A post-disaster search and rescue detection system and method based on radio tomography

Abstract

The application discloses a kind of based on radio tomography's post-disaster search and rescue detection system and method, comprising: ultra-wideband wireless sensor network, robot cluster and cluster control and tomography system.The ultra-wideband wireless sensor network is divided into static ultra-wideband node and mobile ultra-wideband node, static node is used to assist mobile node positioning, all nodes are used for radio tomography;Robot cluster is used to carry mobile ultra-wideband node, each robot is configured with two nodes, and is configured with independent power supply, two-dimensional laser radar and communication module, robot can walk in complex road conditions, wide application range;The cluster control and tomography system is wirelessly interconnected with robot cluster, for controlling the moving track of robot, and according to the direct path signal received intensity signal collected in ultra-wideband wireless sensor network, the environment is tomographically imaged, and the detection of the victim in invisible area (smoke, collapse environment, etc.) is realized.

Description

Technical Field

[0001] This invention relates to a method and apparatus for post-disaster search and rescue detection, and more particularly to a post-disaster search and rescue detection system and method based on radio tomography. Background Technology

[0002] For disaster-stricken areas, post-disaster search and rescue is of paramount importance. At disaster sites, those buried or trapped are highly vulnerable to serious injury or death; without timely rescue, more innocent lives may be lost. Post-disaster victim detection helps rescuers quickly locate victims, significantly improving search and rescue efficiency. Currently, commonly used victim detection methods include manual search, search dogs, and search drones, but all have limitations. Especially in the face of blind spots such as collapsed environments after earthquakes or dense smoke from fires, accurate and rapid detection and location of victims is difficult, hindering the effective implementation of search and rescue operations. Summary of the Invention

[0003] Purpose of the invention: To address the above problems, this invention proposes a disaster search and rescue detection system and method based on radio tomography, which is easy to deploy and suitable for invisible environments such as dense smoke and collapsed buildings.

[0004] Technical solution: The technical solution adopted in this invention is a disaster search and rescue detection system based on radio tomography, including an ultra-wideband wireless sensor network, a robot swarm, and a swarm control and tomography system;

[0005] The ultra-wideband wireless sensor network includes stationary ultra-wideband nodes and mobile ultra-wideband nodes. The stationary ultra-wideband nodes are installed on stationary node devices at fixed locations to assist the positioning of mobile nodes, while the mobile ultra-wideband nodes are mounted on robots. Each ultra-wideband node transmits pulse signals, which are received by other nodes. Each ultra-wideband node is equipped with a signal processing module to process the received signals of the ultra-wideband node, obtain the signal reception strength data of the direct path between each node link, and send it to the cluster control and tomography system in real time through the communication module.

[0006] The robot cluster includes several robots, each equipped with a lidar and two mobile ultra-wideband nodes A and B with a fixed distance between them. The lidar is used by the robots to build environmental maps and identify obstacles in real time, and transmits the data to the cluster control and tomographic imaging system in real time via a communication module.

[0007] The cluster control and tomography system calculates the position of the mobile ultrawideband node in real time with the assistance of stationary nodes, based on the ID of each ultrawideband node and the signal arrival time difference between every two nodes. Based on the real-time position of the mobile ultrawideband node and the signal reception strength of the direct path when each node is linked to different positions, radio tomography is performed, and the robot motion trajectory planning is updated based on the imaging results.

[0008] Furthermore, the robot is a quadruped robot, the lidar is located at the front end of the quadruped robot, and the quadruped robot is also equipped with a first communication module; the stationary node device includes a fixed bracket, and the second communication module and the stationary ultra-wideband node are mounted on the fixed bracket.

[0009] This invention also proposes a post-disaster search and rescue detection method based on the above-mentioned post-disaster search and rescue detection system, comprising the following steps:

[0010] (1) Based on the initial position of the robot cluster, with the goal of spreading the node links of the ultra-wideband wireless sensor network throughout the entire detection area grid, plan the global motion path of each robot.

