A spatiotemporal embodied intelligent system integrating BeiDou and remote sensing

Abstract

This invention provides a spatiotemporal embodied intelligent system integrating BeiDou and remote sensing, belonging to the interdisciplinary research field of Earth spatiotemporal analysis and decision-making and artificial intelligence technologies. It includes an intelligent sensing module, a data fusion processing module, an intelligent decision-making module, and an execution and feedback module. The intelligent sensing module acquires positioning and remote sensing data; the data fusion processing module fuses the positioning and remote sensing data acquired by the intelligent sensing module to obtain spatiotemporal data of the target area; the intelligent decision-making module generates decision instructions for the target area based on the spatiotemporal data; and the execution and feedback module executes corresponding operations according to the decision instructions and provides feedback results. Furthermore, it enables intelligent equipment terminals such as unmanned aerial vehicles, unmanned agricultural machinery, and unmanned ships to execute corresponding operations based on the decision instructions, and continuously learns, evolves, and optimizes the feedback results.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary research field of Earth spatiotemporal analysis and decision-making and artificial intelligence technology. In particular, it relates to a spatiotemporal embodied intelligent system that integrates BeiDou and remote sensing. Background Technology

[0002] With the rapid development of the BeiDou Navigation Satellite System and Earth Observation Systems, these two systems, as the dual core infrastructures for geographic information acquisition, are being widely applied in various fields such as land surveys, traffic monitoring, disaster early warning, and smart agriculture. The BeiDou system possesses centimeter-level high-precision positioning and nanosecond-level time synchronization capabilities, playing a crucial role in the dynamic tracking of point and line targets in areas such as railway safety, urban traffic, and unmanned operations. Meanwhile, high-resolution remote sensing, through its sub-meter spatial resolution and multispectral, multi-temporal imaging capabilities, provides comprehensive areal feature information support for geological disaster monitoring, resource surveys, and ecological assessments.

[0003] Although BeiDou and remote sensing technologies are clearly complementary in terms of spatiotemporal resolution, target type, and information acquisition methods, existing technologies still face the following major challenges in integrated applications:

[0004] (1) Insufficient integration depth and limited level of collaboration: Currently, most "BeiDou + remote sensing" applications are still at the level of functional superposition and information assistance, lacking a systematic and closed-loop technical system. BeiDou provides positioning trajectory, and remote sensing provides environmental images. There is a lack of unified data model and intelligent decision-making mechanism between the two, which makes it difficult to support the high reliability requirements of autonomous system operation in complex environments;

[0005] (2) Insufficient intelligent processing capabilities and a broken perception chain: Traditional remote sensing focuses on image acquisition and manual interpretation, while BeiDou data relies on background rule processing. Overall, it is "strong in perception but weak in cognition" and lacks intelligent agent systems with reasoning, judgment and strategy generation capabilities. Especially in the context of multi-source heterogeneous data, existing models are difficult to achieve semantic understanding, real-time response and dynamic updates, which seriously restricts the widespread deployment of automated operation systems;

[0006] (3) The execution process is disconnected and the perception-decision-execution loop is not closed: At present, the "Beidou + remote sensing" system is mostly used for monitoring or static identification. It has not yet achieved deep integration with physical intelligent equipment such as drones, unmanned vehicles, and agricultural robots. It lacks an embodied intelligent system that can transform "data understanding" into "physical operation", which makes it difficult for the perception results to directly guide the execution. The operation chain is not closed, and both response efficiency and operation accuracy are affected.

[0007] (4) Lack of system evolution capability and lack of self-learning characteristics: Existing systems often rely on static rules and human experience to build models, lacking dynamic feedback mechanisms and continuous learning capabilities, making it difficult to adapt to the operational needs of different regions, complex environments or changing conditions. In complex scenarios such as heterogeneous farmland, mountainous geological disaster areas or marine buoy environments, existing technologies have poor adaptability and weak promotion.

