Wireless emergency command broadcasting system and method based on artificial intelligence

Abstract

The invention discloses a wireless emergency command broadcasting system and method based on artificial intelligence, and the method comprises the following steps: S1, determining the role of a user in an emergency scene through the wireless identity authentication of terminal equipment; s2, acquiring the specific position of each terminal user through a wireless network in real time, and sensing the surrounding environment information of the user; s3, generating a personalized emergency instruction by using a model based on the combination of the graph attention network and the generative adversarial network; s4, a priority optimization algorithm based on hierarchical reinforcement learning is adopted, and the system dynamically adjusts the priority of the broadcast content; s5, a real-time feedback mechanism is established, and the system dynamically adjusts an emergency instruction and a broadcast strategy according to feedback information of a user through a wireless communication technology; and S6, the emergency instruction is pushed to the user terminal, and optimized pushing is carried out according to the characteristics of different users. According to the method, artificial intelligence and wireless communication technologies are fully utilized, personalized emergency instructions are dynamically generated, and the method has the advantages of accuracy, high efficiency and real-time adaptation.

Description

Technical Field

[0001] The present invention relates to the field of emergency management and wireless communication technology, and in particular to an artificial intelligence-based wireless emergency command and broadcast system and method. Background Art

[0002] In modern society, with the frequent occurrence of natural disasters, emergencies and public safety incidents, emergency management and command systems have become increasingly important in ensuring public safety, reducing losses and improving emergency response efficiency. Traditional emergency command systems mainly rely on manual command and broadcast communication technology. These systems usually use a single communication channel, such as radio broadcasting or wired telephone networks, to transmit emergency information to the public and rescuers. However, this single-mode emergency command system has exposed many limitations in practical applications, especially in the face of complex and changing emergency scenarios. These limitations lead to untimely and inaccurate information transmission, affecting the overall efficiency and effectiveness of emergency response.

[0003] First, the information transmission method in the traditional emergency command system is usually a "one-size-fits-all" broadcast mode. Whether it is the commander, rescue personnel or the general public, the information content received is basically the same. This method is difficult to meet the personalized information needs of different roles and scenarios in emergency situations. For example, commanders need to obtain global information to formulate response strategies, rescue personnel need specific operating instructions and path guidance, and the general public needs clearer risk avoidance guidance and safety reminders. However, the broadcast information of the traditional system cannot be distinguished and customized according to the specific identity and role of the recipient, resulting in redundant or insufficient information, which affects the effectiveness of command decisions and rescue operations.

[0004] Secondly, traditional emergency systems have significant deficiencies in positioning and environmental perception capabilities. Typically, these systems rely mainly on a single positioning technology such as GPS. However, in complex emergency scenarios, such as densely built-up areas in cities, underground spaces, or environments with severe signal shielding, the positioning accuracy of GPS often cannot meet the requirements. Moreover, traditional systems lack the ability to perceive the user's surrounding environment in real time, making it difficult to dynamically acquire and process emergency-related environmental information, such as temperature, gas concentration, fire smoke, etc. This information is crucial for formulating and adjusting emergency instructions, but traditional systems find it difficult to update in real time and accurately perceive, resulting in severe limitations on the response speed and decision-making accuracy of emergency command.

[0005] In addition, traditional systems also have great limitations in the generation and push of emergency instructions. Generally speaking, the generation of instructions relies on pre-set fixed templates and lacks the ability to generate dynamically and adjust in real time. As emergency events progress and the environment changes, the content and priority of instructions often need to be adjusted frequently, but traditional systems lack intelligent processing capabilities based on real-time data and feedback information, making it difficult to dynamically adjust the content and priority of emergency instructions. This static way of generating emergency instructions obviously cannot provide sufficient flexibility and adaptability when facing complex and changing emergency scenarios.

[0006] More importantly, traditional emergency command systems lack effective feedback mechanisms and adaptive capabilities. Although some systems are able to receive user feedback, in most cases, this feedback information is not effectively utilized and a closed-loop emergency command process cannot be formed. In addition, during the push of emergency instructions, traditional systems are often unable to make adaptive adjustments based on the user's receiving capabilities, device characteristics, and real-time network conditions, resulting in a significant reduction in the timeliness and accuracy of information transmission. In the case of poor network conditions or limited user device capabilities, the push effect of traditional systems cannot be guaranteed, affecting the effective communication of emergency information.

[0007] Therefore, how to provide a wireless emergency command and broadcast system and method based on artificial intelligence is a problem that technical personnel in this field urgently need to solve. Summary of the invention

[0008] One purpose of the present invention is to propose a wireless emergency command broadcasting system and method based on artificial intelligence. The present invention makes full use of the combination of artificial intelligence technology and wireless communication technology to dynamically generate and push personalized emergency instructions, and make intelligent decisions based on user identity, location and real-time environmental information. At the same time, through hierarchical reinforcement learning algorithms and multimodal communication technology, the system can adjust the priority and push mode of broadcast content in real time to ensure the accuracy and timeliness of information transmission and adapt to complex and changeable emergency scenarios. The system significantly improves the response speed and decision-making support capabilities of emergency command, and has the advantages of accuracy, efficiency, real-time adaptability, flexibility and reliability.

[0009] According to an embodiment of the present invention, a wireless emergency command and broadcast method based on artificial intelligence includes the following steps:

[0010] S1. Automatically identify the user through wireless identity authentication of the terminal device and determine the user's role in the emergency scenario, including rescue personnel, commanders and the general public;

[0011] S2, combining the global positioning system, Wi-Fi signal positioning and IoT sensors, to obtain the specific location of each terminal user through the wireless network in real time and perceive the user's surrounding environment information;

[0012] S3. Based on the user's identity, role, location and real-time context, a model based on a combination of graph attention network and generative adversarial network is used to generate personalized emergency instructions. Commanders receive overall command strategies and decision support information through wireless communication networks, rescuers receive detailed operation instructions and path guidance, and the general public receives avoidance routes and safety reminders.

[0013] S4. Using a priority optimization algorithm based on hierarchical reinforcement learning, the system dynamically adjusts the priority of broadcast content according to the progress of emergency events and real-time feedback from users. The priority optimization algorithm includes high-level strategies and low-level strategies.

