95 results about "Computational intelligence" patented technology

The invention relates to the technical field of artificial intelligence, in particular to a self-organizing knowledge base construction method, system and device based on a multi-agent system and a storage medium, and the method comprises the steps: S1, monitoring a dialogue information flow, calculating the self-reply confidence coefficient of an agent, and if the self-reply confidence coefficient exceeds a preset dynamic threshold value, judging that a potential knowledge value exists and triggering an extraction request; s2, in response to the request, performing semantic coding on a dialogue fragment, extracting knowledge elements, and generating a structured knowledge unit by utilizing graph neural network modeling and relation calibration; s3, through a pre-training semantic coding model, mapping the multi-dimensional semantic coding model into a multi-dimensional semantic vector; s4, carrying out hierarchical clustering comparison and combination, and if a threshold value is exceeded, establishing a new classification and adjusting a knowledge base index; and S5, knowledge fingerprints are generated and compared, and fusion updating is carried out based on the reputation scoring model when semantics conflict or redundancy occurs. According to the method, unstructured dialogue high-precision knowledge extraction is realized, so that the knowledge base structure is dynamically self-organized according to semantics, multi-source knowledge is coordinated, and the knowledge management automation level and quality are improved.

The invention relates to the technical field of edge computing, and discloses an edge computing intelligent decision-making system based on edge cloud collaboration, which comprises a terminal sensing layer, an edge computing layer, a network transmission layer and a cloud center layer. The terminal sensing layer is configured to collect original data; the edge computing layer comprises edge computing equipment, a resource scheduling module and a distributed synchronous triggering module which are respectively used for data preprocessing and model reasoning, dynamic resource allocation based on multi-agent deep reinforcement learning and multi-equipment synchronous acquisition realized through a precision time protocol; the network transmission layer dynamically adjusts a transmission strategy based on a reinforcement learning algorithm during data transmission; and the cloud center layer carries out training optimization on the model through a federal learning mechanism and issues the model to an edge side. According to the method, millisecond-level real-time intelligent decision making of an industrial site and sustainable evolution of the AI model are realized through a side cloud collaborative architecture, and the efficiency, the reliability and the self-adaptive capability of the system are effectively improved.

The invention provides a trailing suction dredger loading operation parameter optimization real-time recommendation system and method, and relates to the technical field of trailing suction dredger dredging engineering.The system comprises a data collection layer, a real-time calculation layer, an intelligent recommendation layer and a feedback control layer, the data collection layer is used for collecting trailing suction dredger loading operation parameters in real time, and the real-time calculation layer is used for calculating the real-time calculation layer; the real-time calculation layer can calculate energy consumption and yield in real time based on edge calculation equipment, the intelligent recommendation layer can generate a Pareto optimal solution set through a multi-objective optimization algorithm, and the dynamic weight adjustment module adjusts energy consumption and yield weight coefficients in real time according to the construction stage. The feedback control layer issues the collected loading operation parameters of the trailing suction dredger to an execution mechanism in real time through a PLC, and monitors the operation state of equipment to update a historical database, so that collaborative optimization of energy consumption and yield can be achieved, and the optimization precision is continuously improved through historical data.

The invention discloses a battlefield multi-agent dynamic cooperation method based on an attention mechanism. The method comprises the following steps: S10, acquiring local observation information; s20, establishing a strategy network and a value network, embedding a multi-head attention mechanism in each of the strategy network and the value network, allowing each agent to dynamically query the internal representation of a teammate, and converting a search problem of an exponential level joint action space into dynamic attention on key information; s30, the strategy network generates action probability distribution of the intelligent agent through the attention layer, and the intelligent agent is forced to depend on a cooperation signal; the value network outputs the value estimation of each agent and provides a reference function; the collaborative loss function calculates similarity penalty between agents based on the embedded representation output by the attention layer; and S40, the training process is updated by adopting a strategy gradient method, strategy network parameters are updated to maximize accumulated rewards, value network parameters are updated by minimizing total loss, and a highly cooperative joint strategy is output. According to the invention, the agent is ensured to always integrate team information during decision making.