[0011] (2) Based on reinforcement learning, the local path of the robot is planned to realize the robot's dynamic obstacle avoidance. If it is determined that the global motion path of a certain robot is completely blocked according to the environmental map and obstacle recognition data, the global motion path of each robot is planned again. The reinforcement learning technology adopts the imitation learning algorithm based on the generative adversarial model.

[0012] (3) Based on the real-time location of the mobile ultra-wideband node and the signal reception strength of the direct path when each node is linked to different locations, radio tomography is performed on the detection area.

[0013] Each ultrawideband node transmits pulse signals to all other nodes. The direct path signal reception strength data between each node link is obtained by processing the node's signal processing module. The signal processing module obtains the energy at the first peak of the ultrawideband pulse reception signal by sampling in the time domain as the direct path signal reception strength between two ultrawideband nodes.

[0014] The radio tomography imaging described herein images the spatial loss field of wireless signals in the detection area by acquiring the received signal strength of multiple direct paths linked to different node locations. Preferably, the radio tomography imaging employs a deep learning method based on a deep sequence learning generative adversarial model (Transformer-GAN). The imaging process of the deep sequence learning generative adversarial model Transformer-GAN includes: using a Transformer deep learning model to feature-encode the sequence of [direct path received signal strength, UWB wireless sensor network node locations], and performing end-to-end tomography imaging based on the generator of the generative adversarial deep learning model GAN.

[0015] (4) Based on the results of the radio tomography, estimate the probability of the victims appearing in the regional grid, take the area with a high probability of the victims appearing as the area to be detected, and further plan the global running trajectory of the robot cluster to increase the density of ultra-wideband node links passing through the area to be detected.

[0016] (5) Repeat steps (2) to (4) to perform detailed radio tomography on the area to be detected and estimate the probability of the victims appearing in the area grid again.

[0017] This invention proposes a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps in the post-disaster search and rescue detection method.

[0018] The present invention proposes a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps in the post-disaster search and rescue detection method.

[0019] Beneficial Effects: Compared with existing technologies, this invention has the following advantages: The post-disaster search and rescue detection method and device based on radio tomography has advantages such as convenient deployment, easy installation and maintenance, high precision and efficiency, and applicability to invisible environments such as dense smoke and collapsed buildings. By reasonably configuring the number of quadruped robot clusters and the scale of the ultra-wideband wireless sensor network, this method can be applied to both small-scale and large-scale search and rescue detection. The direct path signal reception strength acquisition method described in this invention can avoid the high noise problem caused by multipath signal transmission, effectively improving the accuracy of tomography; the post-disaster search and rescue detection system and method first performs radio tomography on the entire detection area, then estimates the potential location of victims based on the imaging results, and then performs fine imaging on this area as the detection area, improving detection accuracy. This scheme can obtain the existence and location of victims in the detection area (including invisible environments) in a short time, providing a foundation and guarantee for search and rescue personnel to carry out rescue work in a timely manner. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the post-disaster search and rescue detection system based on radio tomography as described in this invention;

[0021] Figure 2 This is a schematic diagram of the static node device described in this invention;

[0022] Figure 3 This is a schematic diagram of the robot described in this invention and its mounted lidar and mobile ultra-wideband node. Detailed Implementation

[0023] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0024] The post-disaster search and rescue detection system based on radio tomography described in this invention is illustrated in the following schematic diagram. Figure 1 As shown, the system includes an ultra-wideband wireless sensor network, a robot swarm, and a swarm control and tomographic imaging module. The ultra-wideband wireless sensor network is divided into stationary ultra-wideband nodes 1-2 and mobile ultra-wideband nodes 2-2. The stationary nodes are used to assist the mobile nodes in localization, and all nodes are used for radio tomography. In the ultra-wideband wireless sensor network, every two nodes can form a link. Since wireless signals have different signal attenuation rates when passing through different media, the same link will have different signal propagation paths after the nodes move. This can be determined by the positions of the two nodes. Therefore, in the detection area, by acquiring the signal reception strength of multiple links at different node positions, the spatial loss field of wireless signals in the detection area can be imaged, thereby enabling the detection of victims.

