43results about How to "Improve system robustness" patented technology

The invention discloses a biased binocular visual odometer method and device based on strawberry cuttlefish bionics. The biased binocular visual odometer method comprises the following steps: synchronously acquiring a bright field image and a dark field image collected by a biased binocular camera and inertial data of an inertial measurement unit; constructing a perceptual attention weight grid according to the bright field image and the dark field image, and performing brightness normalization processing on the bright field image and the dark field image; extracting lateral visual features based on the normalized image and the perceptual attention weight grid; and fusing the biased visual features with the inertial data, and solving through nonlinear optimization to obtain a carrier pose estimation result. According to the invention, the dynamic range of the binocular vision system is expanded, and the imaging and sensing capabilities of the system in an HDR scene are improved. And meanwhile, the positioning precision and robustness of the binocular visual odometer in an HDR scene are improved.

The present application relates to a gas tracing method, in particular to a kind of environmental identification and tracing method, system and medium based on multi-modal cooperation, method includes: collection contains gas plume visual image and contains gas response signal olfactory signal;Visual image is preprocessed and visual feature is extracted, and visual depth feature vector is obtained;Olfactory signal is preprocessed and olfactory feature is extracted, and olfactory feature vector is obtained;Cross-modal fusion obtains fusion feature vector;Physical constraint loss function is constructed, and physical consistency correction is carried out, and modified fusion feature vector is obtained;Output gas concentration space-time distribution result.Compared with prior art, the present application solves the problem that prior art cannot realize high-precision online monitoring and tracing of multi-component gas.The scheme improves the identification accuracy of gas component, the multi-component distinguishing ability and real-time response performance in complex environment through the integration of visual monitoring and olfactory monitoring, combined with multi-modal fusion algorithm and physical consistency correction.

The present application relates to a kind of based on multi-target patch type ultrasonic hemodynamic state quantification and classification method, comprising: obtaining the ultrasonic echo signal of multiple channels and carrying out filtering and gain compensation preprocessing;Extract the hemodynamic physical parameter of each channel;The hemodynamic physical parameter of each channel is carried out multi-scale coarse-grained processing to construct multi-scale fuzzy entropy spectrum, and hemodynamic physical parameter is mapped as the local entropy grade score of the channel based on multi-scale fuzzy entropy spectrum;According to the signal quality evaluation result of each channel, determine dynamic weight, based on dynamic weight, the local entropy grade score of each channel is weighted and summed to obtain comprehensive quantification index;The time sequence of the comprehensive quantification index of multiple continuous moments is analyzed and output stability label by recursion characteristic;The local entropy grade score of each channel, comprehensive quantification index and stability label constitute multidimensional feature vector, input preset classifier, and output hemodynamic state classification result.

The present application relates to the technical field of power distribution network load management, and discloses a power distribution network line current out-of-limit load optimization method, system, device, medium and product, the method comprising: obtaining an overload out-of-limit line in a power distribution network, and identifying all tie lines having a physical tie relationship with the overload out-of-limit line in combination with line topology information of the power distribution network; obtaining operation information of each tie line, and selecting effective tie lines from the multiple tie lines according to the operation information; determining a plurality of transferable power supply paths according to the tie relationship between the effective tie lines and the overload out-of-limit line; performing multi-objective transferable power supply capacity evaluation on each transferable power supply path, and selecting an optimal transferable power supply path from the plurality of transferable power supply paths according to the evaluation result; and performing a load transfer operation on the overload out-of-limit line according to the optimal transferable power supply path, thereby significantly enhancing the dynamic adaptability and overall robustness of the power distribution network.

This invention discloses an AI-based data center operation and maintenance security system, belonging to the field of data center operation and maintenance security technology. It constructs an operation and maintenance intent fingerprint set, combines equipment topology relationships, power connection relationships, and airflow organization relationships to determine the disturbed closed domain, and establishes a counterfactual security twin model based on the disturbed closed domain and a reference equipment group. It performs mapping analysis on multi-source real-time data, extracts interpretable disturbance components, and separates residual abnormal sequences. It achieves anomaly propagation tracking through time-delay coupling relationships, determining the initial source point and propagation chain of the anomaly. Furthermore, it calculates the operation and maintenance intent deviation entropy based on path information entropy and the cross-domain coupling hysteresis gain coefficient based on the frequency domain coherence function, comprehensively evaluates each node within the disturbed closed domain, determines security vulnerability locations, and outputs corresponding handling instructions. This invention can effectively distinguish between operation and maintenance disturbances and abnormal disturbances, achieving accurate anomaly location and proactive intervention, thus improving the security and reliability of data center operation and maintenance.