The invention discloses a mobile robot adaptive path planning method based on enhanced Q learning and multi-strategy cooperation, and relates to the field of robot autonomous navigation and computing intelligence. The method comprises the following steps: firstly, establishing a two-dimensional grid map model containing obstacle information through an environment sensing module; the initial position of a gold scrubber population is generated by utilizing the low-difference characteristic of a golden section sequence to enhance the global ergodicity of a solution space, three stages of migration, gold scrubber and cooperation of a gold scrubber optimization algorithm are modeled into a discrete action space, an optimal search action is dynamically selected through a nonlinear attenuation greedy strategy, and the optimal search action is realized. According to the method, adaptive switching of exploration and development behaviors is realized, a search gradient is constructed by using a differential vector of survival of the survival of the survival of the population, and a lens imaging reverse learning strategy is introduced to dynamically disturb an elite individual so as to jump out of a local extremum trap. And finally, outputting an optimal collision-free path through a multi-objective evaluation function including path length, safety and smoothness.

The invention relates to the technical field of public safety personnel evacuation, discloses a public emergency emotional infection simulation method, and aims to solve the problems that an existing emotional infection simulation model does not fully consider a group emotional propagation theoretical mechanism and the influence of personality characteristics of an intelligent agent and is insufficient in simulation precision and authenticity. The method is based on an improved SIRS (Susceptable-Infected-Recovery-Susceptable) model and an OCEAN large five-personality model, and is realized through the following six core steps: 1) constructing a three-dimensional simulation model of a public place; 2) initializing personality characteristics, emotional states and behavior states of the agents; 3) calculating an emotion propagation process (including a propagation range, an influence value, a distance weight and emotion conflict processing) between the intelligent agents; the method comprises the steps of (1) acquiring the emotion of the intelligent agent, (2) calculating the emotion attenuation of the intelligent agent, (3) determining the evacuation speed and completing analogue simulation, (4) realizing the behavior logic of the intelligent agent in different emotion states, (5) calculating the emotion attenuation of the intelligent agent, and (6) determining the evacuation speed and completing analogue simulation. The method can accurately restore the emotion infection and evacuation behaviors of the crowd in an emergency scene, improves the simulation authenticity and lightweight level, and provides decision support for emergency evacuation scheme making and emergency management.

The invention discloses a multi-agent cooperation method, system and device based on shared memory data, and a medium, mainly relates to the technical field of multi-agents, and aims to solve the problem that a traditional reasoning system cannot recognize semantic similarity between requests and cannot sense a context relationship of agent tasks, so that the efficiency is improved. The intelligent agent system is often deployed at multiple nodes, and the cache result cannot be shared across the nodes. Comprising the steps of obtaining an output result and an intermediate result generated by each path node; caching the semantic signature and the output result to a preset result cache layer; caching an intermediate result generated by each path node as context information to a preset context cache layer; and according to a preset time window, detecting inference frequencies of different semantic signatures generated by all the agents, and according to the inference frequencies, adjusting storage time of output results corresponding to the semantic signatures in a preset result cache layer.

The invention discloses a power distribution network situation deduction method and system considering multi-type source-load interaction, and relates to the technical field of power distribution network situation prediction. The method comprises the following steps: acquiring operation states of various types of source loads and power distribution equipment in a power distribution network; capturing a collaborative prediction relationship between the source load and the power distribution equipment by using a multivariable time sequence coupling model; constructing an electric power AI large model, and pre-training the electric power AI large model by using an autoregression method; performing parameter fine tuning on the pre-trained electric power AI large model by adopting a plurality of fine tuning technologies; and deducing the situation of the power distribution network according to the operation states of the source load and the power distribution equipment and the collaborative prediction relationship between the source load and the power distribution equipment by using the fine-tuned power AI large model. According to the method, on the premise that large-scale flexible resources are contained and the flexible resources continuously respond to the interaction strategy of the power distribution network, calculation intelligent deduction of multi-type source-load interaction of the short-term operation situation of the power distribution network can be achieved.

The application discloses a user rate enhancement method of an intelligent reflecting surface assisted FTTR system, and belongs to the field of wireless communication.The method comprises the following steps: S1, determining the signal-to-interference-and-noise ratio of a target user according to the beamforming vectors of a target AP and each interference AP and the reflection phase matrix of an IRS, so as to represent the rate of the target user; and S2, maximizing the rate of the target user to obtain the optimal reflection phase matrix of the IRS.The method takes the rate of the target user as a target function, simultaneously considers the enhancement of the target user signal and the suppression of the interference signal by the intelligent reflecting surface, and adjusts the weight of the reflection unit of the IRS to enhance the user rate.The intelligent reflecting surface becomes a competitive communication resource, thereby greatly reducing the complexity of active and passive beam joint adjustment and greatly improving the achievable rate of the target user.The passive beam of the intelligent reflecting surface is quickly calculated through a Riemannian manifold gradient algorithm.