[0025] Each ultra-wideband (UWB) node is equipped with a signal processing module, designed using an FPGA chip, to process the received signals. A preferred processing scheme is to directly sample in the time domain and obtain the energy at the first peak of the UWB pulse received signal as the direct path signal reception strength between two UWB nodes, thus avoiding the problem of high signal reception noise caused by multipath transmission. The signal processing module processes the received signals of the UWB nodes to obtain the direct path signal reception strength data between each node link, and transmits the real-time link network and the data to the cluster control and tomography module 3 in real time through the communication module.

[0026] In one embodiment, the robot swarm is a quadruped robot swarm. The quadruped robot swarm consists of multiple quadruped robots 2-1, each equipped with a mobile ultra-wideband node 2-2, a two-dimensional lidar 2-3, and a communication module 2-4, which can be a 5G communication module. The quadruped robots can move in complex environments, providing conditions for the effective movement of ultra-wideband nodes in post-disaster areas. Each quadruped robot 2-1 is configured with two mobile ultra-wideband nodes 2-2 at a certain distance to increase the density of effective node links in the ultra-wideband wireless sensor network, thereby improving imaging accuracy and resolution. The quadruped robot 2-1 is equipped with a two-dimensional lidar 2-3 for real-time environmental mapping and obstacle identification; the quadruped robot 2-1 is also equipped with a 5G communication module 2-4 for real-time wireless communication with the swarm control and tomographic imaging system 3. Additionally, the stationary ultra-wideband node 1-2 is mounted on a fixed support 1-1 and, together with the 5G communication module 1-3, constitutes a stationary node module 1.

[0027] The cluster control and tomography module 3 acquires the location of the mobile ultra-wideband node and the signal reception strength of the direct path between each node in real time, performs radio tomography, and updates the motion trajectory planning based on the environmental map and imaging results.

[0028] The system's workflow is as follows: A. Preparation phase: Initial deployment of stationary and mobile nodes of the ultra-wideband wireless sensor network in the area to be detected; B. Based on the initial position of the quadruped robot swarm, rationally plan the global motion path of each quadruped robot to ensure that the node links of the ultra-wideband wireless sensor network are sufficiently dense throughout the detection area; C. During the movement of the quadruped robots, the swarm control and tomography system 3 performs radio tomography on the detection area after acquiring sufficient data.

[0029] The post-disaster search and rescue detection method based on the above-mentioned post-disaster search and rescue detection system includes the following steps:

[0030] (1) Based on the initial position of the robot cluster, with the goal of spreading the node links of the ultra-wideband wireless sensor network throughout the entire detection area grid, plan the global motion path of each robot.

[0031] (2) Based on reinforcement learning, the local path of the robot is planned to realize the robot's dynamic obstacle avoidance. If the global motion path of a robot is completely blocked according to the environmental map and obstacle recognition data, the global motion path of each robot is planned again.

[0032] (3) Based on the real-time location of the mobile ultra-wideband node and the signal reception strength of the direct path when each node is linked to different locations, radio tomography is performed on the detection area.

[0033] (4) Based on the results of the radio tomography, estimate the probability of the victims appearing in the regional grid, take the area with a high probability of the victims appearing as the area to be detected, and further plan the global running trajectory of the robot cluster to increase the density of ultra-wideband node links passing through the area to be detected.

[0034] (5) Repeat steps (2) to (4) to perform detailed radio tomography on the area to be detected and estimate the probability of the victims appearing in the area grid again.

[0035] Radio tomography, in particular, images the spatial loss field of wireless signals in the detection area by acquiring the signal reception strength of multiple links at different node locations. Specifically, it uses the direct path signal reception strength signals acquired by the ultra-wideband wireless sensor network at all time steps during the movement of a quadruped robot to perform tomographic imaging of the detection environment. Preferably, radio tomography employs the deep sequence learning model Transformer-GAN for imaging. While Transformer-GAN is an existing learning model, this invention adjusts its parameters according to the specific data dimensionality and the size of the detection environment without modifying the algorithm structure. The imaging process of the Transformer-GAN model includes: using the Transformer deep learning model to feature-encode the sequence of [direct path signal reception strength, ultra-wideband wireless sensor network node locations], and performing end-to-end tomographic imaging based on the generator of the generative adversarial learning model GAN. In this scheme, radio tomography was performed in both steps 3 and 5. Step 5 involved the robot cluster approaching the area where the victim was likely to be found, and then obtaining more dense node link signals passing through the area to perform fine imaging of the location. The same radio tomography method as in step 3 was used. Here, fine imaging means that it is more refined and has higher imaging accuracy than the first imaging in step 3.

