Search

CN-121615953-B - 5G and AI multisource fusion decision system for earthquake rescue

CN121615953BCN 121615953 BCN121615953 BCN 121615953BCN-121615953-B

Abstract

The invention discloses a 5G and AI multi-source fusion decision system for earthquake rescue, and relates to the technical field of intelligent emergency rescue decision. The method comprises the steps of acquiring multisource original data with space-time identifiers through a 5G network in a low delay mode, extracting features by a stage judging module to generate stage identifiers, respectively carrying out weighting and reconstruction processing on the data based on the stage identifiers by a reliability evaluating module and a data reconstruction module to generate a data representation structure adapting to the current stage, receiving the stage identifiers by a scheduling control module and generating trigger signals, controlling a first analysis module and/or a second analysis module to respond to the trigger signals, respectively analyzing and generating life existence probability data and environment traffic class data from the data representation structure, and generating rescue decision instructions by a fusion decision module through a pre-trained multi-agent reinforcement learning model according to the stage identifiers. The earthquake rescue decision-making method improves timeliness, accuracy and resource utilization efficiency of earthquake rescue decisions.

Inventors

  • REN ZHIYUAN
  • YAN MENGMENG
  • YIN YUZHEN
  • XU XIUJIE

Assignees

  • 山东省地震局

Dates

Publication Date
20260505
Application Date
20260202

Claims (10)

  1. 1. Earthquake rescue-oriented 5G and AI multi-source fusion decision system is characterized by comprising: The data acquisition module is used for acquiring multi-source original data with space-time identification through a 5G network with low delay; The stage judging module is used for extracting stage characteristics of the multi-source original data and generating a stage identifier for representing the current rescue stage; the credibility evaluation module is used for carrying out credibility evaluation on the multi-source original data according to the stage identification to generate a weighted data set; the data reconstruction module is used for carrying out reconstruction processing on the weighted data set based on the stage identifier to generate a data representation structure adapting to the current rescue stage; the scheduling control module is used for receiving the stage identification and generating a trigger signal based on the stage identification; the first analysis module is used for responding to the trigger signal, analyzing the life related information based on the data representation structure and generating life existence probability data associated with the geographic position; The second analysis module is used for responding to the trigger signal, analyzing the environment traffic related information based on the data representation structure and generating traffic class data associated with the geographic position; And the fusion decision module is used for carrying out joint analysis on the life existence probability data and the traffic class data through a pre-trained multi-agent reinforcement learning model according to the stage identification to generate a rescue decision instruction.
  2. 2. The earthquake rescue oriented 5G and AI multisource fusion decision system of claim 1, wherein the specific steps of performing phase feature extraction on the multisource raw data and generating phase identifiers for characterizing a current rescue phase include: Acquiring at least three types of core data in multi-source original data, wherein the core data comprise earthquake magnitude time sequence data, aftershock frequency data and rescue force input distribution data; inputting the data into a pre-trained disaster evolution prediction neural network, and outputting the stage identification; Wherein the phase identification comprises at least a current phase predicted duration and a set of critical risk factors.
  3. 3. The earthquake rescue oriented 5G and AI multisource fusion decision system of claim 1, wherein the specific step of performing credibility assessment on the multisource raw data according to the stage identification comprises: dynamically adjusting weight evaluation strategies of different data sources based on the key risk factor set identified in the current stage; If the stage mark is a gold rescue stage, the weight coefficient of the life detection sensor data is improved, and the weight coefficient of the social media text data is reduced; if the stage mark is a secondary disaster high-rise stage, the weight coefficient of geological and meteorological sensor data is improved; And the weight coefficient is dynamically calculated by combining the stage mark through a fuzzy analytic hierarchy process to generate the weighted data set.
  4. 4. The earthquake rescue oriented 5G and AI multisource fusion decision system of claim 1, wherein the specific step of reconstructing the weighted data set based on the phase identification comprises: selecting a corresponding feature extraction backbone network according to the stage identification, and performing feature coding on the weighted data set; And carrying out data enhancement and missing data generation on the coded features by utilizing a generation countermeasure network associated with the stage mark to form the data representation structure with uniform space-time scale, wherein the data representation structure is a multi-channel space-time feature tensor constructed under the conditions of uniform time step and uniform spatial resolution.
  5. 5. The earthquake rescue oriented 5G and AI multisource fusion decision system of claim 1, wherein the scheduling control module is specifically a central scheduler, and the working mode of the central scheduler specifically includes: Receiving the stage identifier, and generating a corresponding trigger signal of the analysis module according to a preset rule mapping table so as to decide to trigger the first analysis module, the second analysis module or both in parallel; When the stage mark indicates that rescue force is limited and life signs are weak, the central scheduler generates a trigger signal which only triggers the first analysis module, performs spatial clustering and path optimizing preprocessing on the life existence probability data, and inputs the generated high-probability life sign aggregation area information to the fusion decision module as enhancement features.
  6. 6. The earthquake rescue oriented 5G and AI multisource fusion decision system of claim 1, wherein the specific step of performing life related information analysis based on the data representation structure comprises: Receiving the data representation structure, and extracting infrared thermal imaging characteristics, micro-vibration sensing characteristics and sound spectrum characteristics; inputting the characteristics into a plurality of light convolutional neural networks working in parallel for preliminary analysis; And carrying out dynamic weighted fusion on the output confidence coefficient of the multiple networks by using a particle swarm optimization algorithm, optimizing probability distribution model parameters, and generating the meshed life existence probability data.
  7. 7. The earthquake rescue oriented 5G and AI multisource fusion decision system of claim 1, wherein the specific step of performing environmental traffic related information analysis based on the data representation structure comprises: receiving the data representation structure, and extracting unmanned aerial vehicle image characteristics, topography digital elevation model characteristics and real-time weather data; constructing a cost map taking the passing difficulty as an optimization target; And simulating the collaborative exploration behaviors of a plurality of groups of rescue units in the cost map by adopting a wolf's group algorithm, dynamically evaluating and marking the traffic risk grades of different geographic areas through a division mechanism of the wolf's group, and generating the traffic grade data.
  8. 8. The earthquake rescue oriented 5G and AI multisource fusion decision system of claim 1, wherein the pre-trained multi-agent reinforcement learning model in the fusion decision module is based on a reward function Training and decision optimization are carried out, and the formula is as follows: Wherein: Dynamically adjusting the weight coefficient for the stage mark; representing a set of objects to be rescued, The life existence probability of the target i at the time t is represented; Represents a set of action paths of the rescue unit, Representing the passing cost of the path j at the time t; Represents a set of states of the rescue unit, As an indication function, when the estimated energy consumption of the cell k Below the safety threshold The value is 1, otherwise, the value is 0.
  9. 9. The earthquake rescue oriented 5G and AI multisource fusion decision system of claim 1, wherein after generating the rescue decision instruction, the system further performs the steps of: distributing the rescue decision instruction to a rescue terminal through a 5G network; Receiving instruction execution state data and real-time environment change data from a rescue terminal; taking the instruction execution state data, the real-time environment change data and the updated stage identification as inputs, and dynamically adjusting the strategy network parameters of the multi-agent reinforcement learning model by adopting an online strategy gradient algorithm; And generating and distributing an incremental decision instruction based on the multi-agent reinforcement learning model after parameter adjustment.
  10. 10. The earthquake rescue oriented 5G and AI multisource fusion decision system of claim 1, wherein when acquiring multisource raw data with space-time identification through a 5G network with low delay, specifically comprising: Performing space-time alignment and redundancy elimination on the multi-source original data by using a 5G network slicing technology; the multi-source original data at least comprises earthquake wave data from a distributed optical fiber sensing network, multi-mode physiological data from rescue personnel wearing equipment and disaster scene text and image data from social media.