A robot collaborative control system based on rigid formation includes: an image acquisition module for acquiring the geometry of the object to be transported; an undirected perception module for confirming the desired formation based on the geometry of the object and constructing an undirected perception topology graph; calculating the relative distance and relative angle between agents based on the global coordinates of each agent in the undirected perception topology graph, and confirming the adjacency relationship; a motion control module for receiving control commands and constructing a controller based on the relative distance error, relative angle error, and adjacency relationship between agents; and a drive module for generating torque according to the controller to drive the motor to perform the task. This invention uses a distributed rigid formation control method to form a directional rigid formation at the grasping end that matches the shape of the object being transported, thereby avoiding pulling damage to the object during transport and effectively controlling the posture of the object during transport.

The invention discloses a distributed processing method for funnel chest treatment influence data in combination with an edge network, and relates to the technical field of medical information technology and edge computing intelligent processing, and the method comprises the steps: S1, collecting the dynamic image and sign data of a patient through an edge node, carrying out the preliminary classification of the collected real-time data through a distributed processing frame, and carrying out the calculation of the collected real-time data; determining a data priority sequence according to the mode applicable symptom and the historical curative effect data, and obtaining a real-time data group with the priority higher than a preset threshold value; s2, according to the real-time data groups of which the priorities are higher than a preset threshold value, executing on-site conversion operation at edge nodes, grouping converted data features by adopting a k-means clustering algorithm, and determining feature groups meeting conditions; according to the distributed processing method for the funnel chest treatment influence data in combination with the edge network, the data processing real-time performance, the treatment scheme individuation and the cross-node consistency are remarkably improved, and an efficient and accurate treatment intervention path is provided for a patient.

A rice japonica seed classification method and system based on a support vector machine belong to the technical field of computing intelligence and optimization algorithm, and solve the technical problem that the existing technology depends on experience to select support vector machine parameters, thus missing the optimal solution to the problem, resulting in reduced rice japonica seed classification accuracy. High spectral image data of rice japonica seeds is collected; based on a fusion elite optimization strategy, a fog optimization algorithm is initialized to generate an initial solution; a hidden mechanism is used to improve the greedy strategy for population updating; the improved fog optimization algorithm is used to optimize the penalty coefficient and kernel function parameters in the support vector machine, and a mapping rule is designed to iteratively optimize the kernel function in the support vector machine to obtain an optimal solution; the optimal solution is set as the support vector machine parameters, a rice japonica seed variety classification model is constructed, and rice japonica seed variety classification is realized. The present application is used to realize efficient and accurate rice japonica seed classification.

The invention discloses an implementation method and system for AI and Agent agents to carry out autonomous team-forming social contact, and relates to the technical field of AI and Agent agents, and the implementation method comprises the steps: generating a cognitive state vector with a variable time sequence based on a knowledge structure, a reasoning mode and decision preference of the agents, and forming a multi-dimensional semantic representation representing individual cognitive characteristics; calculating cognitive similarity among the agents according to the cognitive state vectors, and when the cognitive similarity exceeds a preset threshold value and the average resonance intensity of the group reaches a critical condition, triggering the multiple agents to spontaneously aggregate into a temporary cooperative community without central control; in the temporary collaboration community, the social identity of each agent is determined based on the trust strength, the communication influence and the task contribution degree among the members, the social identity comprises at least one of a leader, a coordinator, an innovator or an executor, and identity state transition is modeled through a Markov decision process.

The invention provides a multi-agent robust decision-making method fusing a causal attention network and trajectory prediction, and belongs to the field of multi-agent decision-making control. The problem that a traditional decision model is insufficient in safety, robustness and interpretability is solved. The method comprises the following steps: acquiring historical and real-time observation data of all agents in a target scene; extracting the dynamic behavior characteristics and the interaction dependency relationship of the intelligent agent, and outputting comprehensive spatial-temporal characteristics; constructing a causal discovery network, calculating a causal probability between agents, and outputting a causal probability matrix; constructing a causal self-attention network, extracting space-time causal interaction features, inputting the space-time causal interaction features into a trajectory prediction decoder, and generating agent trajectory prediction features; and fusing the space-time causal interaction features and the trajectory prediction features to form comprehensive feature representation, inputting the comprehensive feature representation to a reinforcement learning decision network based on a PPO algorithm, outputting an intelligent agent behavior decision instruction, and completing online optimization and updating of strategy parameters. The method is used in the fields of robot cluster cooperation and intelligent traffic control.