[0014] The high-level strategy analyzes the global information of the event through a decision tree model;

[0015] The low-level strategy adopts a strategy based on temporal difference learning to specifically optimize the priority of emergency instructions in specific scenarios;

[0016] S5. Establish a real-time feedback mechanism. The system dynamically adjusts emergency instructions and broadcast strategies based on user feedback information through wireless communication technology;

[0017] S6. Through multimodal wireless communication technology, emergency instructions are pushed to user terminals based on user identity, role, real-time location and environmental perception data, and the push is optimized according to the characteristics of different users.

[0018] Optionally, the S2 specifically includes:

[0019] S21. Use the global positioning system module to communicate with satellites via wireless networks in a two-way manner to obtain the longitude and latitude coordinates of each terminal user, and locate the user's current location through a global positioning system receiver, with a positioning accuracy of less than 5 meters;

[0020] S22, using Wi-Fi signal positioning technology, scanning the signal strength of wireless access points around the user, inputting received signal strength indicator values ​​of multiple Wi-Fi access point signals into different kernel functions to generate a feature vector;

[0021] S23, performing regression analysis on the generated feature vectors through an improved multiple kernel function support vector regression model to generate a position estimation value;

[0022] S24, adaptively weighting the position estimation value to generate a final position estimation result:

[0023]

[0024] Among them, Pos represents the final calculated relative position of the user, SVR krepresents the support regression model of the kth kernel function, λ k represents the contribution weight of the k-th model to the final position estimation, RSSI represents the received signal strength indicator value, and m represents the number of kernel functions used;

[0025] S25. Combined with IoT sensor data, the environmental parameters around the user are obtained in real time through the sensor network, and the environmental parameters are transmitted to the system through a low-power wireless communication protocol. The sampling frequency of the environmental parameters is 10 times per second.

[0026] S26, the system receives and analyzes the environmental parameters around the user, and generates a multi-dimensional geographic information model of the user;

[0027] S27. When there is a deviation between the user's location and the positioning results of the global positioning system and the Wi-Fi signal, the system automatically triggers the multi-source information fusion algorithm based on Bayesian reasoning to perform location correction:

[0028]

[0029] Among them, Corrected_Pos represents the corrected user position, GPS_Pos represents the position obtained by global positioning system positioning, WiFi_Pos represents the position obtained by Wi-Fi signal positioning, σ GPS represents the uncertainty of GPS positioning, σ WiFi Indicates the uncertainty of Wi-Fi positioning.

[0030] Optionally, the S3 specifically includes:

[0031] S31. Based on user identity and role information, users are divided into three categories: commanders, rescuers, and the general public. Different personalized emergency command generation models are built for each type of user. The commander model focuses on global decision-making information, the rescuer model focuses on operation instructions and path guidance, and the general public model focuses on avoidance routes and safety reminders.

[0032] S32. Use the graph attention network to model the user's identity, role, location and environmental information, regard the user and the surrounding environment as nodes in the graph, and describe the correlation between the user and the environmental information through the edges between the nodes, and calculate the importance weight α of each node ij :

[0033]

[0034] Among them, ReLU represents the rectified linear unit activation function, represents the attention weight vector, W h represents the feature map weight matrix, h irepresents the feature vector of node i, h k represents the feature vector of node k, h j represents the feature vector of node j, || represents the connection operation of feature vectors, Represents the set of neighbor nodes of node i;

[0035] S33. The node output by the graph attention network is represented as h i The user's location information i and environmental information i Combine to form an extended feature vector x i =[h i ||p i ||e i ], input into the generator model;

[0036] S34, generator model generates personalized emergency instruction set C based on conditional generative adversarial network gen :

[0037] C gen =W g tanh(x i ·W cond +b cond );

[0038] Among them, x i represents the extended feature vector, W cond represents the weight matrix of the conditional generative adversarial network, b cond represents the bias vector, tanh represents the hyperbolic tangent activation function, W g represents the output weight matrix of the generator;

[0039] S35, the discriminator receives the generated personalized emergency instruction set C gen and actual scene feedback data C real , and calculate the effectiveness score S of the instruction cmd :

[0040] S cmd =σ(W d ·[C gen ||C real ]);

[0041] Among them, W d represents the weight matrix of the discriminator, C real represents the command feedback data received in the actual emergency scenario, and σ represents the activation function;

[0042] S36, the system updates the node representation h in the graph model in real time during each round of generation and evaluation. i and the importance weight α of each node ij, ensuring the generated personalized emergency instruction set C gen It can improve the accuracy and real-time performance of instructions based on the latest environment and user information.

[0043] Optionally, the S4 specifically includes:

[0044] S41. When an emergency occurs, the system determines the global priority of the emergency through high-level strategies and makes decisions through a multi-dimensional decision tree model. global The calculation formula is:

[0045]

[0046] Among them, w i Indicates the weight coefficient of the i-th factor, Score i represents the score of the ith factor, Risk i represents the risk index of the i-th factor, λ p represents the risk adjustment factor, Latency i Indicates the network delay of the i-th factor, Bandwidth i represents the network bandwidth of the i-th factor, σ represents the activation function, and n represents the total number of evaluation factors;

[0047] S42, based on the calculated global priority Priority global ,The system selects an appropriate low-level strategy to optimize the dynamic priority of ,emergency commands. The low-level strategy adopts a ,temporal difference learning based strategy to refine the ,dynamic adjustment of command priorities;

[0048] S43. The instruction priority update formula in the low-level strategy is:

[0049]

[0050] Among them, Priority t+1 Indicates the update priority at time step t+1, Priority t represents the current priority at time step t, α represents the learning rate, and r t represents the immediate reward of the current time step, γ represents the discount factor, Context t represents the context information of the current time step t, and φ represents the context adjustment coefficient;

[0051] S44, the system based on the user's real-time feedback information F user , dynamically adjust the immediate reward r in the low-level strategy t , calculate the updated immediate reward r t+1 :

[0052]

[0053] Among them, β represents the feedback correction rate, Score feedback represents the user feedback score, ζ represents the response time adjustment coefficient, ResponseTime user Indicates user response time, ErrorRate network Indicates the network error rate;

[0054] S45. After each priority update, the system calculates the current priority of different emergency instructions in real time. t+1 , and push the instructions to the corresponding users in order according to the priority level. The instructions with higher priority will be pushed first;

[0055] S46. During the push process, the system continuously monitors the network status and user feedback information, and re-evaluates the command priority and adjusts the push order when necessary to adapt to the dynamically changing emergency environment and ensure that the information can be delivered to the target user in the shortest time.

