of this global coordinate system, a map coordinate system completely consistent with it is established on a static base. This ensures that the coordinate axis scales of the map coordinate system strictly correspond to the global coordinate system without any deviation. During the establishment process, the edge command unit repeatedly verifies the accuracy of the coordinate system by retrieving point cloud data coordinates of known fixed structures inside the mine, such as the corners of roadways at the edge of sealed sections, and comparing them with the corresponding positions in the map coordinate system. Any minor deviations are corrected to ensure that the map coordinate system accurately maps the actual three-dimensional space of the mine. The map coordinate system will reserve an interface for adaptation with the three-dimensional spatial coordinates of subsequent data packages, ensuring that all perception data with three-dimensional coordinates can accurately match the corresponding positions on the map. This provides a unified coordinate benchmark for determining geographical locations and data marking rendering, avoiding location errors caused by inconsistencies in the coordinate system.

[0094] Step 3.43: Based on the three-dimensional spatial coordinates of each data packet, determine the corresponding geographical location in the map coordinate system. Specifically, this includes: the edge command machine extracting the three-dimensional spatial coordinates from each data packet assigned with the completed spatiotemporal coordinates. These coordinates are the precise location information of the data packet's perceived content in the mine's global coordinate system, including the coordinates of the real-time transmitted toxic gas concentration data, the coordinates of the thermal imaging images of suspected trapped personnel, and the data coordinates of scenes without abnormalities. Subsequently, the edge command system transforms these 3D spatial coordinates according to the parameter rules of the map coordinate system, accurately mapping the 3D coordinates in the global coordinate system to the established map coordinate system. During the transformation process, considering the complex spatial structure inside the mine, the edge command system performs secondary corrections to compensate for minor errors that may exist during the establishment of the map coordinate system and coordinate deviations caused by local spatial deformation of the mine. After the transformation is completed, a unique geographic location point is determined for each data packet in the map coordinate system. The unique geographic location point will accurately correspond to the actual physical location inside the mine, such as the edge of a blockage section in the middle of a tunnel. For the retransmitted suspected trapped personnel location data, the edge command system will focus on verifying the accuracy of its coordinate transformation to ensure that the geographic location point can truly reflect the area where the trapped personnel may be located. This solves the problem of suspected trapped personnel location data being lost or inaccurately located due to unstable links in the background technology, providing rescuers with accurate search directions.

[0095] Step 3.44 defines different visualization layers for different types of sensing data, and dynamically renders the determined geographical locations, along with the associated sensing data content, priority labels, and collection timestamps, as data markers onto the corresponding visualization layers. Specifically, the edge command unit first categorizes and organizes all sensing data, clarifying the different types of sensing data, including thermal imaging video stream data, visible light video stream data, toxic gas concentration data, and environmental audio stream data. An independent visualization layer is defined for each type of sensing data, and each layer is assigned unique display attributes to ensure that different types of data do not interfere with each other when presented on the map, while also facilitating command personnel to... For users who need to view a single type of data individually or multiple types of data overlay, the edge command system integrates the priority tags and collection timestamps of the associated sensor data content for each data packet's corresponding geographic location, generating a unified data marker. The presentation of these data markers is differentiated based on the priority tags: first-priority tags corresponding to confirmed vital signs use a prominent and easily identifiable style; second-priority tags corresponding to suspected signs of life use a moderately prominent style; third-priority tags corresponding to environmental hazards use a style with warning features; and fourth-priority tags corresponding to scenarios without anomalies use a simple, basic style. Finally, the edge command system dynamically renders these data markers onto their respective visualization layers according to their geographic locations, ensuring the accurate positioning of each data marker in the map coordinate system. Simultaneously, the collection timestamps included in the data markers clearly reflect the order in which the data was generated, allowing command personnel to intuitively grasp the status changes of different areas of the mine at different times, eliminating the need for manual comparison of scattered data and improving information acquisition efficiency.