The present application belongs to the technical field of intelligent transportation system, and relates to a signal-vehicle cooperative optimization method based on MPC and trust region Bayesian optimization. Firstly, traffic flow data collection and road network simulation modeling are performed, and initial path allocation of vehicles is performed, then real-time speed control is performed on intelligent connected vehicles entering the road network, and the best driving speed is calculated by using a Bayesian optimization algorithm based on a trust region according to signal light information, distance from the intersection and driving conditions of surrounding vehicles. Then, the road impedance is updated, and the best driving path of the intelligent connected vehicle is recalculated. Then, the future traffic state is predicted by using model predictive control, and the best green light time of each intersection in different phases is calculated by using trust region Bayesian optimization. The best green light time is imported into the simulation platform, and vehicle path optimization and speed control are performed again, and the cycle feedback is performed. Finally, it is judged whether the vehicle leaves the road network or reaches the simulation end time, and the vehicle control and signal optimization are ended.

The invention discloses a highway engineering construction resource dynamic optimization scheduling system based on the Internet of Things and an implementation method, and the system comprises a multi-dimensional perception layer, an edge calculation transmission layer, an intelligent decision scheduling layer and an execution feedback layer, and each layer forms a perception-calculation-decision-execution-feedback closed-loop scheduling system. The multi-dimensional sensing layer is deployed on a highway construction area and various construction resources; the edge calculation transmission layer performs real-time preprocessing and local calculation on the multi-source data acquired by the sensing layer; a dynamic optimization scheduling model based on reinforcement learning is built in the intelligent decision scheduling layer, and a resource scheduling optimization scheme is output; the execution feedback layer is used for driving the construction resources to execute the scheduling instruction. According to the method, the core problems of one-sided perception, calculation lag, decision-making rigidity, feedback deficiency and the like in a traditional mode are solved, road construction resource scheduling is promoted to be transformed from experience-driven to data-driven and intelligent decision-making, and finally optimal configuration of construction resources and efficient management and control of the construction process are achieved.

The application belongs to the technical field of knowledge graph, and particularly relates to a knowledge reasoning method and system based on an agent dynamic path completion strategy, which extracts entities and the relationships between the entities in a target knowledge graph, and mines rules in the target knowledge graph and confidence score corresponding to the rules; a reinforcement learning agent is constructed, the agent is dynamically guided to complete the path of the knowledge graph according to the current entity state and historical path information through rules, and the total reward of the agent is calculated according to the hit reward when the agent hits a target entity and the rule reward when the reasoning path of the agent conforms to the rules, the strategy network of the agent is trained by maximizing the expected value of the total reward of the agent, and the corresponding knowledge reasoning result is obtained by using the trained agent to perform path reasoning in the knowledge graph for a given target condition to be queried. The application dynamically completes the most likely path in the reasoning process by using the dynamic path completion strategy to obtain a complete reasoning path, and solves the reasoning truncation problem caused by the missing path of a sparse knowledge graph.

The invention discloses an oral expired gas data preprocessing and batch correction method based on PTR-TOF-MS. The method comprises the following steps: selecting a spectrogram with a peak height and a stable waveform in a middle section, removing an instrument abnormal section without a peak height and / or lower than a sample inspection threshold value, and calculating an intersection-to-union ratio of an interval marked by an intelligent agent to a high-confidence coefficient generated by manual marking or simulation; if the intersection-to-union ratio exceeds a preset threshold value, giving positive rewards, if the intersection-to-union ratio does not exceed the preset threshold value, continuing to judge whether the starting point, the vertex and the end point of the spectrum peak are missed or repeatedly marked, if so, giving punishment, if not, continuing to judge whether the noise interval is marked as the starting point, the vertex and the end point of the spectrum peak, and if so, giving additional punishment, training the strategy network; and outputting a structured peak list from m / z intervals of a spectrum peak starting point, a peak and an end point for each sample, continuing to perform integral calculation, monotonic transformation and ComBat correction to obtain a final data matrix, and realizing adaptive recognition of complex spectrum peaks and effective elimination of batch effects.