[0056] Optionally, the S5 specifically includes:

[0057] S51, the system collects user feedback information in real time through multimodal sensors and user terminal devices user , including user response time, operation results, success rate of instruction execution and user satisfaction score;

[0058] S52, collect the user feedback information F user Perform preprocessing, including data cleaning, denoising and standardization operations, to obtain standardized feedback data F norm ;

[0059] S53, the system will standardize the feedback data F norm And the user's historical feedback data F hist Fusion is performed to construct the fused feedback feature representation F fusion ;

[0060] S54, based on the fused feedback feature representation F fusion , the system dynamically adjusts the generation and push strategy of emergency instructions through reinforcement learning algorithm, and the objective function L of reinforcement learning is:

[0061]

[0062] Where θ represents the discount factor, Q t represents the reward function at time step t, S env Indicates the current environment status;

[0063] S55. After the command is pushed, the system continues to monitor new user feedback information F user , and update the fusion feedback feature representation F in real time fusion ,According to the real-time updated feedback information, the generation and push strategy of emergency instructions are optimized.,T represents the end time or termination condition of the entire decision-making process, and E represents the expected value.

[0064] Optionally, the S6 specifically includes:

[0065] S61. The system selects the most appropriate push strategy for different user types based on the user's identity, role, location, and environmental perception data. The decision function for the push strategy selection is:

[0066]

[0067] Among them, S strategy represents the selected push strategy, ρ i represents the weight coefficient of the i-th influencing factor, represents the impact evaluation value of the i-th influencing factor on the k-th strategy, and N represents the total number of influencing factors considered;

[0068] S62. According to the selected push strategy, the system integrates voice, text, image and video to generate the emergency instruction push content that is most suitable for the user to accept. The generation formula of the push content is:

[0069]

[0070] Among them, C content Indicates the generated push content, η j Represents the weighting coefficient of the jth modal information, M j represents the content representation of the jth modal information, and M represents the number of types of modal information;

[0071] S63. During the push process, the system adaptively selects the best push mode according to the current network conditions. The decision function for selecting the push mode is:

[0072]

[0073] Among them, M mode Indicates the selected push mode, Latency l Indicates the network delay of the lth push mode, ErrorRate l Indicates the network error rate of the lth push mode, Bandwidth l Indicates the bandwidth of the lth push mode;

[0074] S64. The system adaptively adjusts the format and size of the pushed content according to the user's receiving ability and device characteristics to ensure that the information can be effectively delivered in different devices and environments. The adjustment formula for personalized push content is:

[0075] C adapted =C content ·A device ;

[0076] Among them, C adapted represents the push content after adaptive optimization, C content Indicates the original push content, A device represents the user equipment adaptability matrix;

[0077] S65. The system continuously monitors and optimizes the push status during the push process;

[0078] S66. Based on real-time monitoring and user feedback, the system dynamically adjusts push strategies and modes to ensure that emergency instructions can be efficiently delivered under different network conditions and user devices:

[0079]

[0080] Among them, S new Indicates the new push strategy, M new Indicates the new push mode, Satisfaction k,l represents the user's satisfaction under strategy k and mode l, Cost k,l represents the push cost of the user under strategy k and mode l, Benefit k,l Indicates the push effect of the user under strategy k and mode l.

[0081] A wireless emergency command and broadcast system based on artificial intelligence according to an embodiment of the present invention includes the following modules:

[0082] Identity recognition module: used to automatically identify the user through wireless identity authentication and determine the user's role in the emergency scenario;

[0083] Positioning and environmental perception module: used to combine the global positioning system, Wi-Fi signal positioning and IoT sensors to obtain user location and perceive environmental information in real time;

[0084] Emergency instruction generation module: Based on the user's identity, role, location and real-time context, it uses graph attention network and generative adversarial network to generate personalized emergency instructions;

[0085] Priority adjustment module: uses a hierarchical reinforcement learning algorithm to dynamically adjust the priority of emergency instructions based on the progress of emergency events and user feedback;

[0086] Feedback mechanism module: used to collect user feedback information in real time and optimize the generation and push strategy of emergency instructions through reinforcement learning algorithms;

[0087] Multimodal push module: Based on user characteristics and network conditions, emergency instructions are pushed to user terminals through multimodal communication technology, and optimized according to user receiving capabilities and device characteristics.

[0088] The beneficial effects of the present invention are:

[0089] First, the present invention uses wireless identity authentication technology to accurately identify the identity and role of users. Whether it is a commander, rescuer or the general public, the system can generate personalized emergency instructions based on their specific identity and role. This differentiated instruction generation method not only avoids the problems of information redundancy and misunderstanding in traditional systems, but also ensures that each user can receive the instructions most relevant to their emergency role, thereby greatly improving the accuracy and effectiveness of information transmission.

[0090] Secondly, the innovative design of the present invention in positioning and environmental perception enables it to provide high-precision positioning services in complex emergency scenarios and perceive environmental changes around users in real time. By combining the global positioning system (GPS), Wi-Fi signal positioning and IoT sensors, the system can accurately locate users in different indoor and outdoor environments and obtain environmental data including temperature, gas concentration, etc. These data not only provide a reliable basis for the generation and adjustment of emergency instructions, but also enable the system to dynamically update the user's geographic information model during the progress of the emergency event, thereby more accurately guiding emergency actions.

[0091] In terms of emergency command generation, the present invention adopts a model based on the combination of graph attention network (GAT) and generative adversarial network (GAN), which can dynamically generate personalized emergency commands based on the user's identity, location and environmental context. Compared with the traditional fixed template command generation method, this method greatly improves the flexibility and adaptability of the command, allowing the emergency command to better cope with complex and changing scenario requirements.