[0036] To improve computational efficiency and rapidly shorten the search time for rescue targets, this invention also proposes a preferred solution:

[0037] (1) Initial motion planning for the robot swarm is performed based on reinforcement learning technology and sent to each robot for action. During the action, radio tomography is performed based on the signal reception intensity data of the direct path between each node link of the stationary ultrawideband node and the mobile ultrawideband node A. Reinforcement learning technology is used to plan the local path of the quadruped robot to achieve dynamic obstacle avoidance, such as obstacles like collapsed rubble. If the global motion path of the quadruped robot is observed to be completely blocked, the global motion path of each quadruped robot is replanned.

[0038] (2) Based on the results of radio tomography, estimate the probability of the victims appearing in the regional grid, take the area with a high probability of the victims appearing as the destination, obtain the global motion path plan of each robot, and send the global motion path plan to each robot.

[0039] (3) After the robot has acted according to the global motion path planning for a period of time, it performs fine radio tomography based on the signal reception strength data of the direct path between each node link of the stationary ultra-wideband node, the mobile ultra-wideband node A and the mobile ultra-wideband node B; based on the results of the fine radio tomography, it estimates the probability of the victims appearing in the regional grid again.

[0040] The above scheme allows for flexible adjustment of the number of sensors based on the detection results, which can improve computational efficiency and rapidly shorten the search time for the target.

[0041] In one embodiment, the reinforcement learning technique employs an imitation learning algorithm based on a generative adversarial model. This imitation learning algorithm is an existing algorithm that adjusts model parameters according to data and the environment; the present invention does not modify the structure of this algorithm.

[0042] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0043] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0044] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0045] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0046] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement a post-disaster search and rescue detection method.

[0047] In one embodiment, a computer-readable storage medium is provided on which a computer program is stored, which, when executed by a processor, implements a post-disaster search and rescue detection method.

[0048] The working principle of this scheme is as follows: Wireless signals have different signal attenuation rates when passing through different media. By deploying a dynamic wireless sensor network in post-disaster areas (especially in visual environments) and acquiring the signal reception strength of each pair of nodes at different locations, when the node links in the detection area are sufficiently dense, the spatial signal loss field of the area can be imaged using a large amount of signal reception strength, thereby achieving environmental detection and victim identification. In visual environments such as dense smoke and collapsed buildings, ultra-wideband signals have better penetration and can acquire direct path signal reception strength. They are insensitive to multipath and non-line-of-sight interference, have strong anti-interference capabilities, and can improve detection accuracy. In the area to be detected, stationary ultra-wideband nodes and mobile ultra-wideband nodes are deployed. The stationary nodes can be used for the localization of the mobile nodes. All nodes can collect signal reception strength for radio tomography. The mobile nodes are mounted on quadruped robots, which can autonomously map and navigate, moving in complex environments to collect sufficient signals for environmental imaging. During radio tomography, imaging is first performed to estimate the possible locations of victims. By further acquiring the denser node link signals passing through the area, detailed imaging of that location is performed to improve detection accuracy.