This invention discloses a collaborative training method and system for autonomous driving planning and control, belonging to the field of autonomous driving technology. The method includes: acquiring the current state of the vehicle, the states of surrounding traffic participants, and reference road information; using a Transformer-based planning network to output a control sequence, which is then used to generate a planned trajectory via a vehicle kinematics model; inputting the planned trajectory into a model predictive control-based neural network controller for dynamic tracking optimization to obtain the executed control quantity and predicted vehicle dynamic state; constructing a control loss and backpropagating its gradient to the planning network; and, during training, layering constraints on the planning cost term, using safety and traffic rule-related terms as hard constraints and comfort-related terms as soft constraints, and employing a multi-stage cost scheduling strategy to adjust the constraint strength or weight. This invention achieves collaborative optimization of planning and control, improves the dynamic executability and safety compliance of the trajectory, and enhances training stability and closed-loop robustness.

The application discloses a kind of short message channel intelligent matching system based on multidimensional weight distribution, specifically related to short message channel intelligent matching technical field, to solve the problem of channel matching weight mechanism rigidification, cost control extensive, insufficient sending scheduling stability in prior art, including channel hierarchical screening, dimension weight matching, cost accounting control, four big modules of dispatch execution self-healing;Through customer industry, short message type and so on attribute completes channel preliminary screening, locks available channel pool, in combination with customer subscription grade sets differentiated weight, based on cost, quality and so on multidimensional index calculates channel performance coefficient, selects optimal primary and backup channel;While constructing dynamic premium logic realizes fine pricing, quantifies channel abnormal degree by sliding window, executes gradient shunting and automatic switching, the application can realize channel accurate matching, gross profit effective guarantee and sending stability improvement, adapts to enterprise differentiation service demand.

This invention discloses an intelligent projection positioning and drilling accuracy control system for blast holes at tunnel faces, belonging to the field of tunnel drilling and blasting construction technology. It aims to solve the technical problems of inaccurate blast hole positioning and large drilling direction deviations in existing tunnel face systems, leading to low construction efficiency. The key technical points are: the control system includes a structured light projector mounted on a tunnel drilling and blasting trolley, at least two industrial cameras, a fiber optic shape demodulation module, and a controller. A fiber optic shape sensing unit is installed inside the drill rod at the end of the drill arm of the drilling and blasting trolley; the structured light projector projects structured light, the industrial cameras acquire binocular images, the fiber optic shape demodulation module calculates the curvature and torsion angle sequence of the sensing unit, and the controller controls the drill arm drive mechanism to correct the drill bit position and direction through 3D reconstruction, coordinate comparison, and real-time deviation adjustment. This invention achieves intelligent positioning and drilling accuracy control of blast holes at tunnel faces, improving the accuracy and efficiency of tunnel drilling and blasting construction.

The application discloses a robot grabbing method and device, a computing device and a computer readable storage medium. The method comprises the following steps: obtaining joint state space information of a robot; inputting the joint state space information into a pre-trained reinforcement learning collaborative decision network; outputting chassis control instructions and mechanical arm grabbing instructions in parallel by the reinforcement learning collaborative decision network; controlling the movement of the mobile chassis according to the chassis control instructions, and controlling the mechanical arm to perform a grabbing action according to the mechanical arm grabbing instructions, so that the mobile chassis and the mechanical arm are controlled in parallel. In this way, the grabbing efficiency of the robot in a dynamic environment can be improved.

The invention discloses an augmented state extended Kalman filter-based hardware delay dynamic tracking method, system and device suitable for ground disaster monitoring, and a medium, and belongs to the technical field of satellite navigation and communication, and the method comprises the steps: obtaining historical inter-satellite link observation data of a satellite constellation, performing batch processing on historical inter-satellite link observation data by using a graph neural network model to obtain a hardware delay initial value of each satellite in a constellation; an augmented state vector is constructed, the orbit state, the clock error state and the hardware delay of the satellite are jointly used as to-be-estimated state quantities, and the augmented state vector is initialized by adopting a hardware delay initial value; and iteratively predicting and updating the augmented state vector by using inter-satellite link observation data obtained in real time. According to the method, the problem of dynamic compensation of hardware delay is solved through a two-stage strategy of offline initialization and online tracking.

