Accurate rescue method using micro-miniature intelligent robot in narrow and small space of earthquake

Abstract

The invention relates to the technical field of data processing, and provides an earthquake narrow space accurate rescue method using a micro intelligent robot. The method comprises the following steps: acquiring millimeter wave radar point cloud data and thermal imaging data of an infrared thermal imager based on a target area; a ruin environment three-dimensional model is prepared according to the data, and a passage path map is generated; self-adaptive networking is deployed for the micro robot cluster, and structural stability evaluation and residual vibration risk prediction are carried out in combination with the atlas; distributing detection partitions and rescue tasks of the micro robot cluster in real time based on an evaluation result; vital sign information is introduced, detection partition and task balanced distribution are optimized, and a parallel scheduling strategy is output. The technical problem that the rescue efficiency and accuracy in the narrow space are low due to the complex environment and incomplete sensing data is solved, and the technical effects that life search and rescue and task allocation are conducted in the narrow space through the micro robot cluster, and the accuracy and efficiency of earthquake narrow space rescue are improved are achieved.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a method for precise earthquake rescue in a narrow space using a micro-intelligent robot. Background Art

[0002] Earthquake disasters are characterized by suddenness, great destructiveness, and high difficulty in rescue. In particular, the collapse of buildings caused by earthquakes will form a large number of small and complex ruins. These spaces are often difficult to enter through traditional rescue methods, which seriously restricts the efficiency of rescue. Inside the ruins, the trapped people may be in an extremely vulnerable state, with limited living space, and face risks such as aftershocks and secondary structural collapse at any time. Therefore, how to achieve accurate, fast, and safe rescue in a small space has become a key issue that needs to be solved in the field of earthquake rescue.

[0003] Traditional rescue methods mainly rely on manual search and rescue and large mechanical equipment, but these methods have obvious limitations in small spaces: manual search and rescue is inefficient and risky, and large equipment is difficult to enter complex ruins. With the development of intelligent technology, micro-robots have gradually become an important tool for earthquake rescue in small spaces due to their small size, high flexibility, and strong adaptability. However, existing micro-robots still have deficiencies in environmental perception, path planning, and task allocation, making it difficult to meet the needs of precise rescue in complex ruins. Summary of the invention

[0004] This application provides a precise earthquake rescue method in a small space using a micro-intelligent robot, aiming to solve the technical problem of low rescue efficiency and accuracy in a small space due to complex environment and incomplete perception data.

[0005] The present application provides a method for precise rescue in a narrow space during an earthquake using a micro-intelligent robot, the method comprising: obtaining point cloud data corresponding to a millimeter-wave radar and thermal imaging data corresponding to an infrared thermal imager based on a target area; formulating a three-dimensional internal model of a ruin environment corresponding to a narrow space during an earthquake based on the point cloud data corresponding to the millimeter-wave radar and the thermal imaging data corresponding to the infrared thermal imager, and generating a traffic path map; deploying an adaptive network for a micro-robot cluster, and performing a structural stability assessment and residual vibration risk prediction in combination with the traffic path map; based on the structural stability assessment results and the residual vibration risk prediction results, using the adaptive network to allocate detection partitions and rescue tasks of the micro-robot cluster in real time; introducing vital sign sampling information, performing balanced allocation and optimization of the detection partitions and rescue tasks of the micro-robot cluster, and outputting a parallel scheduling strategy.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0007] The above-mentioned precise rescue method for earthquake narrow spaces using micro and mini intelligent robots comprehensively utilizes millimeter-wave radar and infrared thermal imagers to obtain data of the target area, constructs a three-dimensional model of the interior of the ruins based on this data, and generates a path map suitable for the passage of micro and mini robots. The robot cluster combines adaptive networking with the path map to conduct structural stability assessment and aftershock risk prediction to ensure that the robots operate in safe areas. Based on these assessment results, the detection areas and rescue tasks of the robot cluster are allocated in real time, and the tasks are balanced and optimized by collecting vital sign information (such as body temperature, micro-vibration, etc.). Finally, a parallel scheduling strategy is output to improve rescue efficiency and ensure the rationality of task priorities and allocations, ensuring the efficiency and precision of rescue.

[0008] The above description is only an overview of the technical solution of this application. In order to be able to understand the technical means of this application more clearly, it can be implemented in accordance with the content of the description. And in order to make the above and other purposes, features and advantages of this application more obvious and understandable, the specific embodiments of this application are specifically given below. Brief Description of the Drawings

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for description in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0010] Figure 1 It is a schematic flowchart of a precise rescue method for earthquake narrow spaces using micro and mini intelligent robots in an embodiment.

[0011] Figure 2 It is a schematic flowchart of the accuracy calibration of the three-dimensional model of the interior of the ruins environment in a precise rescue method for earthquake narrow spaces using micro and mini intelligent robots in an embodiment. Detailed Description of the Embodiments

[0012] The embodiments of this application solve the technical problems of low rescue efficiency and precision in narrow spaces due to complex environments and incomplete perception data by providing a precise rescue method for earthquake narrow spaces using micro and mini intelligent robots.

[0013] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the drawings in the embodiments of this application. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of this application without creative efforts belong to the scope of protection of this application.

[0014] It should be noted that the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules that are not explicitly listed or inherent to these processes, methods, products or devices.

[0015] Examples, such as Figure 1 As shown, the present application provides a precise rescue method for earthquake narrow spaces using a micro-sized intelligent robot, the method comprising:

[0016] Based on the target area, the point cloud data corresponding to the millimeter wave radar and the thermal imaging data corresponding to the infrared thermal imager are obtained.