[0092] In addition, the priority adjustment mechanism of the present invention uses a hierarchical reinforcement learning algorithm to dynamically optimize the push of emergency instructions according to the progress of the event and the real-time feedback of users. Through the combination of high-level and low-level strategies, the system can ensure that the most important emergency information is delivered first, and always maintain the timeliness and effectiveness of information delivery at different stages of the emergency event. The system's real-time feedback mechanism further enhances the closed-loop management of the command process, and can continuously optimize the generation and push strategies of emergency instructions based on user feedback information, thereby ensuring the flexibility and adaptability of emergency response.

[0093] Finally, the multimodal communication technology used in the present invention enables it to adaptively select the best push mode and content format under different network conditions and device environments. This not only ensures the efficient delivery of emergency information, but also improves the overall communication capability of the system in complex scenarios. Whether in an urban environment with good network conditions or in remote areas with insufficient network coverage, the system can flexibly adjust the push strategy to ensure that information can be delivered to the target users in a timely and accurate manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0094] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0095] Figure 1 A flowchart of a wireless emergency command and broadcast method based on artificial intelligence proposed by the present invention;

[0096] Figure 2 This is a structural schematic diagram of an artificial intelligence-based wireless emergency command and broadcast system proposed by the present invention. DETAILED DESCRIPTION

[0097] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0098] refer to Figure 1 and Figure 2 , a wireless emergency command and broadcast method based on artificial intelligence, comprising the following steps:

[0099] S1. Automatically identify the user through wireless identity authentication of the terminal device and determine the user's role in the emergency scenario, including rescue personnel, commanders and the general public;

[0100] S2, combining the global positioning system, Wi-Fi signal positioning and IoT sensors, to obtain the specific location of each terminal user through the wireless network in real time and perceive the user's surrounding environment information;

[0101] S3. Based on the user's identity, role, location and real-time context, a model based on a combination of graph attention network and generative adversarial network is used to generate personalized emergency instructions. Commanders receive overall command strategies and decision support information through wireless communication networks, rescuers receive detailed operation instructions and path guidance, and the general public receives avoidance routes and safety reminders.

[0102] S4. Using a priority optimization algorithm based on hierarchical reinforcement learning, the system dynamically adjusts the priority of broadcast content according to the progress of emergency events and real-time feedback from users. The priority optimization algorithm includes high-level strategies and low-level strategies.

[0103] The high-level strategy analyzes the global information of the event through a decision tree model;

[0104] The low-level strategy adopts a strategy based on temporal difference learning to specifically optimize the priority of emergency instructions in specific scenarios;

[0105] S5. Establish a real-time feedback mechanism. The system dynamically adjusts emergency instructions and broadcast strategies based on user feedback information through wireless communication technology;

[0106] S6. Through multimodal wireless communication technology, emergency instructions are pushed to user terminals based on user identity, role, real-time location and environmental perception data, and the push is optimized according to the characteristics of different users.

[0107] In this implementation, S2 specifically includes:

[0108] S21. Use the global positioning system module to communicate with satellites via wireless networks in a two-way manner to obtain the longitude and latitude coordinates of each terminal user, and locate the user's current location through a global positioning system receiver, with a positioning accuracy of less than 5 meters;

[0109] S22, using Wi-Fi signal positioning technology, scanning the signal strength of wireless access points around the user, inputting received signal strength indicator values ​​of multiple Wi-Fi access point signals into different kernel functions to generate a feature vector;

[0110] S23, performing regression analysis on the generated feature vectors through an improved multiple kernel function support vector regression model to generate a position estimation value;

[0111] S24, adaptively weighting the position estimation value to generate a final position estimation result:

[0112]

[0113] Among them, Pos represents the final calculated relative position of the user, SVR k represents the support regression model of the kth kernel function, λ k represents the contribution weight of the k-th model to the final position estimation, RSSI represents the received signal strength indicator value, and m represents the number of kernel functions used;

[0114] S25. Combined with IoT sensor data, the environmental parameters around the user are obtained in real time through the sensor network, and the environmental parameters are transmitted to the system through a low-power wireless communication protocol. The sampling frequency of the environmental parameters is 10 times per second.

[0115] S26, the system receives and analyzes the environmental parameters around the user, and generates a multi-dimensional geographic information model of the user;

[0116] S27. When there is a deviation between the user's location and the positioning results of the global positioning system and the Wi-Fi signal, the system automatically triggers the multi-source information fusion algorithm based on Bayesian reasoning to perform location correction:

[0117]

[0118] Among them, Corrected_Pos represents the corrected user position, GPS_Pos represents the position obtained by global positioning system positioning, WiFi_Pos represents the position obtained by Wi-Fi signal positioning, σ GPS represents the uncertainty of GPS positioning, σ WiFi Indicates the uncertainty of Wi-Fi positioning.

[0119] In this implementation, S3 specifically includes:

[0120] S31. Based on user identity and role information, users are divided into three categories: commanders, rescuers, and the general public. Different personalized emergency command generation models are built for each type of user. The commander model focuses on global decision-making information, the rescuer model focuses on operation instructions and path guidance, and the general public model focuses on avoidance routes and safety reminders.

[0121] S32. Use the graph attention network to model the user's identity, role, location and environmental information, regard the user and the surrounding environment as nodes in the graph, and describe the correlation between the user and the environmental information through the edges between the nodes, and calculate the importance weight α of each node ij :

[0122]

[0123] Among them, ReLU represents the rectified linear unit activation function, represents the attention weight vector, W h represents the feature map weight matrix, h i represents the feature vector of node i, h k represents the feature vector of node k, h j represents the feature vector of node j, || represents the connection operation of feature vectors, Represents the set of neighbor nodes of node i;

[0124] S33. The node output by the graph attention network is represented as h i The user's location information i and environmental information i Combine to form an extended feature vector x i =[h i ||p i ||e i ], input into the generator model;

[0125] S34, generator model generates personalized emergency instruction set C based on conditional generative adversarial network gen :

[0126] C gen =W g tanh(x i ·W cond +b cond );

[0127] Among them, x i represents the extended feature vector, W cond represents the weight matrix of the conditional generative adversarial network, b cond represents the bias vector, tanh represents the hyperbolic tangent activation function, W g represents the output weight matrix of the generator;

[0128] S35, the discriminator receives the generated personalized emergency instruction set C gen and actual scene feedback data C real , and calculate the effectiveness score S of the instruction cmd :

[0129] S cmd =σ(W d ·[C gen ||C real ]);

[0130] Among them, W d represents the weight matrix of the discriminator, C real represents the command feedback data received in the actual emergency scenario, and σ represents the activation function;

[0131] S36, the system updates the node representation h in the graph model in real time during each round of generation and evaluation. i and the importance weight α of each node ij , ensuring the generated personalized emergency instruction set C gen It can improve the accuracy and real-time performance of instructions based on the latest environment and user information.