This application provides a mobile robot task scheduling method and system based on a heterogeneous computing architecture, relating to the field of intelligent scheduling technology. The method includes: constructing a workspace geometric model based on the geometric structure of a planting row; obtaining the motion state parameters of the mobile robot using a first computing unit; determining the look-ahead compensation amount by combining the statistical value of the inference time of the target recognition task by a second computing unit; determining a spatiotemporal pre-aiming slice in the visual field region based on the look-ahead compensation amount and the workspace geometric model, and cropping image data located within the slice from the original image data; the second computing unit performing target recognition on the image data, generating a work instruction corresponding to the recognized target, and controlling the lateral actuator to perform the work according to the work instruction. This can alleviate the spatiotemporal asynchrony problem between perception results and execution position in a moving work scenario, reduce missed and misaligned tasks, and reduce the communication and computational load of the heterogeneous computing system.

The invention relates to the technical field of crane fault detection, in particular to a fault diagnosis method and system for a crane motor. Comprising the following steps: acquiring motor operation data in real time through a sensor, extracting time domain and frequency domain statistical characteristics, and generating a working condition classification label; judging an operation mode according to the working condition classification label, determining a deviation vector of an operation sequence and a standard health sequence, and calculating an abnormal deviation threshold value; calculating a similarity matrix of a current sequence and a historical sequence, and optimizing a weight set; a comprehensive diagnosis index is formed, and whether a potential fault signal is generated is judged; and if the fault signal exists, updating the diagnosis model parameters to form an optimized diagnosis framework, analyzing the operation data based on the framework, and fusing the working condition classification label and the deviation vector to obtain a final fault diagnosis result. According to the method, the problems of low fault identification precision, diagnosis lagging and non-adaptive model of the crane motor under multiple working conditions are solved, and intelligent monitoring and dynamic accurate diagnosis of the running state of the motor are realized.

The invention discloses a performance constraint control method and system for multi-robot collaborative carrying of large workpieces. The method comprises the steps that firstly, a unit quaternion is adopted to establish a coupling dynamic model containing closed-chain constraint under a task reference coordinate system; secondly, designing a completely distributed finite time observer, and performing consistency estimation on a global reference trajectory only through neighborhood communication; secondly, designing a consistency indicator function based on the estimated value, gradually activating a performance constraint function only after an observation error is converged so as to apply transient and steady-state performance boundaries, and generating an expected motion state of each robot; and finally, designing a distributed robust adaptive controller to realize trajectory tracking. According to the method, singularity is avoided through quaternion modeling, controller design of a task reference coordinate system is achieved through rigid body transformation, dependence on a centroid coordinate system is avoided, the requirement for a priori track is eliminated through a distributed observer, constraint conflicts are avoided through a strategy of'feasibility first and then constraint ', and the robustness of the system is improved. And the safety, the precision and the robustness of collaborative carrying are effectively improved.

The invention discloses a log analysis method, system and equipment of a database and a computer readable storage medium. According to the method, a log set is analyzed, if no abnormal log exists in the log set, a check point is generated based on the log with the latest generation time in the log set, logs included in the log set are marked as analyzed logs, a new log set is determined from to-be-analyzed logs, the new log set serves as the log set, and the log set is analyzed according to the new log set. Analyzing the log set; if the abnormal log exists in the log set, determining a target check point from the generated check points according to the error type of the abnormal log; the new log set is obtained according to the target log and the error type of the abnormal log, the new log set is used as the log set, the log set is analyzed, and by generating and dynamically selecting the check points, different check points can be selected for backspacing when analysis abnormity occurs in the data synchronization process, so that the load of a source library is reduced, and the data synchronization efficiency is improved. And the data synchronization efficiency is improved.

The present application belongs to the technical field of automatic filling and management and control of disposable consumables, and discloses a digital filling system for medical iodophor cotton swabs based on closed-loop control, which comprises: obtaining a physical trigger signal of a feeding station, generating a space-time ID of a single cotton swab, activating an exclusive physical storage space, collecting initial production data and binding; synchronously driving filling action execution and data collection, performing real-time deviation determination, triggering hierarchical closed-loop correction action when deviation is detected; positioning corresponding cross-station data, generating a feedforward correction instruction, integrating multi-station data, and forming a full-link cross-station data chain; screening filling quality verification indicators, verifying filling quality status, triggering reverse correction routing and precise rejection action for unqualified products, forming a closed-loop data flow and a reverse correction instruction set; archiving the full-process production data of a single cotton swab, generating a single production file, performing basic statistics during non-production period, dynamically optimizing control benchmarks, and forming full-process closed-loop iterative management and control.