[0017] In an embodiment of the present application, a millimeter-wave radar is first used to scan the target area to obtain point cloud data reflecting the distribution of objects in the area. This data can provide accurate distance and spatial information, which helps to identify structures, obstacles and gaps in the ruins. At the same time, an infrared thermal imager is used to obtain thermal imaging data of the target area. This data can show the temperature differences in the ruins and help identify potential living spaces and heat source signals of trapped people. By combining these two different types of data, the spatial structure and temperature distribution of the ruins environment can be fully understood, providing accurate basic information for subsequent modeling and rescue work.

[0018] According to the point cloud data corresponding to the millimeter wave radar and the thermal imaging data corresponding to the infrared thermal imager, a three-dimensional internal model of the ruins environment corresponding to the narrow space of the earthquake is drawn up, and a traffic path map is generated.

[0019] In one embodiment, based on the data obtained by the millimeter wave radar and the infrared thermal imager, the ruins environment will be analyzed in detail to establish an accurate three-dimensional space model. Specifically, first, the point cloud data obtained by the millimeter wave radar is aligned and fused with the thermal imaging data of the infrared thermal imager. The millimeter wave radar data provides the location, shape and spatial structure of objects in the ruins, while the thermal imaging data reveals the temperature changes in the ruins, especially the location of the heat source, which can help determine the location of trapped people or key equipment. Through data fusion technology, the information of both can be comprehensively considered to reduce the error that may be caused by a single data source. Subsequently, using the fused data, the point cloud data is converted into a three-dimensional space model through computer modeling software. This three-dimensional model will show the structural features of the walls, rubble accumulation, passages, gaps, etc. inside the ruins. At the same time, based on the thermal imaging data, the heat source areas will be marked in the model. These areas may be the locations of trapped people or places that need priority rescue. Through continuous iteration and optimization of the model, it is ensured that it restores the actual situation of the ruins as realistically as possible, thereby obtaining the final three-dimensional model of the ruins environment corresponding to the narrow space of the earthquake. After that, based on the constructed 3D model, it is necessary to generate a path map for the micro robot to safely pass through the ruins. In this process, the motion constraints of the robot itself are first defined, such as the size of the body, the range of motion of the joints, etc., and then combined with the spatial structure of the 3D model of the ruins, a preliminary feasible path is calculated. The path will avoid obstacles and unstable structures in the ruins. Then, the safety and efficiency of the path are evaluated to ensure that the robot can avoid dangerous areas and complete the task efficiently in the shortest possible time. Finally, the path map obtained through the evaluation will guide the micro robot to the optimal route and ensure that its movement in the ruins is both safe and efficient, so that it can perform effective rescue tasks in complex ruins environments.

[0020] Further, if Figure 2 As shown, the present application provides a method for preparing a three-dimensional model of the interior of a ruin environment corresponding to a small earthquake space based on the point cloud data corresponding to the millimeter wave radar and the thermal imaging data corresponding to the infrared thermal imager, and the method also includes:

[0021] Detect the internal cavity structure of the ruins in the target area and generate a dielectric constant distribution map to identify the potential survival space of the three-dimensional model inside the ruins environment; based on the potential survival space of the three-dimensional model inside the ruins environment, monitor the structural deformation caused by aftershocks and dynamically update the danger zone markers; use the updated danger zone markers to accurately calibrate the three-dimensional model inside the ruins environment.

[0022] Preferably, after obtaining the three-dimensional model of the interior of the ruins environment, the three-dimensional model will be optimized. First, it is detected by high-frequency electromagnetic waves (such as microwaves or radar waves), and the propagation characteristics (reflection, refraction, absorption, etc.) in different media are used to identify the cavity structures in the ruins. The electromagnetic waves will generate different reflection waveforms when encountering different substances. The electromagnetic wave reflection characteristics of the cavities or voids inside the ruins usually have significant differences from the surrounding solid structures (such as bricks, stones, steel bars, etc.). By measuring the intensity, time delay, and frequency change of the reflected electromagnetic waves, the shape, position, and size of the cavities can be determined. After obtaining the electromagnetic wave reflection signals in the ruins area, the dielectric constant of each area in the ruins will be calculated according to the characteristics of the reflected waves. The dielectric constant is the ability of a substance to respond to an electric field, which directly affects the propagation speed and reflection intensity of electromagnetic waves in the medium. In the ruins environment, different substances (such as metals, concrete, wood, etc.) have different dielectric constants, and the cavities show obvious different electromagnetic reflection characteristics from the surrounding substances. Subsequently, according to the electromagnetic wave reflection information at each position, a dielectric constant distribution map of the ruins area will be generated. This map will show the dielectric constant values of different areas in the ruins. The cavities or voids usually appear as low dielectric constant areas, while solid substances (such as bricks, stones, steel bars, etc.) will show higher dielectric constants. This distribution map can help identify potential living spaces in the ruins. These cavities may be the hiding places of trapped people. According to the generated dielectric constant distribution map, the low dielectric constant areas are marked as potential living spaces. These areas are usually voids, caves, or incompletely collapsed spaces in the ruins, which are the areas that the robot should give priority to detect and search. After an earthquake occurs, the structure of the ruins will continue to be affected by aftershocks. By installing devices such as seismic sensors, accelerometers, or displacement sensors, the deformation conditions of each key structural part in the ruins are monitored in real time. These sensors can detect the tiny vibrations and displacements inside the ruins, especially those changes that may cause the structure to collapse further. Then, based on the monitored structural deformation data, combined with earthquake parameters such as the epicenter location, magnitude, and duration of the earthquake, a mathematical model is used to predict aftershocks. By simulating the possible impacts of aftershocks on the ruins structure, it is evaluated which areas may further collapse or be damaged in future aftershocks, and an aftershock collapse probability is generated. This prediction process can identify high-risk areas in advance, so as to guide subsequent rescue operations. After obtaining the aftershock collapse probability, this aftershock collapse probability will be compared with a preset probability threshold. When the aftershock collapse probability is greater than or equal to the preset probability threshold, the corresponding potential living spaces will be marked as dangerous areas. These marked areas are usually unstable due to the structural deformation caused by aftershocks and may lead to further collapse or crack expansion.As these dangerous area markers are updated, the original 3D model will be calibrated for accuracy, that is, the current potential living space will be mapped to the 3D model inside the ruins environment, and the dangerous area markers will also be mapped to ensure that the ruins model can reflect the latest structural changes. After accuracy calibration, the robot cluster can accurately navigate according to the latest 3D model, avoid new dangerous areas, and more accurately locate potential living spaces, improving the safety and efficiency of search and rescue work.