Claims

1. A disaster search and rescue detection system based on radio tomography, characterized in that: This includes ultra-wideband wireless sensor networks, robot swarms, and swarm control and tomographic imaging systems; The ultra-wideband wireless sensor network includes stationary ultra-wideband nodes and mobile ultra-wideband nodes. The stationary ultra-wideband nodes are installed on stationary node devices (1) at fixed locations to assist the positioning of mobile nodes. The mobile ultra-wideband nodes are mounted on robots (2). Each ultra-wideband node emits a pulse signal, which is received by other nodes. Each ultra-wideband node is equipped with a signal processing module to process the received signals of the ultra-wideband node, obtain the signal reception strength data of the direct path between each node link, and send it to the cluster control and tomography system (3) in real time through the communication module. The robot cluster includes several robots (2), each equipped with a lidar (2-3) and two mobile ultra-wideband nodes A and B (2-2) with a fixed distance between them; the lidar (2-3) is used by the robot to build an environmental map and identify obstacles in real time, and sends the data to the cluster control and tomography system (3) in real time through the communication module. The cluster control and tomography system (3) calculates the position of the moving ultrawideband node in real time with the assistance of the stationary node, based on the ID of each ultrawideband node and the signal arrival time difference between each two nodes. Based on the real-time location of the mobile ultra-wideband node and the signal reception strength of the direct path when each node is linked to different locations, radio tomography is performed, and the robot's motion trajectory planning is updated based on the imaging results.

2. The post-disaster search and rescue detection system based on radio tomography according to claim 1, characterized in that: The robot is a quadruped robot (2-1), the lidar (2-3) is located at the front end of the quadruped robot (2-1), and the quadruped robot (2-1) is also equipped with a first communication module (2-4); the stationary node device (1) includes a fixed bracket (1-1), the second communication module (1-3) and the stationary ultra-wideband node (1-2) are installed on the fixed bracket (1-1).

3. A post-disaster search and rescue detection method applied to the post-disaster search and rescue detection system of claim 1, characterized in that, Includes the following steps: (1) Based on the initial position of the robot cluster, with the goal of the node links of the ultra-wideband wireless sensor network covering the entire detection area grid, plan the global motion path of each robot. (2) Based on reinforcement learning, the local path of the robot is planned to realize the robot's dynamic obstacle avoidance. If the global motion path of a robot is completely blocked according to the environmental map and obstacle recognition data, the global motion path of each robot is planned again. (3) Based on the real-time location of the mobile ultra-wideband node and the signal reception strength of the direct path when each node is linked to different locations, radio tomography is performed on the detection area. (4) Based on the results of the radio tomography, estimate the probability of the victims appearing in the regional grid, take the area with a high probability of the victims appearing as the area to be detected, and further plan the global running trajectory of the robot cluster to increase the density of ultra-wideband node links passing through the area to be detected. (5) Repeat steps (2) to (4) to perform fine radio tomography on the area to be detected and estimate the probability of the victims appearing in the area grid again.

4. The post-disaster search and rescue detection method according to claim 3, characterized in that: The aforementioned radio tomography imaging images the spatial loss field of wireless signals in the detection area by acquiring the received signal strength of multiple direct paths linked at different node locations.

5. The post-disaster search and rescue detection method according to claim 4, characterized in that: The radio tomography employs a deep learning method based on the deep sequence learning generative adversarial model Transformer-GAN for imaging.

6. The post-disaster search and rescue detection method according to claim 5, characterized in that: The imaging process of the deep sequence learning generative adversarial model Transformer-GAN includes: using the Transformer deep learning model to encode the features of the [x, y] sequence, where x represents the signal reception strength of the direct path and y represents the location of the ultra-wideband wireless sensor network node, and performing end-to-end tomographic imaging based on the generator of the generative adversarial deep learning model GAN.

7. The post-disaster search and rescue detection method according to claim 3, characterized in that: Each ultrawideband node transmits pulse signals to all other nodes; the direct path signal reception strength data between each node link is obtained by the node's signal processing module. The signal processing module obtains the energy at the first peak of the ultrawideband pulse reception signal by sampling in the time domain as the direct path signal reception strength between two ultrawideband nodes.

8. The post-disaster search and rescue detection method according to claim 3, characterized in that: The reinforcement learning method employs an imitation learning algorithm based on a generative adversarial model.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps in the post-disaster search and rescue detection method according to any one of claims 3 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps in the post-disaster search and rescue detection method according to any one of claims 3 to 8.

Images