[0132] In this implementation manner, the S4 specifically includes:

[0133] S41. When an emergency occurs, the system determines the global priority of the emergency through high-level strategies and makes decisions through a multi-dimensional decision tree model. global The calculation formula is:

[0134]

[0135] Among them, w i Indicates the weight coefficient of the i-th factor, Score i represents the score of the ith factor, Risk i represents the risk index of the i-th factor, λ p represents the risk adjustment factor, Latency i Indicates the network delay of the i-th factor, Bandwidth i represents the network bandwidth of the i-th factor, σ represents the activation function, and n represents the total number of evaluation factors;

[0136] S42, based on the calculated global priority Priority global ,The system selects an appropriate low-level strategy to optimize the dynamic priority of ,emergency commands. The low-level strategy adopts a ,temporal difference learning based strategy to refine the ,dynamic adjustment of command priorities;

[0137] S43. The instruction priority update formula in the low-level strategy is:

[0138]

[0139] Among them, Priority t+1 Indicates the update priority at time step t+1, Priority t represents the current priority at time step t, α represents the learning rate, and r t represents the immediate reward of the current time step, γ represents the discount factor, Context t represents the context information of the current time step t, and φ represents the context adjustment coefficient;

[0140] S44, the system based on the user's real-time feedback information F user , dynamically adjust the immediate reward r in the low-level strategy t , calculate the updated immediate reward r t+1 :

[0141]

[0142] Among them, β represents the feedback correction rate, Score feedbackrepresents the user feedback score, ζ represents the response time adjustment coefficient, ResponseTime user Indicates user response time, ErrorRate network Indicates the network error rate;

[0143] S45. After each priority update, the system calculates the current priority of different emergency instructions in real time. t+1 , and push the instructions to the corresponding users in order according to the priority level. The instructions with higher priority will be pushed first;

[0144] S46. During the push process, the system continuously monitors the network status and user feedback information, and re-evaluates the command priority and adjusts the push order when necessary to adapt to the dynamically changing emergency environment and ensure that the information can be delivered to the target user in the shortest time.

[0145] In this implementation manner, S5 specifically includes:

[0146] S51, the system collects user feedback information in real time through multimodal sensors and user terminal devices user , including user response time, operation results, success rate of instruction execution and user satisfaction score;

[0147] S52, collect the user feedback information F user Perform preprocessing, including data cleaning, denoising and standardization operations, to obtain standardized feedback data F norm ;

[0148] S53, the system will standardize the feedback data F norm And the user's historical feedback data F hist Fusion is performed to construct the fused feedback feature representation F fusion ;

[0149] S54, based on the fused feedback feature representation F fusion , the system dynamically adjusts the generation and push strategy of emergency instructions through reinforcement learning algorithm, and the objective function L of reinforcement learning is:

[0150]

[0151] Where θ represents the discount factor, Q t represents the reward function at time step t, S env Indicates the current environment status;

[0152] S55. After the command is pushed, the system continues to monitor new user feedback information F user , and update the fusion feedback feature representation F in real time fusion,According to the real-time updated feedback information, the generation and push strategy of emergency instructions are optimized.,T represents the end time or termination condition of the entire decision-making process, and E represents the expected value.

[0153] In this implementation manner, S6 specifically includes:

[0154] S61. The system selects the most appropriate push strategy for different user types based on the user's identity, role, location, and environmental perception data. The decision function for the push strategy selection is:

[0155]

[0156] Among them, S strategy represents the selected push strategy, ρ i represents the weight coefficient of the i-th influencing factor, represents the impact evaluation value of the i-th influencing factor on the k-th strategy, and N represents the total number of influencing factors considered;

[0157] S62. According to the selected push strategy, the system integrates voice, text, image and video to generate the emergency instruction push content that is most suitable for the user to accept. The generation formula of the push content is:

[0158]

[0159] Among them, C content Indicates the generated push content, η j Represents the weighting coefficient of the jth modal information, M j represents the content representation of the jth modal information, and M represents the number of types of modal information;

[0160] S63. During the push process, the system adaptively selects the best push mode according to the current network conditions. The decision function for selecting the push mode is:

[0161]

[0162] Among them, M mode Indicates the selected push mode, Latency l Indicates the network delay of the lth push mode, ErrorRate l Indicates the network error rate of the lth push mode, Bandwidth l Indicates the bandwidth of the lth push mode;

[0163] S64. The system adaptively adjusts the format and size of the pushed content according to the user's receiving ability and device characteristics to ensure that the information can be effectively delivered in different devices and environments. The adjustment formula for personalized push content is:

[0164] C adapted =C content ·A device ;

[0165] Among them, C adapted represents the push content after adaptive optimization, C content Indicates the original push content, A device represents the user equipment adaptability matrix;

[0166] S65. The system continuously monitors and optimizes the push status during the push process;

[0167] S66. Based on real-time monitoring and user feedback, the system dynamically adjusts push strategies and modes to ensure that emergency instructions can be efficiently delivered under different network conditions and user devices:

[0168]

[0169] Among them, S new Indicates the new push strategy, M new Indicates the new push mode, Satisfaction k,l represents the user's satisfaction under strategy k and mode l, Cost k,l represents the push cost of the user under strategy k and mode l, Benefit k,l Indicates the push effect of the user under strategy k and mode l.