[0023] The mathematical model used in aftershock prediction can be constructed based on the long short-term memory network (LSTM). The training data used in the construction process include sample structural deformation data, sample earthquake parameters, sample ruins structural features (such as structural component types, material properties, etc.), and sample aftershock collapse probability. During the training process, the model structure is constructed, including the input layer, LSTM hidden layer, fully connected layer, and output layer. The input layer of the model is then defined, and the sample structural deformation data, sample earthquake parameters, and sample ruins structural features (such as structural component types, material properties, etc.) are combined to form a feature sequence as the input data of the input layer. After the input layer, an LSTM hidden layer is constructed to extract the time series features and dynamic change trends in the input data, and then one or more fully connected layers are connected to further integrate the features. Finally, the predicted value of the sample aftershock collapse probability is output through the output layer. After the model structure is built, the weights of the LSTM model are initialized by the random initialization method, and the training data is input into the initialized LSTM prediction model for forward propagation. The data is passed to the LSTM hidden layer through the input layer for feature extraction and learning of temporal dependencies. The feature vector is further integrated through the fully connected layer, and the output layer generates the prediction result of the aftershock collapse probability. After obtaining the prediction result, the mean square error (MSE) loss function is used to calculate the loss value between the predicted output and the sample aftershock collapse probability, and the gradient of the loss value to each neuron weight in the LSTM hidden layer, fully connected layer and output layer is calculated layer by layer through the back propagation algorithm. Subsequently, the Adam optimizer is used to iteratively optimize and update the model parameters, and the network weights and biases are dynamically adjusted to minimize the loss function value. The above process is repeated until the preset maximum number of iterations is reached, and the model parameters gradually converge to a stable state. After the training is completed, the model performance is evaluated using the validation set (data not used for training) data to calculate the accuracy of predicting the probability of aftershock structural collapse and the generalization ability of the model. If the verification results meet the expected requirements, the current LSTM structural model will be output as the final aftershock structural deformation prediction model; if the expectations are not met, the hyperparameters such as learning rate, number of hidden layer nodes, training batch size, etc. will be further adjusted to continuously improve the prediction accuracy of the model.

[0024] Furthermore, the present application provides a method for generating a traffic path map, the method comprising:

[0025] The robot motion capability constraint matrix is ​​defined, including body size, joint range of motion, and obstacle crossing quantitative parameters; based on the robot motion capability constraint matrix, an initial feasible path is calculated, and the path nodes are embedded in the potential living space of the three-dimensional model inside the ruin environment; the safety and efficiency of the initial feasible path are evaluated to determine the pass path map.

[0026] Optionally, in order to accurately plan the movement route of the micro-robot in the earthquake ruins, it is necessary to define a constraint matrix for the robot's movement ability. This matrix includes the body size of the robot (such as length, width, height), the maximum range of joint movement (such as the rotation angle and extension length of the robotic arm), and the quantitative parameters of the obstacle-crossing ability such as the height of the obstacle that the robot can cross and the maximum slope. These parameters are used to clarify the spatial limitations and ability range of the robot's movement in the actual environment. Subsequently, the three-dimensional model of the ruins environment established previously is voxelized or meshed to form a structured environmental grid. Each grid cell stores structural features such as the storage space size, obstacle distribution, and cavity position. After meshing, a data structure that is easy to perform path search calculations is formed, which is convenient for the implementation of the path planning algorithm. Then, the positions of the potential survival spaces determined previously by means of millimeter-wave radar, thermal imaging, dielectric constant detection, etc. are associated with the environmental grid data to clarify the coordinates of the survival space positions and the corresponding grid cells. These potential survival spaces are embedded as important path nodes into the path planning to ensure that the generated path can cover all the space areas that need to be searched and rescued. After that, according to the robot movement ability constraint matrix, a path search algorithm (such as A*, Dijkstra algorithm or RRT rapid search tree algorithm) is used to calculate the initial feasible path in the three-dimensional ruins environment. Exemplarily: taking the starting position of the robot as the starting point of the path; taking the positions of the potential survival spaces as the node targets that the path must pass through; gradually checking each path node through the movement constraint matrix, and eliminating the nodes that do not match the size, are difficult to cross obstacles, have joint limitations, or have obvious blockages; calculating the initial paths connecting each potential survival space node in turn according to the areas that the robot can actually reach, so as to obtain a set of initial feasible paths. The starting point of the robot's movement, the target search and rescue space, the path turning points, etc. are clearly marked in these path nodes. Then, the determined path nodes are spatially matched with the positions of the potential survival spaces. That is, when the initial path passes through or is close to the potential survival space, the central position of the survival space is accurately marked on the path node, so that the robot can clearly stay at the node, perform detailed detection and rescue operations during subsequent task execution; if it is found that the initial path does not pass through the potential survival space sufficiently, the path nodes can be further optimized, and the path can be added or adjusted to cover all the survival spaces as much as possible. Finally, a further evaluation and analysis of the generated initial feasible path is carried out.Specifically, each path will be evaluated for safety and efficiency. The safety assessment focuses on the stability of the structure around the path, whether there is a potential risk of collapse or aftershocks, and the path safety score of the path is obtained by weighting the probability of no collapse due to aftershocks (obtained by subtracting the probability of collapse due to aftershocks from 1) of each node on the path, where the weight of each node is determined according to the number of obstacles to reach the node; the efficiency assessment focuses on the path length, the estimated time to reach the target area, etc., and the path efficiency score is obtained by weighting the total path length and total time. By weighting the path safety score and path efficiency score, one or several optimal paths are determined, thereby forming a complete robot action path map to guide the robot cluster to efficiently and safely implement precise rescue operations.