[0096] Step 3.45: Integrate the visualization layers and the digital map base, and provide a map interaction interface; receive query commands for specific locations on the map through the map interaction interface, aggregate and return all multi-source historical sensing data related to the location from the spatiotemporal fusion database, and realize query and aggregate display by map location. Specifically, the edge command unit deeply integrates all generated visualization layers with the established static digital map base to ensure that the positions of each layer and the base are accurately aligned and there will be no layer offset or misalignment, forming a complete and coherent closed-space digital map. A simple and easy-to-use map interaction interface is built. This interface can clearly display the integrated digital map and support basic operations such as zooming and panning for command personnel to easily view the detailed situation of different areas of the mine. When command personnel need to understand the situation of a specific area, they can click on the map location corresponding to that area through the map interaction interface, and the interface will immediately send a query command to the spatiotemporal fusion database. The command contains the map coordinate information of that location. Upon receiving a query command, the spatiotemporal fusion database retrieves all associated multi-source historical sensing data for the location and adjacent areas based on coordinate information. This includes thermal imaging images, changes in toxic gas concentrations, and environmental audio recordings from different acquisition timestamps. Regardless of whether this data is transmitted in real-time or retransmitted after the link is restored, it is completely extracted. The database categorizes and organizes the extracted multi-source data according to the order of acquisition timestamps and data types, forming a comprehensive data report for the area. This report is then fed back to the map interface in real-time for aggregation and display. Command personnel can obtain all relevant information for the area at once through the interface, clearly grasp the status changes of the area from the start of detection to the current moment, quickly determine whether there are trapped personnel and whether the environment is safe, effectively solving the problem of delayed and scattered key information requiring cumbersome manual comparison in the background technology, and providing strong support for rapid positioning and precise rescue.

[0097] In this embodiment of the invention, the spatial point cloud data is first denoised, registered, and reconstructed to obtain a three-dimensional mesh model or two-dimensional planar projection map representing the closed spatial structure as the static base of the digital map. Then, a map coordinate system consistent with the global coordinate system is established on this static base. Subsequently, the corresponding geographical locations are accurately determined in the map coordinate system based on the three-dimensional spatial coordinates of each data packet. Next, dedicated visualization layers are defined for different types of sensing data, and the determined geographical locations, along with the associated sensing data content, priority labels, and collection timestamps, are dynamically rendered to the corresponding layers as data markers. Finally, all visualization layers are integrated with the digital map base, and a map interaction interface is provided to receive queries. This technology, which aggregates and returns all multi-source historical sensing data related to a specific location from a spatiotemporal fusion database, overcomes the technical problems of existing technologies, such as the lack of a unified visualization carrier for closed space rescue data, the scattered presentation of multi-source sensing data, and the ambiguity in the correspondence between data and physical location. These problems lead to commanders having to manually compare various types of data, being unable to quickly and intuitively grasp the full-dimensional information of a specific area, and having low decision-making efficiency. As a result, it achieves the technical effect of constructing an intuitive, accurate, and interactive digital map of closed spaces, realizing the spatiotemporal and hierarchical visualization of multi-source sensing data, and allowing commanders to quickly obtain complete historical data of any location through simple operations, reducing the difficulty of data interpretation and improving the intuitiveness and accuracy of rescue decisions.

[0098] like Figure 2 As shown, embodiments of the present invention also provide a data transmission and storage system for autonomous exploration in confined spaces, comprising:

[0099] The acquisition module is used for the autonomous detection device to move in a confined space and collect multi-source sensing data, and to generate spatial point cloud data using the real-time positioning and mapping SLAM module on the autonomous detection device; the front-end computing unit of the autonomous detection device performs real-time analysis on the multi-source sensing data to obtain analysis results, and dynamically generates priority tags that characterize the data rescue effectiveness based on the analysis results, so as to encapsulate the raw data into data packets with priority tags.