[0170] At the same time, the present invention proposes a wireless emergency command and broadcast system based on artificial intelligence, including the following modules:

[0171] Identity recognition module: used to automatically identify the user through wireless identity authentication and determine the user's role in the emergency scenario;

[0172] Positioning and environmental perception module: used to combine the global positioning system, Wi-Fi signal positioning and IoT sensors to obtain user location and perceive environmental information in real time;

[0173] Emergency instruction generation module: Based on the user's identity, role, location and real-time context, it uses graph attention network and generative adversarial network to generate personalized emergency instructions;

[0174] Priority adjustment module: uses a hierarchical reinforcement learning algorithm to dynamically adjust the priority of emergency instructions based on the progress of emergency events and user feedback;

[0175] Feedback mechanism module: used to collect user feedback information in real time and optimize the generation and push strategy of emergency instructions through reinforcement learning algorithms;

[0176] Multimodal push module: Based on user characteristics and network conditions, emergency instructions are pushed to user terminals through multimodal communication technology, and optimized according to user receiving capabilities and device characteristics.

[0177] Embodiment 1:

[0178] In order to verify the feasibility of the present invention in implementation, the present invention is applied to a coastal city, which faces disaster risks such as strong winds, rainstorms, and seawater backflow every summer due to frequent seasonal typhoons. To cope with this situation, the municipal government introduced an artificial intelligence-based wireless emergency command and broadcast system to provide accurate and efficient emergency instructions and information push to residents, rescuers, and government commanders before and after the typhoon.

[0179] Before the typhoon landed, the system was widely used in the city's emergency management and command. The entire city is divided into multiple emergency management areas, each of which includes different user groups, such as the general public, rescue workers, and commanders. The system performs wireless identity authentication through user terminal devices, automatically identifies user identities and roles, and pushes corresponding emergency instructions to each type of user according to the preset emergency response strategy.

[0180] The day before the typhoon arrived, the meteorological department provided the system with accurate typhoon paths and real-time meteorological data through the global positioning system and the IoT sensor network. The system first evaluated the areas most affected by the typhoon based on these data, and updated the geographic information model of users in each area in real time. Since the typhoon path involved multiple urban areas, the system used Wi-Fi signal positioning technology to accurately locate the locations of users in the urban areas, especially those inside buildings and underground spaces where GPS signals were limited. By comparing GPS and Wi-Fi positioning data, the system automatically triggered a multi-source information fusion algorithm to correct the location information of tens of thousands of residents, and improved the positioning accuracy to within 2 meters.

[0181] In residential areas, the system generates personalized emergency instructions based on the identity, role, and real-time location information of the residents. For residents close to the coastline, the system pushes detailed evacuation routes and shelter locations, and prompts them to evacuate immediately to designated safe areas through voice and text. For residents far from the coastline but still in high-risk areas, the system pushes specific guidance on risk avoidance advice and stockpiling supplies, and reminds them to be prepared for long-term power and water outages. In addition, the system also monitors residents' feedback in real time and finds that some residents are hesitant because they are unfamiliar with the location of the shelter. In response to this problem, the system quickly adjusted the push content, generated instructions with detailed map navigation for these residents, and pushed the information to their smartphones and home smart devices through multimodal communication technology.

[0182] In terms of rescue, the system identified the location and responsibilities of the rescuers through the identity recognition module, and pushed detailed rescue route maps and operation instructions. In the city center, due to severe traffic congestion, the system adjusted the rescue route in real time based on the traffic flow data collected by the IoT sensor, avoiding traffic congestion areas, and pushed the adjusted route instructions to the terminal devices of the rescue team in real time. During the rescue process, the system received feedback from the rescue team in real time through the wireless network, and optimized the priority and content of the instructions through the hierarchical reinforcement learning algorithm. For example, in a rescue in a large area of ​​waterlogged area, the system found that some rescuers could not reach the designated rescue site as planned due to road closures. The system then adjusted the rescue strategy based on the latest feedback data, focusing on areas with shallow water and passable, and notified other rescue team members to go to support. Through this dynamic adjustment, the rescuers completed a large-scale evacuation of personnel and material transportation tasks in a very short time, avoiding more casualties.

[0183] For commanders, the system provides global decision-making support information to help them better coordinate the city's emergency resources. The command center obtains the latest disaster situation, rescue progress and material distribution in various regions of the city through the emergency command generation module. This information is presented in real time on the large screen of the command center in the form of charts and data reports to assist commanders in making the best decision. With the assistance of the system, commanders can accurately determine which area needs priority rescue and which area needs further protection, and quickly convey instructions to relevant departments and rescue teams.

[0184] Within 24 hours after the typhoon made landfall, the system pushed more than 500,000 personalized emergency instructions, covering a city population of more than 2 million. Through the system's real-time feedback and dynamic adjustment mechanism, the vast majority of residents were able to evacuate in time or take effective risk avoidance measures before the typhoon arrived. Compared with the traditional emergency command method that relied solely on radio broadcasts, the average evacuation time for residents in this typhoon was shortened by 30%, the rescue response time was shortened by 25%, and the disaster losses in the entire city were reduced by about 20%.

[0185] Table 1 Data comparison between traditional emergency response system and emergency response system based on artificial intelligence

[0186]

[0187] From the data analysis in Table 1, we can see that during the typhoon, the AI-based wireless emergency command and broadcast system showed significant advantages over the traditional emergency system, which is reflected in several key indicators. First, in terms of evacuation time, the average evacuation time using the AI ​​system was 42 minutes, which is 30% shorter than the 60 minutes of the traditional system. This shows that the AI-based system can more quickly identify residents in high-risk areas and push evacuation instructions in a timely manner, thereby greatly improving the efficiency of emergency response.

[0188] The rescue response time has also been significantly shortened. The response time of the traditional system is 30 minutes, while the AI-based system shortens it to 20 minutes, a reduction of 25%. This improvement is mainly due to the system's dynamic processing and priority adjustment capabilities of real-time feedback information, which enables rescuers to quickly obtain and execute adjusted rescue instructions, avoiding time waste.

[0189] In terms of disaster losses, the AI-based system reduced the city's overall disaster losses by 20%. This achievement is closely related to the system's ability to accurately identify and relocate residents in high-risk areas before a disaster occurs. Data shows that based on the system, about 10,000 residents in high-risk areas were successfully relocated, avoiding potential casualties. In addition, the system performed well during the rescue process and successfully evacuated 5,000 trapped residents, further highlighting its practicality in disaster response.