[0008] In summary, the integration of BeiDou and remote sensing technologies in practical applications has not yet formed a closed-loop system from high-precision positioning and environmental perception to intelligent decision-making and autonomous execution, and lacks a complete technical architecture with the five-in-one functions of "perception-cognition-decision-execution-learning". Summary of the Invention

[0009] To address the above technical problems, this invention provides a spatiotemporal embodied intelligent system that integrates BeiDou and remote sensing, with the specific technical solution as follows:

[0010] A spatiotemporal embodied intelligent system integrating BeiDou and remote sensing includes an intelligent sensing module, a data fusion processing module, an intelligent decision-making module, and an execution and feedback module. The intelligent sensing module acquires positioning data and remote sensing data; the data fusion processing module fuses the positioning and remote sensing data acquired by the intelligent sensing module to obtain spatiotemporal data of the target area; the intelligent decision-making module generates decision instructions for the target area based on the spatiotemporal data; and the execution and feedback module executes corresponding operations according to the decision instructions and provides feedback on the results.

[0011] The present invention has the following beneficial effects:

[0012] This invention deeply integrates spatiotemporal intelligence and embodied intelligence to construct a full-link intelligent closed-loop system. For the first time, this invention systematically integrates "BeiDou spatiotemporal intelligence" and "embodied intelligence," constructing a five-in-one closed-loop technology chain of "perception-cognition-decision-execution-learning" to achieve autonomous control throughout the entire process from environmental data acquisition to the physical operation of the intelligent agent. This collaborative approach breaks through the traditional isolated operation mode of GIS systems, remote sensing analysis platforms, and individual intelligent equipment, establishing a multimodal, highly perceptive, and highly executable intelligent operation system under a unified spatiotemporal benchmark.

[0013] This invention achieves deep coupling of the original layer with the BeiDou positioning and navigation system and remote sensing at the signal layer, breaking through the bottleneck of perception accuracy: At the signal layer, this system achieves collaborative analysis of parameters such as delay, Doppler shift, and power distribution through joint modeling of BeiDou multi-band GNSS signals and original echo signals from the remote sensing platform. This is used for accurate inversion of environmental factors such as sea surface roughness, soil moisture, and terrain height, effectively improving the positioning reliability and perception capability of BeiDou in complex environments.

[0014] This invention introduces a "remote sensing-assisted BeiDou positioning" mechanism to improve the robustness of the navigation system: by integrating remote sensing image feature extraction with BeiDou trajectory points, visual-assisted navigation and map matching reverse correction are achieved, which greatly enhances the path continuity and accuracy of the system in areas with signal obstruction or loss of lock, and is especially suitable for continuous navigation tasks in complex environments such as tunnels, forest areas, and mining areas.

[0015] This invention improves decision-making credibility by using a remote sensing inversion model based on the Jacobian matrix and the error propagation law. The system introduces a "remote sensing inversion mechanism with control area network adjustment" at the decision-making level. By constructing a time evolution model of the inversion factors and combining it with a high-precision control network constructed from sampling points, the Jacobian matrix is ​​introduced to perform accuracy extrapolation and error estimation. This ensures that the generated prescription maps and control commands have quantitative verification capabilities and theoretical credibility, solving the problems of existing remote sensing decision results being uninterpretable and unverifiable.

[0016] This invention opens up the "data-behavior" transformation channel through multi-terminal collaboration of intelligent agent execution modules: This invention supports the execution capabilities of embodied intelligent agents on multiple platforms such as air, ground, underwater, and indoors. Through the task parsing engine, decision instructions are transformed into "atomic" execution actions, realizing unmanned execution of tasks such as variable fertilization in agriculture, precision seeding, mine inspection, and indoor handling, truly opening up a closed loop channel from digital information to physical behavior.

[0017] This invention enables the learning layer to support self-evolution, achieving model self-optimization and scenario adaptation: during operation, the system drives model parameter updates and rule base optimization through feedback links, and combines federated learning and reinforcement learning strategies to achieve cross-device and cross-regional generalization training capabilities. It has the advantages of strong scenario adaptability, model transferability, and system evolution, and supports the continuous and efficient operation of intelligent agent clusters in changing environments. Attached Figure Description

[0018] Figure 1 A structural diagram of the BeiDou spatiotemporal embodied intelligent model;

[0019] Figure 2 This is a diagram of the architecture of a Beidou-based smart unmanned farm, representing a specific embodiment of the present invention.