[0027] Adaptive networking is deployed for the micro-robot cluster, and the structural stability assessment and residual vibration risk prediction are performed in combination with the travel path map.

[0028] In one embodiment, in the rescue mission of a micro-robot cluster, deploying an adaptive networking is the key to ensuring the effective cooperation of the cluster and its ability to respond flexibly to emergencies. The adaptive networking can adjust the communication and collaboration modes of the robots in real time according to the mission requirements and environmental changes. First, based on the mission requirements, the robot cluster will connect to the central control system through a wireless communication network. Each robot dynamically connects and disconnects from other robots through the adaptive networking technology to ensure stable communication and coordinated work among the robots in the cluster. The adaptive networking can adjust the network topology according to the position, mission importance, and communication quality of each robot, thus avoiding information delay or interruption caused by communication problems. During this process, the communication network of the robots will monitor and update the network status in real time. When the robots move or the environment changes, the network topology will be automatically adjusted to ensure the stability and real-time nature of data transmission. For example, if the signal is poor in a certain area, the robot cluster will dynamically select a more stable path for information transmission. Subsequently, according to the previously constructed passage path map, the structural stability of the ruins is evaluated. The path map not only contains the specific routes for the robots to travel but also marks the safety indicators of each area in the ruins. The robot cluster will travel along the paths in the map and continuously monitor environmental changes through sensors (such as millimeter-wave radars, infrared thermal imagers, etc.). If a new breakage or instability phenomenon is found in the structure when a robot passes through a certain area, data such as strain and stress distribution in that area will be collected and transmitted back in real time. Through the real-time transmitted environmental data, the possible risk areas in the ruins can be updated on the three-dimensional model of the ruins environment, and a structural stability analysis will be carried out, that is, a finite element solver (such as ANSYS, ABAQUS) will perform numerical calculations on the updated model. The solver will solve the displacement, stress, and strain distribution of the structure through numerical methods (such as direct methods, iterative methods), obtain the response of the entire ruins structure, and evaluate the bearing capacity of the risk area based on this to determine its stability, thus obtaining the structural stability evaluation result. In addition, a residual vibration risk prediction will also be carried out for the risk area, that is, using the previously constructed mathematical model, calculating the aftershock collapse probability according to the updated three-dimensional model of the ruins environment, and taking it as the residual vibration risk prediction result. The obtained structural stability evaluation result and residual vibration risk prediction result will be used to dynamically adjust the task allocation to ensure the efficiency and safety of the rescue work.

[0029] Based on the structural stability evaluation result and the residual vibration risk prediction result, the detection areas and rescue tasks of the micro-robot cluster are allocated in real time by the adaptive networking.

[0030] In one embodiment, based on the evaluation results of the stability of the ruins structure and the prediction results of the residual vibration risk, the detection areas and rescue tasks of the micro-robot cluster can be dynamically adjusted. First, the structural stability evaluation provides the bearing capacity of each risk area of the ruins, and the residual vibration risk prediction provides the collapse probability of each risk area. These data are combined to form a comprehensive ruins risk map. On this basis, through the adaptive networking technology, according to the current position, function, health status of each robot and the real-time feedback in the environment, tasks are dynamically allocated. The robot cluster will, according to the instructions, go to high-risk areas for more in-depth detection, while avoiding areas that have been marked as high-risk to ensure the safety of the cluster and the efficient execution of tasks. For example, if the stability of a certain area is lower than the lower limit of stability and the collapse probability is higher than the upper limit of the collapse probability, it will be marked as a high-risk area, and robots in other areas will be preferentially assigned to perform search and rescue tasks to avoid robots in high-risk areas becoming potential traps. At the same time, the adaptive networking technology ensures that detection data can be shared in real time among various robots, coordinate with each other, and cooperate to complete tasks. If a certain robot detects a new danger signal (such as further structural damage, changes in vital signs, etc.), the data it discovers will be transmitted to other robots and systems in the cluster in a timely manner, and these feedback data will immediately update the task allocation to ensure the work progress and efficiency of the entire robot cluster. Finally, all robots will perform balanced allocation and optimization of tasks according to the structural stability and risk assessment results to ensure that the areas most in need of help can receive timely and sufficient rescue, while avoiding the danger caused by structural instability to the greatest extent.

[0031] Introduce the vital sign sampling information, perform balanced allocation and optimization on the detection areas and rescue tasks of the micro-robot cluster, and output a parallel scheduling strategy.