The “Industrial Digestive System” encapsulates a novel paradigm in waste management and resource recovery, bridging the gap between sophisticated computational intelligence and industrial automation. The hardware component replicates the natural digestive system's efficiency in processing a diverse range of inputs, including municipal and industrial waste, transforming them into valuable outputs like biofuels, chemicals, and advanced composite / conglomerate materials. This process is achieved through a series of mechanized operations. Complementing the hardware, the software facet of the invention is rooted in advanced machine learning and optimization algorithms that optimize multivariable functions for the most efficient conversion of input to resources, orchestrating the transformation from varied waste streams, elements, and molecules into solid, liquid, and gaseous outputs. The result is an automated, intelligent manufacturing machine that enhances product quality and operational efficacy. “The Industrial Digestive System” can be seen as a new branch of cybernetics, a cyber-physical system inspired by nature.

The application relates to an expandable multi-agent hierarchical navigation method and device based on a graph network, wherein the method comprises the following steps: receiving state information of all agents and position information of target points, and respectively generating an agent relationship graph of all agents and a target point relationship graph of all target points according to the state information and the position information, so as to calculate an agent-target point relationship matrix, and match a corresponding optimal target point for a current agent through the relationship matrix; fusing the state information and the position information through a subgraph fusion strategy to obtain a fusion result, and encoding the fusion result according to a multilayer perception strategy, extracting an encoding relationship between the current agent and the optimal target point, and generating an optimal path between the current agent and the optimal target point. Therefore, the problems that the existing multi-agent navigation method has large calculation amount and poor performance, the action space and the state space of multiple agents are too complex, overfitting is prone to occur, the migration ability is poor, and large-scale application is difficult are solved.

The invention discloses a multi-modal large model ground penetrating radar disease data intelligent interpretation method and device, and belongs to the computing intelligence and information processing technology. The method is based on a computing device of machine learning, artificial intelligence and a specific mathematical model, and an inference model of specific ground penetrating radar knowledge is fused for information processing. Comprising the following steps: acquiring underground disease data by using multi-frequency ground penetrating radar equipment, and converting the underground disease data into a serialization format adaptive to a multi-modal large model; and then sequentially executing detection, segmentation and image description tasks. Wherein a segmented label mask is generated by adopting a Bayesian optimization method based on a target detection result, and an image description data set is constructed by utilizing a segmentation result; and finally, performing intelligent interpretation on the data by using the multi-modal large model to generate an interpretation report. According to the method, through multi-task collaborative learning, full-process automatic interpretation from disease detection to semantic description is realized, the interpretation efficiency and reliability are improved, and an intelligent basis is provided for road maintenance decision making.

The present disclosure provides a search term recognition method and device, electronic equipment and storage medium, relates to the technical field of data processing, in particular to the technical field of cloud computing and intelligent search, and includes: obtaining a search term input by a user; identifying the search term by using a first search term recognition model to obtain a search term recognition result and a search term identifier; determining whether the search term is of a first specific type according to the search term recognition result; in response to the search term being of the first specific type, matching the search term identifier in a first preset database; and in response to a successful match, obtaining a matching result corresponding to the search term and sending the matching result to the user.

The disclosed embodiments disclose techniques for recovering from a ransomware attack using computational intelligence. During operation, a monitoring system detects that a ransomware attack has encrypted multiple subsections of a data storage that is simultaneously being accessed by multiple processing units, and receives a set of encryption keys that were harvested for the multiple processing units. The system analyzes the contents of the data storage to determine a pattern in the multiple subsections that have been encrypted, and then uses the encryption keys and the pattern to decrypt the multiple encrypted subsections.

The invention discloses an AI training service-oriented intra-network computing intelligent arrangement method and device, and the method comprises the steps: obtaining a system state, and constructing a mixed integer linear programming model MILP according to the system state; inputting the system state into the LLM, generating a new heuristic algorithm by adopting a computing resource layout heuristic algorithm based on the LLM, and iteratively selecting an optimal layout heuristic algorithm; an optimal layout heuristic algorithm is adopted to select a plurality of switches to deploy an intra-network computing function, namely, the switches serve as programmable switches to form a computing resource layout; after the MILP is relaxed, a solver is adopted to solve a flow distribution rate and an ideal sending rate of a working node of the AI training task for sending a gradient to a programmable switch and a PS in the computing resource layout; according to the flow distribution rate, the AI training task is distributed to a programmable switch and a PS in the computing resource layout, and a gradient routing path is formed; and sending the computing resource layout and the gradient routing path to a physical network device where the AI training service cluster is located for execution.