[0190] The system also performed well in resource allocation and decision-making efficiency. Through intelligent decision support and dynamic optimization of emergency instructions, the efficiency of relief material distribution has increased by 15%, and the decision-making efficiency of the command center has also increased by 15%. These improvements ensure the maximum utilization of resources and make emergency operations more coordinated and efficient.

[0191] In terms of information push, the AI-based system pushed more than 500,000 personalized emergency instructions during the typhoon, covering more than 2 million people. The system's emergency information push success rate reached 98%, a 13% increase from the 85% of traditional systems. This high success rate is due to the system's multimodal communication technology and adaptability to network conditions. Even under poor network conditions, the system can still effectively transmit key information to ensure that every user can receive emergency instructions in a timely manner.

[0192] Finally, the AI-based system demonstrated strong real-time response capabilities in terms of feedback processing and dynamic adjustment. Compared with the lack of real-time performance of traditional systems, the AI ​​system can dynamically adjust push strategies and content based on real-time feedback to ensure the accuracy and timeliness of information transmission. This capability enables the system to respond flexibly to complex and changing disaster scenarios, further improving the overall efficiency of emergency command.

[0193] In summary, the wireless emergency command and broadcast system based on artificial intelligence is significantly superior to the traditional emergency system in many key indicators. It is not only more efficient in terms of time, but also performs well in resource utilization and accuracy of information transmission, providing strong technical support for responding to sudden disasters and greatly improving the overall effect of emergency management.

[0194] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A wireless emergency command and broadcast method based on artificial intelligence, characterized in that: The steps include: S1. Automatically identify the user through wireless identity authentication of the terminal device and determine the user's role in the emergency scenario, including rescue personnel, commanders and the general public; S2, combining the global positioning system, Wi-Fi signal positioning and IoT sensors, to obtain the specific location of each terminal user through the wireless network in real time and perceive the user's surrounding environment information; S3. Based on the user's identity, role, location and real-time context, a model based on a combination of graph attention network and generative adversarial network is used to generate personalized emergency instructions. Commanders receive overall command strategies and decision support information through wireless communication networks, rescuers receive detailed operation instructions and path guidance, and the general public receives avoidance routes and safety reminders. S4. Using a priority optimization algorithm based on hierarchical reinforcement learning, the system dynamically adjusts the priority of broadcast content according to the progress of emergency events and real-time feedback from users. The priority optimization algorithm includes high-level strategies and low-level strategies. The high-level strategy analyzes the global information of the event through a decision tree model; The low-level strategy adopts a strategy based on temporal difference learning to specifically optimize the priority of emergency instructions in specific scenarios; S5. Establish a real-time feedback mechanism. The system dynamically adjusts emergency instructions and broadcast strategies based on user feedback information through wireless communication technology; S6. Through multimodal wireless communication technology, emergency instructions are pushed to user terminals based on user identity, role, real-time location and environmental perception data, and the push is optimized according to the characteristics of different users.

2. According to the artificial intelligence-based wireless emergency command and broadcast method of claim 1, it is characterized in that: The S2 specifically includes: S21. Use the global positioning system module to communicate with satellites via wireless networks in a two-way manner to obtain the longitude and latitude coordinates of each terminal user, and locate the user's current location through a global positioning system receiver, with a positioning accuracy of less than 5 meters; S22, using Wi-Fi signal positioning technology, scanning the signal strength of wireless access points around the user, inputting received signal strength indicator values ​​of multiple Wi-Fi access point signals into different kernel functions to generate a feature vector; S23, performing regression analysis on the generated feature vectors through an improved multiple kernel function support vector regression model to generate a position estimation value; S24, adaptively weighting the position estimation value to generate a final position estimation result: Among them, Pos represents the final calculated relative position of the user, SVR k represents the support regression model of the kth kernel function, λ k represents the contribution weight of the k-th model to the final position estimation, RSSI represents the received signal strength indicator value, and m represents the number of kernel functions used; S25. Combined with IoT sensor data, the environmental parameters around the user are obtained in real time through the sensor network, and the environmental parameters are transmitted to the system through a low-power wireless communication protocol. The sampling frequency of the environmental parameters is 10 times per second. S26, the system receives and analyzes the environmental parameters around the user, and generates a multi-dimensional geographic information model of the user; S27. When there is a deviation between the user's location and the positioning results of the global positioning system and the Wi-Fi signal, the system automatically triggers the multi-source information fusion algorithm based on Bayesian reasoning to perform location correction: Among them, Corrected_Pos represents the corrected user position, GPS_Pos represents the position obtained by global positioning system positioning, WiFi_Pos represents the position obtained by Wi-Fi signal positioning, σ GPS represents the uncertainty of GPS positioning, σ WiFi Indicates the uncertainty of Wi-Fi positioning.

3. According to the artificial intelligence-based wireless emergency command and broadcast method of claim 1, it is characterized in that: The S3 specifically includes: S31. Based on user identity and role information, users are divided into three categories: commanders, rescuers, and the general public. Different personalized emergency command generation models are built for each type of user. The commander model focuses on global decision-making information, the rescuer model focuses on operation instructions and path guidance, and the general public model focuses on avoidance routes and safety reminders. S32. Use the graph attention network to model the user's identity, role, location and environmental information, regard the user and the surrounding environment as nodes in the graph, and describe the correlation between the user and the environmental information through the edges between the nodes, and calculate the importance weight α of each node ij : Among them, ReLU represents the rectified linear unit activation function, represents the attention weight vector, W h represents the feature map weight matrix, h i represents the feature vector of node i, h k represents the feature vector of node k, h j represents the feature vector of node j, ∥ represents the connection operation of feature vectors, Represents the set of neighbor nodes of node i; S33. The node output by the graph attention network is represented as h i The user's location information i and environmental information i Combine to form an extended feature vector x i =[h i ∥p i ∥e i ], input into the generator model; S34, generator model generates personalized emergency instruction set C based on conditional generative adversarial network gen : C gen =W g ·tanh(x i ·W cond +b cond ); Among them, x i represents the extended feature vector, W cond represents the weight matrix of the conditional generative adversarial network, b cond represents the bias vector, tanh represents the hyperbolic tangent activation function, W g represents the output weight matrix of the generator; S35, the discriminator receives the generated personalized emergency instruction set C gen and actual scene feedback data C real , and calculate the effectiveness score S of the instruction cmd : S cmd =σ(W d ·[C gen ∥C real ]); Among them, W d represents the weight matrix of the discriminator, C real represents the command feedback data received in the actual emergency scenario, and σ represents the activation function; S36, the system updates the node representation h in the graph model in real time during each round of generation and evaluation. i and the importance weight α of each node ij .