The application provides a label printer multi-node access edge cloud collaborative control method and system, relates to the technical field of edge cloud collaborative control, a local agent node is connected with a label printer, and real-time state parameters of the label printer are collected; a to-be-printed task is acquired, and the to-be-printed task is disassembled into a subtask unit according to the real-time state parameters; it is judged whether the execution burden of the subtask unit exceeds the local carrying capacity; the execution result of the local agent node is instruction-level fused with the execution result returned by the cloud, a printing instruction sequence is generated, and the printing instruction sequence is issued to the label printer for execution. The application can realize the calculation load balancing in the label printer multi-node access scene, and improve the processing efficiency of the printing task and the system robustness.

This invention relates to the field of intelligent control technology for tunnel boring equipment, and more specifically, to an automated adjustment method and system for the TBM shield tunneling process. It addresses the shortcomings of existing technologies, which cannot construct a dynamic causal knowledge graph rooted in geological conditions, cannot integrate Bayesian structural learning, model-independent meta-learning, hierarchical reinforcement learning, and multi-objective game mechanisms, and struggle to achieve interpretable modeling and adaptive decision-making for the "geology-equipment-control" coupling relationship. This invention constructs a dynamic causal knowledge graph rooted in geological conditions, integrates Bayesian structural learning, model-independent meta-learning, hierarchical reinforcement learning, and multi-objective game mechanisms, and achieves interpretable modeling and adaptive decision-making for the "geology-equipment-control" coupling relationship. It can capture causal changes caused by geological evolution online, support rapid strategy transfer and hierarchical action generation, and output precise control commands that conform to engineering priorities.

The invention discloses an intelligent archive checking method and system based on a thermodynamic diagram, and relates to the technical field of information management, and the system comprises the following modules: a bimodal reference construction module which is used for generating a radio frequency-vision bimodal mapping tensor; the wide-area scanning and exception locking module is used for generating a visual wake-up queue; the high-resolution sampling flow generation module is used for generating an original local image flow; the manifold constraint denoising generation module is used for improving the MaxViT network based on manifold geometric constraints and predicting and generating a real-time thermodynamic diagram; the dual fault discrimination module is used for constructing a differential matrix and logic and dual fault discrimination mechanism; the virtual identification and index increment updating module is used for updating index increment; and the tensor iteration updating module is used for executing iteration updating. According to the method, the limitation that a traditional inventory method is easily interfered, lacks visual verification and is weak in physical topology constraint is overcome, and an efficient solution is provided for intelligent archive inventory.

A strain sensor for a leg brace and a real-time interactive detection system for lower limb rehabilitation are disclosed. The sensor includes a leg brace body and strain gauges mounted on the leg brace body. A double-hole, narrow-beam stress concentration structure is provided at the point of maximum torque near the exoskeleton drive end of the leg brace body. The strain gauges are mounted on the stress concentration structure. The stress concentration structure is designed based on finite element mechanical simulation and is used to concentrate stress and facilitate strain detection. This system, through a self-developed strain gauge leg brace-type human-machine interactive force sensor structure, combined with multi-level noise reduction filtering and high-precision force calculation algorithms, achieves real-time, accurate, and stable detection of human-machine interactive forces during lower limb rehabilitation training. This provides reliable data support for intelligent closed-loop control, active / passive mode switching, and quantitative evaluation of rehabilitation effects of lower limb rehabilitation robots, reducing equipment costs and improving clinical adaptability and widespread adoption.

This invention relates to the field of underground pipeline construction equipment, specifically disclosing a method for determining the pushing stroke in a pipe-rolling machine based on multi-action posture coordination, and the pipe-rolling machine itself. The method includes: establishing a multi-actuator-soil coupled physical model and inverting it to obtain the initial pushing stroke path; real-time acquisition of multi-action posture data during construction, constructing an actual response vector and comparing it with the model's predicted values ​​to form a dynamic error vector; segmenting along the time axis to identify inferior response path segments, applying adaptive perturbations to the control parameters of these inferior segments to generate a new generation path; introducing a feedback weight matrix to weight the error components, iteratively optimizing using the weighted error norm as a criterion until convergence, outputting the optimal pushing stroke path, and executing it in a closed loop. This invention achieves adaptive dynamic optimization of the pushing stroke through deep integration of pre-construction parameter inversion and multi-generation path iteration during construction, significantly improving the multi-cylinder coordination accuracy and overall machine verticality control capability.