[0020] Figure 3 This is a diagram illustrating the deployment of equipment on an open farm site according to a specific embodiment of the present invention;

[0021] Figure 4 These are some agricultural parameters from a specific embodiment of the present invention;

[0022] Figure 5 This is a prescription diagram of a Beidou smart unmanned farm, a specific embodiment of the present invention.

[0023] Figure 6 This is a diagram illustrating the unmanned farm operation of the present invention. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0025] To achieve the above objectives, the present invention adopts the following technical solution.

[0026] This invention provides a spatiotemporal embodied intelligent system integrating BeiDou and remote sensing, comprising an intelligent sensing module, a data fusion processing module, an intelligent decision-making module, and an execution and feedback module. The intelligent sensing module acquires positioning data and combines it with remote sensing data. The data fusion processing module fuses the positioning and remote sensing data to obtain spatiotemporal data of the target area. The intelligent decision-making module generates decision instructions for the target area based on the fused spatiotemporal data. The execution and feedback module executes corresponding operations according to the decision instructions and provides feedback on the results.

[0027] Earth observation and navigation technologies, through precise understanding of the Earth's environment in terms of time and space, can only generate new productive forces by transforming the physical world through human social actions, thus serving the harmonious coexistence of humanity and nature. Based on this, this invention addresses the shortcomings of artificial intelligence in geosciences, spatiotemporal intelligence, and geographic intelligence research at the behavioral execution level by proposing the concept of spatiotemporal embodied intelligence. It defines the connotation and extension of the concept of spatiotemporal embodied intelligence and studies its implementation methods and technical approaches.

[0028] The connotation and extension of the concept of spatiotemporal embodied intelligence: The connotation of the concept of spatiotemporal embodied intelligence refers to the science and technology by which physical entities use high-precision navigation, positioning and timing, high-resolution remote sensing and artificial intelligence methods to conduct high-precision perception, accurate cognition, intelligent decision-making, scientific behavior and autonomous optimization learning of natural environmental elements and spatiotemporal characteristics of human society. It is an advanced stage of artificial intelligence development. As the extension of the concept of spatiotemporal embodied intelligence, Beidou spatiotemporal embodied intelligence is a physical entity equipment based on the Beidou satellite navigation system and high-resolution remote sensing integration technology, integrating geographic spatiotemporal intelligence and embodied intelligence technology to achieve precise execution control of natural environment and human social objects. It is a frontier field of multidisciplinary intersection of aerospace information, physical cognition, embodied intelligence and other disciplines and a new round of technological revolution. Its core technologies include: (1) deep integration of centimeter-level Beidou spatiotemporal reference and multimodal environmental perception; (2) spatiotemporal coupling analysis of multi-source heterogeneous data (satellite remote sensing, positioning trajectory, environmental sensing); (3) dynamic decision optimization driven by geoscience knowledge; (4) real-time generation and adaptive control of autonomous behavior in complex scenarios. This system achieves full-process automation from large-scale environmental cognition to refined operation execution through the organic combination of an integrated air-space-ground sensing network, a cloud-based collaborative computing architecture, and a dynamic response mechanism for intelligent agents.

[0029] The spatiotemporal embodied intelligence system, combining BeiDou and remote sensing, possesses spatiotemporal cognitive capabilities and embodied intelligence, exhibiting the characteristics of "perception-cognition-execution-learning," such as... Figure 1 As shown. These correspond to the components of human intelligence, including the five sensory systems, the left brain, the right brain, the heart, and the limbs.

[0030] The intelligent sensing module is equivalent to the five senses perception system. It obtains positioning data through the BeiDou satellite navigation system, remote sensing image data through satellite remote sensing system and airborne remote sensing system, and visual images, navigation trajectories and other data through ground Internet of Things system. It is used for space-air-ground collaborative observation to achieve accurate perception of the time and space dimensions of the Earth's environment and obtain spatiotemporal data of the target area.