Description

5G and AI multisource fusion decision system for earthquake rescue Technical Field The invention relates to the technical field of intelligent emergency rescue decision making, in particular to a 5G and AI multi-source fusion decision making system for earthquake rescue. Background The intelligent emergency rescue decision technology for earthquake is an important development direction in the field of disaster emergency management, and aims to generate an accurate and dynamic rescue action scheme by rapidly fusing multi-source heterogeneous information and applying an intelligent algorithm. In recent years, with the progress of 5G communication technology, internet of things and artificial intelligence algorithms, rescue decisions are developing toward data-driven, real-time response. In the prior art, various monitoring data such as satellite remote sensing, unmanned aerial vehicle aerial photography, ground sensors and the like are integrated, and a path planning or resource scheduling algorithm is combined to generate a static or quasi-dynamic rescue guidance scheme. The method can utilize information from different sources to a certain extent, form preliminary judgment on life signs of disaster areas or road damage conditions, provide effective auxiliary decision support when data are relatively stable, and have certain information integration and application capability. However, earthquake disaster has obvious dynamic evolution characteristics, and the core contradiction, key risks and data reliability at different stages can be radically changed from 'golden rescue' to 'secondary disaster high incidence' to 'post-disaster recovery'. Most of the prior schemes are static or single-target optimization in architecture design, lack of a core mechanism capable of perceiving and accurately describing the current rescue stage in real time, and fail to dynamically and adaptively globally coordinate the data processing weight, the analysis model emphasis point and the final decision target of the subsequent full link based on the cognition of the stage. This results in a system that has difficulty making optimal trade-offs between multiple objectives of "fight for seconds to search and rescue life" and "evade risk to ensure safety" in the face of rapidly changing complex disasters, with significant bottlenecks in timeliness, accuracy, and overall rescue efficiency of the output instructions. Disclosure of Invention Aiming at the defects existing in the prior art, the invention provides a 5G and AI multi-source fusion decision system for earthquake rescue. In order to achieve the above object, the technical scheme of the present invention is as follows: The invention discloses a 5G and AI multi-source fusion decision system for earthquake rescue, which comprises: The data acquisition module is used for acquiring multi-source original data with space-time identification through a 5G network with low delay; The stage judging module is used for extracting stage characteristics of the multi-source original data and generating a stage identifier for representing the current rescue stage; the credibility evaluation module is used for carrying out credibility evaluation on the multi-source original data according to the stage identification to generate a weighted data set; the data reconstruction module is used for carrying out reconstruction processing on the weighted data set based on the stage identifier to generate a data representation structure adapting to the current rescue stage; the scheduling control module is used for receiving the stage identification and generating a trigger signal based on the stage identification; the first analysis module is used for responding to the trigger signal, analyzing the life related information based on the data representation structure and generating life existence probability data associated with the geographic position; The second analysis module is used for responding to the trigger signal, analyzing the environment traffic related information based on the data representation structure and generating traffic class data associated with the geographic position; And the fusion decision module is used for carrying out joint analysis on the life existence probability data and the traffic class data through a pre-trained multi-agent reinforcement learning model according to the stage identification to generate a rescue decision instruction. Compared with the prior art, the invention has the beneficial effects that: 1. According to the invention, by introducing the stage mark with the collusion effect, a global cognition and self-adaption cooperative mechanism of the dynamic evolution of the rescue system is established. The method can automatically reconstruct the data processing key points and decision targets according to the real evolution law of rescue work, overcomes the inherent defect of insufficient adaptability of the traditional static architecture under multi-stage a