[0100] The execution module is used to autonomously probe the status of the communication link with the backend network, obtain the assessment results, and perform adaptive hierarchical transmission based on the assessment results and data packets with priority tags: when the link is connected, data packets are scheduled for back transmission according to the order of priority tags, with high-priority data packets enjoying a higher transmission guarantee level; when the link is interrupted, the autonomous probe caches data packets in the local storage unit according to priority order and continues to execute the probe task. After the communication link is restored, the high-priority data packets in the cache are retransmitted first.

[0101] The processing module is used by the edge command unit to receive high-priority data packets and spatial point cloud data retransmitted by autonomous detection devices. The edge command unit uses the spatial point cloud data as a unified spatial reference, assigns spatial coordinates to each data packet, and associates and fuses different types and time-related sensing data with the same or adjacent spatial coordinates to construct a digital map based on a unified spatiotemporal index.

[0102] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A data transmission and storage method for autonomous exploration in confined spaces, characterized in that, The method includes: Step 1: The autonomous detection device moves within a confined space and collects multi-source sensing data. It then uses the real-time positioning and mapping SLAM module on the autonomous detection device to generate spatial point cloud data. The front-end computing unit of the autonomous detection device performs real-time analysis on the multi-source sensing data to obtain analysis results. Based on the analysis results, it dynamically generates priority tags that characterize the data rescue effectiveness, and encapsulates the raw data into data packets with priority tags. Step 2: The autonomous detection device assesses the communication link status with the backend network, obtains the assessment results, and performs adaptive hierarchical transmission based on the assessment results and data packets with priority tags: When the link is connected, data packets are scheduled for back transmission according to the order of priority tags, with high-priority data packets enjoying a higher transmission guarantee level; when the link is interrupted, the autonomous detection device caches data packets in local storage units according to priority order and continues to perform the detection task. After the communication link is restored, the high-priority data packets in the cache are retransmitted first. Step 3: The edge command unit receives high-priority data packets and spatial point cloud data retransmitted by the autonomous detection equipment. The edge command unit uses the spatial point cloud data as a unified spatial reference, assigns spatial coordinates to each data packet, and associates and fuses different types and time-related sensing data with the same or adjacent spatial coordinates to construct a digital map based on a unified spatiotemporal index.

2. The data transmission and storage method for autonomous exploration in confined spaces according to claim 1, characterized in that, Step 1: The autonomous detection device moves within a confined space and collects multi-source sensing data. It then uses the onboard Simultaneous Localization and Mapping (SLAM) module to generate spatial point cloud data, including: Step 1.1: The front-end computing unit processes multi-source perception data in parallel, including at least thermal imaging video stream, visible light video stream and environmental audio stream, and performs human target detection, posture recognition and abnormal sound recognition respectively to obtain recognition results; Step 1.2: Based on the identification results, a comprehensive judgment is made. When a confirmed vital sign is identified, a first priority label is generated; when a suspicious vital sign is identified, a second priority label is generated; when environmental hazard data is identified, a third priority label is generated; and when a scene without abnormalities is identified, a fourth priority label is generated. Step 1.3: Differentiate the encapsulation of the raw sensing data according to the priority labels: encapsulate the data carrying the first priority label into a data packet containing high-fidelity compressed sensing data; encapsulate the data carrying the second and third priority labels into a compressed data packet containing only key feature information; and encapsulate the data carrying the fourth priority label into a data packet containing only status reports.

3. The data transmission and storage method for autonomous exploration in confined spaces according to claim 2, characterized in that, The front-end computing unit of the autonomous detection equipment performs real-time analysis on multi-source sensing data to obtain analysis results, and dynamically generates priority tags representing the data rescue effectiveness based on the analysis results. This encapsulates the raw data into data packets with priority tags, including: Step 1.4: Receive the recognition results, including human target detection boxes and pose classification results; Step 1.5: Confidence assessment and spatiotemporal alignment of each recognition result: Match and fuse detection boxes with spatial location correlation within the same time window with the recognition results to obtain comprehensive judgment data; Step 1.6: Based on the comprehensive analysis data and according to the predefined analysis rules, perform logical association judgment: if the stable human target in the thermal imaging video stream and the active human posture in the visible light video stream are successfully matched in time and space, then the first priority label is obtained; if only the stable human target in the thermal imaging video stream exists, then the second priority label is generated. Step 1.7: Based on the judgment conclusion, obtain the final priority label.