[0032] In one embodiment, during rescue operations in earthquake ruins, the task allocation of the robot swarm not only needs to consider the structural stability of the ruins and the risk of aftershocks, but also the vital sign information of the trapped people. Introducing vital sign sampling information can greatly improve the rescue efficiency and the rationality of priority allocation. The vital sign information includes important parameters such as body temperature and micro-vibrations, and this information can be collected in real time by sensors carried by the robots. During the task execution, each robot will detect and collect the vital sign data of the trapped people within its coverage area. When the vital sign data in a certain area indicates the presence of trapped people and the vital signs show that they are in a critical state, a higher priority will be assigned to this area, and robots will be promptly dispatched for rescue. Through the real-time collected vital sign information, all detection areas can be dynamically optimized, and the task allocation strategy can be adjusted according to the life conditions of the trapped people, the danger of the environment they are in, and the available resources of the robot swarm. For example, if the vital sign data indicates that the trapped people in some areas may be in a critical state due to lack of oxygen or severe injuries, the tasks will be automatically adjusted to prioritize sending robots to these areas for emergency rescue. On this basis, balanced scheduling will also be carried out according to the cooperation situation among the robots. Considering the scale of the robot swarm and the working capabilities of different robots (for example, some robots may need more time for material replenishment or maintenance), the task load of each robot will be dynamically adjusted to ensure the efficient operation of the robot swarm and prevent the situation of some robots being overloaded or resource waste. The parallel scheduling strategy enables multiple tasks to be carried out simultaneously, minimizing the rescue time and reducing secondary injuries caused by structural instability or vibrations. In summary, by introducing vital sign information and combining real-time data analysis, the detection partitions and rescue tasks can be comprehensively optimized to ensure that the rescue activities are carried out efficiently, scientifically, and accurately, while ensuring that the most critical life rescue tasks are given priority.

[0033] Furthermore, the present application provides that the vital sign sampling information includes body temperature marker data, micro-vibration marker data, and sweat marker data; based on the micro-vibration marker data, the surface micro-vibration signal of the trapped person is extracted, and the respiratory frequency feature and the heartbeat waveform feature are separated; through the respiratory frequency feature and the heartbeat waveform feature, the critical state level is evaluated, and the critical state level is used to rank the priorities of the rescue tasks.

[0034] Preferably, during earthquake rescue operations, a swarm of robots collects vital sign information of the trapped victims through various sensors to understand their health status in real time. Especially in the rubble, many trapped victims may face life-threatening situations due to the long duration and harsh environment. To quickly determine the criticality of each trapped victim, rescue priorities are established by analyzing vital sign markers. The vital sign information collected mainly includes body temperature marker data, micro-vibration marker data, and sweat marker data. Body temperature markers can help determine whether the trapped victim is in a low or high temperature environment, and then infer their physical exertion, whether there is shock, etc. Sweat markers can help identify the stress response of the trapped victim and understand whether their body is in an abnormal state due to dehydration, overexertion, or excessive stress. Among them, micro-vibration marker data is particularly crucial. Through micro-vibration sensors, the robots can capture the minute vibration signals on the surface of the trapped victims. These micro-vibration signals contain information such as the breathing frequency and heartbeat waveform of the trapped victims. By extracting and analyzing these micro-vibration signals, the breathing frequency characteristics and heartbeat waveform characteristics can be accurately separated. The breathing frequency can reflect the breathing condition of the trapped victim. If the breathing is too fast or too slow, it may indicate unstable vital signs. The heartbeat waveform can show the working state of the heart, and an abnormal waveform may indicate the risk of internal injuries or other life-threatening diseases. Based on the breathing frequency and heartbeat waveform characteristics, the criticality level of the trapped victim can be further evaluated. If the breathing frequency is abnormal, the heartbeat waveform is disordered, or the body temperature is too low / high, it is considered that the trapped victim is in a high-risk state, and resources will be preferentially allocated for rescue. Through this evaluation, the swarm of robots can intelligently rank the priorities of various rescue tasks to ensure that the most critical trapped victims receive the fastest rescue. Therefore, by comprehensively analyzing vital sign data, especially micro-vibration marker data, the health status of the trapped victims can be efficiently evaluated, and based on the evaluation results, the rescue tasks of the swarm of robots can be reasonably arranged to ensure the maximum utilization of rescue resources and the safety of the trapped victims' lives.

[0035] Further, the criticality level of the state is used to rank the priorities of the rescue tasks, and the method further includes:

[0036] Deploy a digital twin engine in the adaptive networking to real-time simulate the interaction forces between the movement trajectories of the micro-robot swarm and the rubble structure; through the simulation data of the interaction forces between the movement trajectories of the micro-robot swarm and the rubble structure, predict the risk of secondary collapse caused by manipulator operations; according to the secondary collapse risk, generate a structural support reinforcement task to the rescue task queue.

[0037] Optionally, in the adaptive networking, a digital twin engine is deployed to simulate in real time the interaction forces between the movement trajectories of the micro and small robot clusters and the structure of the ruins. The digital twin engine is a virtual model that can reflect in real time the mutual influence between the robot and the structure when the robot moves in the ruins by highly accurately reconstructing the three-dimensional environment of the ruins and the dynamic behavior of the robot cluster. Whenever the robot cluster executes a task, the digital twin engine interacts through sensor data, robot movement data, and the internal three-dimensional model of the ruins environment to update in real time the relative positions, contact forces between the robot and the ruins, and the pressure generated by the robot movement on the structure. During this real-time simulation process, special attention is paid to the operation of the robotic arm because the robotic arm usually needs to perform high-intensity physical operations, such as carrying heavy objects or clearing obstacles, and the actions of the robotic arm may affect the structural components in the ruins, especially when the structure is already partially damaged or unstable. Through the simulation of the digital twin engine and the calculation of finite element analysis, the risk of secondary collapse that may be caused by the robotic arm when performing tasks can be understood in advance. Secondary collapse refers to the phenomenon that the structure after the initial damage may further collapse when subjected to external forces, which poses a great threat to the rescue work. Based on the secondary collapse risk data obtained through the simulation of the digital twin engine, the areas with high risks are automatically evaluated, and structural support reinforcement tasks are generated based on this information. These reinforcement tasks will be preferentially added to the rescue task queue. The purpose of the reinforcement tasks is to reduce the possibility of secondary collapse by providing additional support or repairing the weaknesses of the structure before other rescue operations are carried out, thereby ensuring the safety of the robot cluster and the trapped people. The reinforcement tasks may include strengthening the structure, stabilizing key areas with temporary supports, or marking and isolating the areas in the ruins that may cause collapse to ensure that subsequent rescue tasks can be carried out smoothly without causing more serious damage. This process can dynamically adjust the rescue plan through the real-time feedback of the digital twin engine to ensure that not only can the trapped people be quickly found during the rescue operation, but also potential secondary collapse risks can be effectively avoided, improving the overall rescue efficiency and safety.