This invention proposes a six-axis robot operation control method, device, six-axis robot, and storage medium, relating to the field of robot control technology. The method employs a decoupled hybrid interpolation strategy when the end effector is detected to be traversing a singular region. This strategy involves independently planning the joint motions affecting the posture and combining them with position interpolation. This avoids the joint overspeed, overcurrent, and motion loss-of-control problems caused by traditional pose-based joint interpolation methods near singular points. While ensuring the end effector always moves strictly along the target straight trajectory, it reasonably sacrifices minor posture deviations, improving the robot's motion stability and system robustness when traversing singular regions. This enables continuous and controllable output of joint commands, thereby ensuring equipment safety and process quality, and has significant engineering application value.

The present application belongs to the technical field of fire detection and early warning, and particularly relates to a fire hazard intelligent early warning method based on multi-dimensional feature fusion, which comprises the following steps: first, collecting video frame sequences and multi-dimensional normalized environment features of a monitoring area, using a convolutional neural network to extract a visual feature map, and simultaneously predicting a local water vapor condensation potential index based on a temperature and humidity gradient through a BP neural network; analyzing smoke-like activation intensity and absolute smoke change rate, constructing a cross-modal consistency logic gating mechanism, and generating a spatial attention mask matrix to suppress artifact features. The present application can identify and eliminate high-concentration process water vapor interference, reduce the false alarm rate under complex industrial conditions, and improve the physical consistency and robustness of the fire hazard intelligent early warning.

This invention relates to the field of agricultural automation equipment control technology, and discloses an adaptive control system for fruit bag adsorption and gripping based on pressure feedback. The system includes a pressure sensing unit that collects the contact pressure between the vacuum suction cup and the fruit bag, as well as the air pressure in the vacuum pipeline cavity, combining them to form a two-dimensional pressure state vector. An adsorption force estimation module generates real-time effective adsorption force estimates through static mapping and dynamic correction sub-models. An adsorption force adjustment and control module uses a preset target adsorption force as a reference and outputs a control signal through a feedback control algorithm. A parameter management module adapts to different fruit bag materials, providing model parameters, gain coefficients, and control parameters. An execution and protection module drives the vacuum generator and auxiliary gripper to perform actions and generates logs. This invention uses two-dimensional pressure sensing to compensate for the shortcomings of single-dimensional air pressure sensing, effectively improving the reliability of fruit bag adsorption and gripping.

The application discloses a track following dynamic correction method and system for a caterpillar electric chassis, and belongs to the field of intelligent control of agricultural robots. The method is aimed at the yellow sand substrate environment of a sunlight greenhouse. Firstly, a reference trajectory for inter-ridge operation is constructed, and a turning area is calibrated. Multi-source data are collected through an encoder, an IMU, a laser radar and a visual sensor, and a chassis pose, a deviation and a caterpillar slip state are fused and estimated. A prediction control model with time sequence embedding and dimension expansion is constructed, control weights and constraint conditions are adjusted online in combination with working condition identification. A caterpillar speed correction amount is solved through rolling optimization, and is output after being limited in amplitude. Meanwhile, disturbance estimation is updated according to an execution result, so that closed-loop control is realized. The application effectively improves the trajectory tracking precision and robustness under a soft ground, suppresses deviation accumulation during turning, and guarantees stable operation of the chassis under complex working conditions.

The application provides an AmBC-NOMA communication resource joint optimization method for a high-mobility V2X scene, and relates to the technical field of communication resource joint optimization. In view of the channel dynamic change caused by high-speed movement of vehicles and the influence of the channel dynamic change on short packet communication reliability and rate performance, the application combines reflective communication and non-orthogonal multiple access technology, fully considers the time-varying characteristics of the channel, and jointly optimizes the reflection coefficient, the power distribution coefficient and the block length resource allocation in the system, so that the communication rate performance and resource utilization efficiency under the requirements of short time delay and low error rate are improved. Through the method, the short packet transmission rate of the system can be significantly improved under the conditions of limited time delay and channel uncertainty, and the practical application requirements of the future vehicle networking system for efficient, low time delay, high reliable short packet communication are met.