4. The data transmission and storage method for autonomous exploration in confined spaces according to claim 3, characterized in that, Step 2: The autonomous probe assesses the communication link status with the backend network, obtains the assessment results, and performs adaptive hierarchical transmission based on the assessment results and data packets with priority tags: When the link is connected, data packets are scheduled for backhaul according to the order of priority tags, with higher priority data packets enjoying a higher transmission guarantee level; when the link is interrupted, the autonomous probe caches data packets in local storage units according to priority order and continues to perform the probe task. After the communication link is restored, the high-priority data packets in the cache are retransmitted first, including: Step 2.1: The communication module of the autonomous detection device periodically detects and evaluates the quality of the communication link with the backend network to obtain the link status evaluation result; Step 2.2: Execute a branch decision based on the link status assessment result. If the link status assessment result is connected and stable, proceed to step 2.3; if the link status assessment result is interrupted, proceed to step 2.

5. Step 2.3: Establish a real-time transmission queue and insert the data packets with priority tags into the real-time transmission queue in descending order of priority tags. Among them, data packets with the first priority tag are assigned the highest transmission priority and are scheduled for transmission with the highest transmission guarantee level, including dedicated channel, enhanced forward error correction and shortened retransmission interval. Step 2.4: During network transmission, monitor the load status of the real-time transmission queue in real time. If the load is lower than the preset threshold, schedule and send data packets with second, third and fourth priority tags in sequence. Step 2.5: When the link is interrupted, the autonomous detection device switches to offline working mode and starts the data caching mechanism: the data packets to be transmitted are stored in the local shockproof and waterproof storage unit in the order of their priority tags, and the detection task continues to be executed. Step 2.6: In offline working mode, the communication module continuously attempts to restore the communication link; when the link is detected to be restored, it prioritizes extracting cached data packets with the first priority tag from the shockproof and waterproof storage unit for retransmission. After the high priority data packets are retransmitted or the network is idle, it retransmits the cached data of other priorities in sequence.

5. The data transmission and storage method for autonomous exploration in confined spaces according to claim 4, characterized in that, Step 3: The edge command unit receives high-priority data packets and spatial point cloud data retransmitted by the autonomous detection equipment. Using the spatial point cloud data as a unified spatial reference, the edge command unit assigns spatial coordinates to each data packet and correlates and fuses sensing data of different types and times with the same or adjacent spatial coordinates to construct a digital map based on a unified spatiotemporal index, including: Step 3.1: The edge command unit receives and parses the data transmitted back from the autonomous detection device, and obtains the perception data, corresponding priority tags, and data acquisition timestamps encapsulated in the data packet; at the same time, it receives and processes the spatial point cloud data stream generated by the SLAM module. Step 3.2, assign spatiotemporal coordinates to each received data packet: based on the data packet's acquisition timestamp, match the corresponding acquisition time and device pose in the spatial point cloud data stream, and calculate the three-dimensional spatial coordinates of the data packet's perceived content within the enclosed space; Step 3.3: Using the three-dimensional spatial coordinates and the acquisition timestamp as the joint index key, store the data packet with the completed coordinate assignment into the spatiotemporal fusion database; the spatiotemporal fusion database automatically associates sensing data with adjacent three-dimensional spatial coordinates, different acquisition timestamps, different types and different priorities to the same data record; Step 3.4: Based on the spatiotemporal fusion database and spatial point cloud data, construct a digital map of the enclosed space, and dynamically overlay the perceived data as an interactive layer onto the corresponding coordinate position of the digital map to obtain a digital map based on a unified spatiotemporal index.