[0038] Furthermore, this application provides that when the rescue task queue is updated, a federated learning mechanism is triggered to integrate the operation data of each robot corresponding to the micro and small robot cluster; through the federated learning mechanism, combined with the criticality level, the detection partition of the micro and small robot cluster is updated online.

[0039] Optionally, when the rescue task queue is updated, the federated learning mechanism is triggered. The purpose is to integrate the operation data of each robot in the micro and small robot cluster and dynamically adjust the detection tasks based on this data. Specifically, as the on-site rescue situation changes, for example, newly discovered high-risk areas, changes in the status of trapped persons, or changes in the progress of robot work, the rescue task queue will be updated in real time. The update of the queue is usually sorted and adjusted according to multiple factors such as task priority, vital signs of rescue personnel, and structural stability of the ruins. Once the task queue changes, the federated learning mechanism will be triggered. Federated learning is a distributed machine learning method. Under this method, each robot in the robot cluster will process and learn its own operation data (such as path information, task completion status, environmental changes, etc.) locally without sending the data to the system. This mechanism can ensure the privacy and real-time nature of the data, while avoiding the system response lag caused by data transmission delay. Each robot analyzes the local data and updates its own learning model. For example, it optimizes itself by combining its own task execution situation (such as obstacles encountered, efficiency of path selection, etc.). These data are not directly transmitted to the system, but are aggregated into a higher-level global model through the federated learning mechanism. Robots optimize their individual models by collaborating and sharing information, thereby improving the working efficiency and rescue effect of the entire robot cluster. On the basis of federated learning, further task optimization will be carried out in combination with the criticality level of the status of trapped persons on site. That is, the criticality of each trapped person is judged according to their vital signs (such as respiratory rate, heartbeat waveform, etc.). More robot resources will be preferentially allocated to areas with a higher criticality level. This priority adjustment will be updated in real time through the control system of the robot cluster. Subsequently, using the global model integrated by federated learning, the division of the detection area will be updated in real time. The robot cluster will re-plan its working area and task objectives according to the latest learning model of each robot. For example, if the danger level of a certain area increases or new vital sign information is fed back, the task allocation will be automatically adjusted to increase the detection intensity of the robot in that area, or shift the task focus to areas that require more resources. Each robot reallocates its work within the updated detection area to ensure maximum efficiency while avoiding over-concentration in certain areas. Through the federated learning mechanism, each member of the robot cluster can continuously learn and adapt to the task environment locally and dynamically cooperate with other robots. When the task queue is updated or the on-site situation changes, the robot cluster can quickly respond and optimize resource scheduling to ensure the real-time and efficient execution of tasks. The cooperation between robots enables the entire cluster to efficiently complete search and rescue tasks in a complex and dynamic ruins environment, while avoiding resource waste and task duplication. This can not only improve the rescue efficiency, but also enhance the safety and flexibility during the rescue process.

[0040] Furthermore, the present application provides freezing of high-risk area operation instructions when the arrival of a P wave is detected, and activating a preset obstacle avoidance path to guide the micro-robot cluster to withdraw to a safe area.

[0041] Optionally, when an earthquake occurs, especially when the P wave (first arrival wave) arrives, the initial signal of the vibration can be detected in real time through the sensor. The P wave is the fastest arriving wave in an earthquake and can usually be sensed within a few seconds after the earthquake, so it can provide early warning for subsequent strong earthquake waves (S waves and surface waves). The micro-robot cluster monitors the vibration of the ground in real time in the ruins area through its sensors (such as accelerometers, vibration sensors, etc.). When these vibration sensors capture the vibration signal of the P wave, they will check all the tasks currently in progress, especially those in high-risk areas. Subsequently, all the operation instructions in these high-risk areas (such as secondary collapse, etc.) will be frozen, that is, the current task execution will be stopped to prevent the robot from continuing to move in these dangerous areas, and the obstacle avoidance mechanism will be immediately activated to guide the robot away from these potential dangerous areas through the preset obstacle avoidance path. Each robot will adjust its route of action according to the passage path map, structural stability data and real-time changes in the environment. Through dynamic path planning and obstacle avoidance guidance, the robot cluster will follow the safe evacuation route, quickly evacuate the high-risk area, and return to the safe area. The path during the evacuation process will take into account obstacles, damage, and the dynamic conditions of other robot clusters in the ruins environment to ensure the safe evacuation of the robots and avoid collisions or being trapped. Although the robot cluster will be guided to a safe area after the P wave arrives, the risk level of different areas will be assessed based on actual conditions. Once aftershocks and other dangers are over, the robots will re-enter the ruins based on the latest structural assessment data and continue to perform rescue missions. The entire evacuation and recovery process is highly automated, and the robots can continuously adjust their behavior and paths based on real-time data to ensure that the mission can be performed safely and efficiently. Through this mechanism, not only can secondary disasters be effectively prevented when an earthquake occurs, but it can also ensure that the robot cluster is always in the safest working state and reduce unnecessary losses.