The present disclosure relates to the technical field of communications, and in particular provides a communication system and method, a network function, an electronic device, a storage medium, and a product. The communication system of the present disclosure comprises first network functions and second network functions. The first network functions are used to implement communication connection functions. The second network functions are communicatively connected to at least one of the first network functions, and are used to implement at least one of the following first functions: a data management function, a data analysis function, an artificial intelligence / machine learning (AI / ML) model-related function, a computing function, and a resource collaborative control function. The technical solution provided in the present disclosure can improve system flexibility and scalability by means of a hierarchical network architecture, and provide system support for multi-dimensional service capabilities such as computing, intelligence, and perception.

The present application belongs to the cross technical field of geotechnical engineering precision monitoring and calculation intelligence, and discloses a kind of deep foundation pit support pile micro-deformation capture and prediction method and system fusing physical mechanism constraint, method includes following steps: obtaining and fusing the static physical field and dynamic displacement field fusion of deep foundation pit support pile, construct enhanced space-time field atlas containing physical boundary condition;Micro-deformation prediction model is constructed, and physical information loss function is constructed based on support pile beam unit four-order differential equation, and micro-deformation prediction model is trained based on physical information loss function, and the model after training is obtained;Enhanced space-time field atlas is input into the model after training, and the spatial form feature of the micro-strain of pile body in enhanced space-time field atlas is extracted, and the deformation prediction result of support pile is obtained.The present application breaks through the technical bottleneck of "micro-feature annihilation" of general model, and realizes high-fidelity capture of micron-level deformation.

The invention discloses a spectrum quality control system and method based on a dynamic repeatability limit rule base and a storage medium, and relates to the technical field of analytical chemistry and spectrum analysis quality control. The system comprises a data import module, a dynamic repeatability limit rule base, an intelligent judgment module, a result export module and a data storage module. The system manages complex repeatability limit rules related to the content through a structured rule base, and dynamic calculation of the rules is achieved through a built-in formula analysis engine. The intelligent judgment module takes an absolute difference value of two times of measurement as a core judgment basis, performs statistical expansion processing on multiple times of measurement conditions, completes comparison and judgment in batches through a full-automatic process, and finally outputs a visual report and archives full-process data. According to the invention, the technical problems of complex repeatability limit rule, low manual processing efficiency and error proneness in high-throughput spectrum analysis are solved, and efficient, accurate and traceable automatic quality control is realized.

This invention relates to a system and method for human authentication utilizing a graphene membrane for human skin scanning, imaging, and communication to determine unique neuron patterns for use in authentication and psychological profiling. The system employs a graphene membrane, which integrates unobstructed optical, ultrasound, and capacitance sensors within a distributed computing smartcard format device. The graphene membrane optical sensors capture high-resolution images of fingerprint ridges and valleys, while the ultrasound sensors generate detailed 3D maps of both surface and subsurface skin structures. Capacitance sensors measure the electrical properties of the skin, further enhancing the biometric data obtained. The biometric data is combined and used to identify distinct neuron patterns embedded in the skin. The system may be manufactured in compact card format and durable, ensuring seamless integration into standard identification card formats and to act as a consensus node for Proof-of-KYC (Know Your Customer).

This invention discloses a method for analyzing massive amounts of data in smart cities based on edge computing, and relates to the field of data analysis technology. The method includes the following steps: constructing edge computing task orders based on the original time-series data stream and business processing intentions of the smart city; collecting static and dynamic data from edge computing nodes to generate a global node capability profile set; constructing an edge computing intelligent scheduling model, inputting the task orders and the node capability profile set into the model, and outputting scheduling decision instructions; ensuring the reliable execution of these instructions, and producing edge local analysis results. This invention enables the analysis of massive amounts of data in smart cities based on edge local analysis results.

The application provides a multi-agent interaction simulation method based on stereoscopic vision information, which comprises the following steps: providing each agent with a virtual binocular camera, dividing the field of view into multiple sectors and performing nearest neighbor search, generating a focus list, and then calculating the physical speed of the agent; at the same time, training the agent navigation strategy by using a reinforcement learning network, generating a navigation decision speed in combination with a multi-level reward mechanism; adaptively weighting and fusing the navigation decision speed and the physical speed, dynamically adjusting the weight proportion according to the environmental conditions, and generating a final speed instruction; finally, executing the instruction in a simulation environment, driving the agent to move, and outputting simulation results containing trajectories and performance indicators. The application breaks through the assumption that the agent has perfect perception ability in traditional crowd simulation, significantly improves the realism, behavior rationality and robustness in complex scenes of the simulation, and can be widely applied to the fields of evacuation simulation, game development and virtual reality.