6. The data transmission and storage method for autonomous exploration in confined spaces according to claim 5, characterized in that, Step 3.2, assign spatiotemporal coordinates to each received data packet: Based on the data packet's acquisition timestamp, match the corresponding acquisition time and device pose in the spatial point cloud data stream, and calculate the three-dimensional spatial coordinates of the data packet's perceived content within the enclosed space, including: Step 3.21: Extract the data collection timestamp carried in the data packet; Step 3.22: Based on the data acquisition timestamp, search for the device pose data synchronized with the timestamp in the spatial point cloud data stream. The device pose data includes the three-dimensional position and attitude of the autonomous detection device in the global coordinate system. Step 3.23: Obtain the fixed installation parameters of the sensing sensor corresponding to the data packet on the autonomous detection device. The installation parameters define the transformation relationship between the sensor coordinate system and the device body coordinate system. Step 3.24: Combining the device pose data and installation parameters, calculate the three-dimensional spatial coordinates of the data packet sensing content in the global coordinate system through coordinate transformation.

7. The data transmission and storage method for autonomous exploration in confined spaces according to claim 6, characterized in that, Step 3.4: Based on the spatiotemporal fusion database and spatial point cloud data, construct a digital map of the enclosed space, and dynamically overlay the perceived data as an interactive layer onto the corresponding coordinate positions of the digital map to obtain a digital map based on a unified spatiotemporal index, including: Step 3.41: Process the spatial point cloud data, including noise reduction, registration and reconstruction, to obtain a three-dimensional mesh model of the closed spatial structure, which serves as the static base of the digital map; Step 3.42: On the static base of the digital map, establish a map coordinate system consistent with the global coordinate system; Step 3.43: Determine the corresponding geographical location in the map coordinate system based on the three-dimensional spatial coordinates of each data packet; Step 3.44: Define different visualization layers for different types of sensing data, and dynamically render the determined geographical locations, along with the associated sensing data content, priority labels, and collection timestamps, as data markers onto the corresponding visualization layers. Step 3.45: Integrate the visualization layer and digital map base, and provide a map interaction interface; receive query commands for specific locations on the map through the map interaction interface, aggregate and return all multi-source historical sensing data related to the location from the spatiotemporal fusion database, and realize query and aggregate display by map location.

8. A data transmission and storage system for autonomous exploration in confined spaces, the system implementing the method as described in any one of claims 1 to 7, characterized in that, include: The acquisition module is used to move the autonomous detection device in a confined space and collect multi-source sensing data, and to generate spatial point cloud data using the real-time localization and mapping (SLAM) module on the autonomous detection device. The front-end computing unit of the autonomous detection equipment performs real-time analysis on multi-source sensing data to obtain analysis results, and dynamically generates priority tags that characterize the data rescue effectiveness based on the analysis results, so as to encapsulate the raw data into data packets with priority tags. The execution module is used to autonomously probe the status of the communication link with the backend network, obtain the assessment results, and perform adaptive hierarchical transmission based on the assessment results and data packets with priority tags: when the link is connected, data packets are scheduled for back transmission according to the order of priority tags, with high-priority data packets enjoying a higher transmission guarantee level; when the link is interrupted, the autonomous probe caches data packets in the local storage unit according to priority order and continues to execute the probe task. After the communication link is restored, the high-priority data packets in the cache are retransmitted first. The processing module is used by the edge command unit to receive high-priority data packets and spatial point cloud data retransmitted by autonomous detection devices. The edge command unit uses the spatial point cloud data as a unified spatial reference, assigns spatial coordinates to each data packet, and associates and fuses different types and time-related sensing data with the same or adjacent spatial coordinates to construct a digital map based on a unified spatiotemporal index.

9. A computing device, characterized in that, include: One or more processors; A storage device for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program that, when executed by a processor, implements the method as described in any one of claims 1 to 7.

Images