[0042] Furthermore, the present application provides loading a pre-earthquake building BIM model in a digital twin environment, performing non-rigid registration evaluation with post-earthquake point cloud data, and identifying the residual bearing capacity attenuation coefficient of key load-bearing components; introducing micro-crack topology data, combined with the residual bearing capacity attenuation coefficient of key load-bearing components, to identify the safe area of ​​the internal three-dimensional model of the ruins environment.

[0043] Optionally, load the BIM (Building Information Model) of the pre-earthquake building, which is an accurate three-dimensional digital model that shows detailed information such as the structure, materials, and layout of the building. The pre-earthquake BIM model provides the structural state of the building under normal conditions, including the design and expected performance indicators of each load-bearing component. After an earthquake occurs, use point cloud data to conduct a detailed three-dimensional scan of the post-earthquake ruins. These point cloud data represent the actual structure of the ruins and can reflect the deformation, damage, and collapse states of the building after the earthquake. Compare and analyze these post-earthquake point cloud data with the pre-earthquake BIM model. This process is called non-rigid registration. Non-rigid registration is to match the pre-earthquake BIM model with the post-earthquake point cloud data in space. Different from rigid registration (which only considers rotation and translation), non-rigid registration allows local deformations to occur in the model, such as cracks, structural loosening, or deformation, etc., which can all be captured through non-rigid registration. The registration algorithm is usually based on a deformation model, allowing two different data sets (the BIM model and the point cloud data) to be matched and compared. During the registration process, the post-earthquake point cloud data will be cleaned and denoised to remove irrelevant data or incorrect points. Then, common features in the point cloud data and the BIM model, such as key structural nodes or geometric features, are selected for matching. The matching results are optimized through the deformation model to find an optimal matching state. This process allows the point cloud data to deform in certain areas so as to align with the pre-earthquake BIM model. After registration, identify the main load-bearing components (such as load-bearing walls, beams, columns, etc.) in the building through the BIM model. These components are crucial for the stability of the entire structure. Subsequently, combine the post-earthquake point cloud data and the BIM model. By comparing and analyzing the deformation, cracks, fractures, etc. of each load-bearing component, evaluate its damage degree. For example, cracks or deformations occurring in beams and columns can indicate the attenuation of their load-bearing capacity. Then use finite element analysis to calculate the residual load-bearing capacity of each key load-bearing component. By comparing the post-earthquake damage with the pre-earthquake designed load-bearing capacity, obtain the load-bearing capacity attenuation coefficient of each component. This coefficient represents the ratio of the load that the component can still bear after experiencing the earthquake. After that, introduce micro-crack topology data. Micro-crack topology data refers to detecting the cracks in the building through high-precision sensors (such as laser scanning, digital image processing, or ultrasonic testing, etc.) to obtain information such as the distribution, shape, width, and depth of the cracks. These data provide detailed damage information of the building after the earthquake, especially the microscopic morphology of the cracks, which helps to analyze the damage degree and safety of the structure in different parts. By combining the crack data with the residual load-bearing capacity attenuation coefficient of the key load-bearing components, analyze whether the cracks penetrate these key parts and affect their load-bearing capacity. For example, if the cracks in the load-bearing wall are large, it may mean that the load-bearing capacity in this area has decreased significantly, thereby affecting the stability of the overall structure. Then use the information such as the depth, length, and width of the cracks to correct the attenuation coefficient of each load-bearing component. For areas with more serious cracks, the attenuation coefficient will be lower, and vice versa.Then, by combining the microscopic crack topology data and the residual bearing capacity attenuation coefficient, we can generate the safety index of each area in the three-dimensional model of the ruins. The safety of these areas is jointly determined by the following factors: First, areas with dense cracks or large crack extensions indicate that these places may have lost their original bearing capacity and are at risk of collapse, which means that the safety is low; second, combined with the crack data, check whether the bearing capacity of each load-bearing component is lower than the safety threshold. If the bearing capacity of the load-bearing components in a certain area is significantly reduced due to the influence of cracks, the safety is low. After the risk assessment, the safe areas in the ruins will be determined based on the crack topology data and the bearing capacity analysis results. These safe areas refer to areas where the structure is still stable and has sufficient bearing capacity, divided according to the crack distribution and structural attenuation coefficient. As safe areas, these areas usually have fewer cracks and the bearing capacity of key load-bearing components has not been significantly attenuated. The identification of safe areas is very critical for rescue operations because it can help rescuers or robot clusters avoid dangerous areas and concentrate their efforts on rescue missions in safer areas. In summary, this process combines the BIM model of the pre-earthquake building with the post-earthquake point cloud data for non-rigid registration and analysis, which can accurately assess the structural damage and bearing capacity changes of the building, and identify safe areas in the ruins, providing a scientific basis for subsequent rescue operations.