4. The wireless emergency command and broadcast method based on artificial intelligence according to claim 1 is characterized in that: The S4 specifically includes: S41. When an emergency occurs, the system determines the global priority of the emergency through high-level strategies and makes decisions through a multi-dimensional decision tree model. global The calculation formula is: Among them, w i Indicates the weight coefficient of the i-th factor, Score i represents the score of the ith factor, Risk i represents the risk index of the i-th factor, λ p represents the risk adjustment factor, Latency i Indicates the network delay of the i-th factor, Bandwidth i represents the network bandwidth of the i-th factor, σ represents the activation function, and n represents the total number of evaluation factors; S42, based on the calculated global priority Priority global ,The system selects an appropriate low-level strategy to optimize the dynamic priority of ,emergency commands. The low-level strategy adopts a ,temporal difference learning based strategy to refine the ,dynamic adjustment of command priorities; S43. The instruction priority update formula in the low-level strategy is: Among them, Priority t+1 Indicates the update priority at time step t+1, Priority t represents the current priority at time step t, α represents the learning rate, and r t represents the immediate reward of the current time step, γ represents the discount factor, Context t represents the context information of the current time step t, and φ represents the context adjustment coefficient; S44, the system based on the user's real-time feedback information F user , dynamically adjust the immediate reward r in the low-level strategy t , calculate the updated immediate reward r t+1 : Among them, β represents the feedback correction rate, Score feedback represents the user feedback score, ζ represents the response time adjustment coefficient, ResponseTime user Indicates user response time, ErrorRate network Indicates the network error rate; S45. After each priority update, the system calculates the current priority of different emergency instructions in real time. t+1 , and push the instructions to the corresponding users in order according to the priority level. The instructions with higher priority will be pushed first; S46. During the push process, the system continuously monitors the network status and user feedback information, and re-evaluates the command priority and adjusts the push order when necessary.

5. The method for wireless emergency command and broadcasting based on artificial intelligence according to claim 1, characterized in that: The S5 specifically includes: S51, the system collects user feedback information in real time through multimodal sensors and user terminal devices user , including user response time, operation results, success rate of instruction execution and user satisfaction score; S52, collect the user feedback information F user Perform preprocessing, including data cleaning, denoising and standardization operations, to obtain standardized feedback data F norm ; S53, the system will standardize the feedback data F norm And the user's historical feedback data F hist Fusion is performed to construct the fused feedback feature representation F fusion ; S54, based on the fused feedback feature representation F fusion , the system dynamically adjusts the generation and push strategy of emergency instructions through reinforcement learning algorithm, and the objective function L of reinforcement learning is: Where θ represents the discount factor, Q t represents the reward function at time step t, S env Indicates the current environment status; S55. After the command is pushed, the system continues to monitor new user feedback information F user , and update the fusion feedback feature representation F in real time fusion ,According to the real-time updated feedback information, the generation and push strategy of emergency instructions are optimized.,T represents the end time or termination condition of the entire decision-making process, and E represents the expected value.

6. The method for wireless emergency command and broadcasting based on artificial intelligence according to claim 1, characterized in that: The S6 specifically includes: S61. The system selects the most appropriate push strategy for different user types based on the user's identity, role, location, and environmental perception data. The decision function for the push strategy selection is: Among them, S strategy represents the selected push strategy, ρ i represents the weight coefficient of the i-th influencing factor, represents the impact evaluation value of the i-th influencing factor on the k-th strategy, and N represents the total number of influencing factors considered; S62. According to the selected push strategy, the system integrates voice, text, image and video to generate the emergency instruction push content that is most suitable for the user to accept. The generation formula of the push content is: Among them, C content Indicates the generated push content, η j Represents the weighting coefficient of the jth modal information, M j represents the content representation of the jth modal information, and M represents the number of types of modal information; S63. During the push process, the system adaptively selects the best push mode according to the current network conditions. The decision function for selecting the push mode is: Among them, M mode Indicates the selected push mode, Latency l Indicates the network delay of the lth push mode, ErrorRate l Indicates the network error rate of the lth push mode, Bandwidth l Indicates the bandwidth of the lth push mode; S64. The system adaptively adjusts the format and size of the pushed content according to the user's receiving ability and device characteristics. The adjustment formula for personalized pushed content is: C adapted =C content ·A device ; Among them, C adapted represents the push content after adaptive optimization, C content Indicates the original push content, A device represents the user equipment adaptability matrix; S65. The system continuously monitors and optimizes the push status during the push process; S66. Based on real-time monitoring and user feedback, the system dynamically adjusts push strategies and modes: Among them, S new Indicates the new push strategy, M new Indicates the new push mode, Satisfaction k,l represents the user's satisfaction under strategy k and mode l, Cost k,l represents the push cost of the user under strategy k and mode l, Benefit k,l Indicates the push effect of the user under strategy k and mode l.

7. An artificial intelligence-based wireless emergency command and broadcast system, which executes the artificial intelligence-based wireless emergency command and broadcast method according to any one of claims 1 to 6, characterized in that: Includes the following modules: Identity recognition module: used to automatically identify the user through wireless identity authentication and determine the user's role in the emergency scenario; Positioning and environmental perception module: used to combine the global positioning system, Wi-Fi signal positioning and IoT sensors to obtain user location and perceive environmental information in real time; Emergency instruction generation module: Based on the user's identity, role, location and real-time context, it uses graph attention network and generative adversarial network to generate personalized emergency instructions; Priority adjustment module: uses a hierarchical reinforcement learning algorithm to dynamically adjust the priority of emergency instructions based on the progress of emergency events and user feedback; Feedback mechanism module: used to collect user feedback information in real time and optimize the generation and push strategy of emergency instructions through reinforcement learning algorithms; Multimodal push module: Based on user characteristics and network conditions, emergency instructions are pushed to user terminals through multimodal communication technology, and optimized according to user receiving capabilities and device characteristics.

Images