This invention relates to the fields of computer vision and action recognition, and discloses a method and device for multimodal sports motion detection. The method includes: selecting cyclic action segments from an input video; extracting video frames and key nodes, whereby key nodes include skeletal joints and specific key points defined based on the motion type; obtaining temporal feature representations of the video frames through a frame encoder and a temporal encoder; converting the key nodes into a graph structure, calculating learnable embedding vectors through in-degree and out-degree calculations, and determining the bias term based on the shortest path distance between nodes; inputting the temporal feature representation and the learnable embedding vectors into a Transformer encoder, and obtaining cross-modal enhanced features through a self-attention mechanism adjusted by the bias term; and outputting a standardized action classification result based on the cross-modal enhanced features and spatiotemporal topological features. This invention effectively improves the accuracy, robustness, and generalization ability of recognizing subtle and irregular actions through multimodal feature fusion.

The invention provides a sequential circuit node feature representation system construction method, which comprises the following steps: S1, acquiring sequential circuit RTL codes, synthesizing into a gate-level netlist, and describing the text of each node; s2, obtaining a fine-tuning large language model for generating high-dimensional semantic embedding of each node in the gate-level netlist; s3, constructing a node feature fusion module for enhanced feature representation of each node in the gate-level netlist; s4, constructing a circuit construction module, which is used for constructing a circuit diagram taking nodes in the gate-level netlist as nodes and directed edges in the gate-level netlist as connecting lines, and representing node features of the nodes in the circuit diagram by enhanced feature representation of the nodes; s5, acquiring a graph neural network, and performing end-to-end training on the graph neural network by adopting the circuit diagram in the step S4 until convergence; and S6, constructing a sequential circuit node feature representation system by using a circuit synthesis tool, the fine-tuned large language model, a node feature fusion module, a circuit construction module and the converged graph neural network.

This invention discloses an intelligent beam tracking method and apparatus for mobile environments, relating to the field of wireless network technology. The method includes: S1 constructing a beam tracking model; S2 training and optimizing the leader and follower agents; S3 acquiring MCS statistics and SSB scan results during the k-th window in real time; S4 constructing a leader state vector; S5 predicting the final leader action; S6 generating a candidate action set for followers; S7 generating follower state vectors; S8 predicting the final follower action; S9 constructing the actual transmission beam; S10 adjusting the MCS to transmit data using the transmission beam; and S11 determining whether it is time to enter the next SSB scan cycle. If yes, return to S3; otherwise, return to S9. This method significantly improves the effective user throughput and system robustness of 5G millimeter-wave systems in mobile scenarios.

The invention discloses an FMCW laser ranging light source spectrum degradation suppression method based on strategy online deployment. Belongs to the laser radar field. The objective of the invention is to solve the technical problems of spectrum degradation and measurement precision reduction caused by frequency modulation nonlinearity of an FMCW laser ranging light source. Comprising the following steps: building an FMCW laser nonlinear dynamic system platform, and constructing a deep learning model to simulate a dynamic environment; designing a nonlinear correction process as a Markov decision process, and training a reinforcement learning agent by using a TD3 algorithm; carrying out lightweight processing on an Actor network in the policy network, wherein the lightweight processing comprises pruning and quantization operations so as to reduce the complexity of the model; a lightweight strategy network is deployed on a hardware platform, strategy network forward reasoning is achieved on an FPGA, the strategy network is cascaded with a state extraction module and an action output module to form a closed-loop correction system, and dynamic modulation current optimization is achieved. According to the invention, frequency spectrum degradation caused by frequency modulation nonlinearity can be inhibited on line, and the measurement precision is improved.

The application provides a method for realizing high-precision underwater integrated navigation by fusing AUV kinetic velocity and single-beacon acoustic data, and belongs to the technical field of underwater navigation and positioning. The method comprises the following steps: estimating the speed against water in the carrier coordinate system by using the propeller rotating speed and the motor current, decomposing the speed against water to the navigation coordinate system by fusing the SINS attitude, obtaining the three-dimensional motion velocity of the AUV, and obtaining the preliminary position based on the dead reckoning; adaptively selecting the navigation strategy according to the operation time length and acoustic availability: the pure kinetic dead reckoning is adopted for the medium-short time task, and the single-beacon distance information is introduced for the long time operation, the navigation position after correction is obtained by periodically constraining the dead reckoning result through the geometric projection filtering algorithm of resetting the starting point. Compared with a single data source, the application significantly improves the navigation precision and reliability, and effectively solves the technical problem that the application of the acoustic sensor is limited in the deep sea environment.