[0044] In summary, the embodiments of the present application have at least the following technical effects:

[0045] The embodiment of the present application first obtains the point cloud data corresponding to the millimeter wave radar and the thermal imaging data corresponding to the infrared thermal imager based on the target area; then, according to the point cloud data corresponding to the millimeter wave radar and the thermal imaging data corresponding to the infrared thermal imager, a three-dimensional model of the internal environment of the ruins corresponding to the narrow space of the earthquake is drawn up, and a passage path map is generated; then, an adaptive network is deployed for the micro-robot cluster, and the structural stability evaluation and residual vibration risk prediction are performed in combination with the passage path map; then, based on the structural stability evaluation results and the residual vibration risk prediction results, the detection partitions and rescue tasks of the micro-robot cluster are allocated in real time by the adaptive network; finally, vital signs sampling information is introduced to optimize the balanced allocation of the detection partitions and rescue tasks of the micro-robot cluster, and a parallel scheduling strategy is output. These technical effects jointly solve the technical problem of low efficiency and accuracy of rescue in narrow spaces due to complex environments and incomplete perception data, and realize the technical effect of using micro-robot clusters to search and rescue lives and allocate tasks in narrow spaces, and improve the accuracy and efficiency of rescue in narrow spaces during earthquakes.

[0046] It should be noted that the above order of the embodiments of the present application is only for description and does not represent the superiority or inferiority of the embodiments. And the above description of specific embodiments of this specification has been made. The processes depicted in the drawings do not necessarily require the specific order and continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0047] The above are only the preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included within the protection scope of the present application.

[0048] This specification and the drawings are only exemplary descriptions of the present application and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of the present application. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalent technologies, the present application is intended to include these changes and modifications.

Claims

1. A precise earthquake rescue method in a narrow space using a micro-intelligent robot, characterized in that: The method comprises: Based on the target area, obtain the point cloud data corresponding to the millimeter wave radar and the thermal imaging data corresponding to the infrared thermal imager; According to the point cloud data corresponding to the millimeter wave radar and the thermal imaging data corresponding to the infrared thermal imager, a three-dimensional model of the interior of the ruins environment corresponding to the narrow earthquake space is drawn up, and a traffic path map is generated; Adaptive networking is deployed for the micro-robot cluster, and structural stability assessment and residual vibration risk prediction are performed in combination with the passage path map; Based on the structural stability assessment results and the residual vibration risk prediction results, the detection partitions and rescue tasks of the micro-robot cluster are allocated in real time by the adaptive networking; Vital sign sampling information is introduced to optimize the balanced allocation of detection partitions and rescue tasks of the micro-robot cluster, and a parallel scheduling strategy is output.

2. The earthquake narrow space precision rescue method using a micro-sized intelligent robot as claimed in claim 1, characterized in that: According to the point cloud data corresponding to the millimeter wave radar and the thermal imaging data corresponding to the infrared thermal imager, a three-dimensional model of the interior of the ruins environment corresponding to the narrow earthquake space is drawn up, and the method further includes: Detecting the internal cavity structure of the ruins in the target area and generating a dielectric constant distribution map to identify potential living space in the three-dimensional model of the interior of the ruins environment; Based on the potential living space of the three-dimensional model inside the ruins environment, the structural deformation caused by aftershocks is monitored and the dangerous area marking is dynamically updated; The updated danger zone markers are used to perform precision calibration on the internal three-dimensional model of the ruin environment.

3. The earthquake narrow space precision rescue method using a micro-sized intelligent robot as claimed in claim 2, characterized in that: Generate a traffic path map, the method comprising: Define the robot's motion capability constraint matrix, including body size, joint range of motion, and obstacle crossing quantitative parameters; Based on the robot motion capability constraint matrix, an initial feasible path is calculated, and the path nodes are embedded in the potential living space of the three-dimensional model inside the ruin environment; The safety and efficiency of the initial feasible path are evaluated to determine the pass path map.

4. The earthquake narrow space precision rescue method using a micro-sized intelligent robot as claimed in claim 1, characterized in that: The vital sign sampling information includes body temperature marker data, micro-vibration marker data and sweat marker data; Based on the micro-vibration marker data, extract the micro-vibration signal of the trapped person's body surface, and separate the respiratory frequency feature and the heartbeat waveform feature; The state criticality level is evaluated through the respiratory rate characteristics and the heartbeat waveform characteristics, and the state criticality level is used to prioritize the rescue tasks.

5. The earthquake narrow space precision rescue method using a micro-sized intelligent robot as claimed in claim 4, characterized in that: The state criticality level is used to prioritize the rescue missions, and the method further includes: Deploy a digital twin engine in the adaptive network to simulate the interaction force between the motion trajectory of the micro-robot cluster and the ruins structure in real time; The risk of secondary collapse caused by the operation of the robot arm is predicted by simulating the interaction force between the motion trajectory of the micro-robot cluster and the ruin structure; Based on the secondary collapse risk, a structural support reinforcement task is generated to the rescue task queue.

6. The precise earthquake rescue method in a narrow space using a micro-intelligent robot as claimed in claim 5, characterized in that: When the rescue task queue is updated, the federated learning mechanism is triggered to integrate the operation data of each robot corresponding to the micro-robot cluster; Through the federated learning mechanism, combined with the state criticality level, the detection partition of the micro-robot cluster is updated online.

7. The precise earthquake rescue method in a narrow space using a micro-intelligent robot as claimed in claim 6, characterized in that: When the arrival of the P wave is detected, the operation instructions in the high-risk area are frozen, and the preset obstacle avoidance path is activated to guide the micro-robot cluster to withdraw to the safe area.

8. The precise earthquake rescue method in a narrow space using a micro-sized intelligent robot as claimed in claim 7, characterized in that: Load the pre-earthquake building BIM model in the digital twin environment, perform non-rigid registration evaluation with the post-earthquake point cloud data, and identify the residual bearing capacity attenuation coefficient of key load-bearing components; The micro-crack topology data is introduced and combined with the residual bearing capacity attenuation coefficient of key load-bearing components to identify the safe area of ​​the internal three-dimensional model of the ruin environment.

Images