[0031] The data fusion processing module, which is the implementation of the BeiDou Navigation and Remote Sensing Integrated Service Platform, is like the heart. It integrates the multimodal sensing module to form data products that fuse BeiDou navigation and remote sensing.

[0032] The intelligent decision-making module is analogous to the human brain, comprising a positioning and navigation control submodule and a spatial cognition and decision-making submodule. The positioning and navigation control submodule, based on positioning data from the multi-band BeiDou satellite navigation system and ground control point information, performs geometric correction and image registration on remote sensing images. It utilizes an inertial navigation-satellite navigation combined positioning algorithm to eliminate inertial drift errors, generating 3D pose data of the target object, achieving precise navigation, trajectory reconstruction, and path control in dynamic scenarios. The spatial cognition and decision-making submodule is analogous to the right brain. Taking remote sensing images, visual images, and navigation trajectories provided by the intelligent perception module as input, it comprehensively utilizes algorithms such as deep convolutional neural networks to achieve land cover classification, state recognition, environmental understanding, and risk assessment of the target area. Furthermore, based on reinforcement learning algorithms, it generates task execution strategies and operation path planning results with context-responsive capabilities. Outputs include a decision prescription map, target priority ranking, and operation instruction set, realizing a closed-loop control of "perception-cognition-reasoning-decision," directly driving the system's intelligent execution and dynamic scheduling.

[0033] The execution and feedback modules correspond to the four limbs of a human. They include: an aerial operation unit (equivalent to the left hand), comprising aerial remote sensing aircraft and drones, enabling variable-time aerial operations based on decision prescription maps; a ground operation unit (equivalent to the right hand), comprising unmanned agricultural machinery, implements, and unmanned vessels for ground production operations, enabling ground and water-based production and operations based on decision prescription maps; an indoor operation unit (equivalent to the left foot), utilizing unmanned vehicles and unmanned manufacturing equipment to achieve intelligent indoor operations and manufacturing based on decision prescription maps and operational instructions; and an underwater operation unit (equivalent to the right foot), comprising underwater robots, enabling precise underwater operations based on decision prescription maps.

[0034] The data fusion processing module is used to construct a spatiotemporal data cube, achieving spatiotemporal alignment and reconstruction of multi-source data to form standardized, traceable data products. The goal is to utilize positioning data provided by the BeiDou Navigation Satellite System, including positioning and timing information, combined with data from satellite remote sensing systems, to achieve geometric correction, spatiotemporal benchmark unification, and high-precision alignment of remote sensing imagery, including:

[0035] The specific algorithm for geometric correction of remote sensing imagery involves using a polynomial geometric correction model based on ground control points (GCPs) to map the image coordinate system (rows and columns) of the remote sensing imagery to the geographic coordinate system (latitude, longitude, or projected coordinates). The process is as follows:

[0036] Selecting GCPs (Ground Control Points): High-precision control points (with an accuracy of up to centimeter level) measured by the BeiDou satellite navigation system positioning terminal are used as GCPs and are evenly distributed in the remote sensing image area.

[0037] Construct an affine or higher-order polynomial transformation model: Use first- to third-order polynomial models to establish the mapping relationship between remote sensing image coordinates (x, y) and ground coordinates (X, Y), as follows:

[0038] ;

[0039] ;

[0040] , These are the model coefficients.

[0041] By constructing a set of observation equations and solving the model coefficients using least squares, the fitting error of the control point coordinates is minimized.

[0042] Using the calculated transformation relationship, geometric correction is performed on remote sensing images, and orthorectified remote sensing images are generated through bilinear interpolation or cubic convolution resampling. This method performs stably in urban areas or scenes with large terrain undulations, and can control the registration accuracy of high-resolution remote sensing images to the sub-pixel level (≤1 pixel).

[0043] The unification of spatiotemporal references specifically includes: achieving spatiotemporal reference unification through high-precision time synchronization information and coordinate reference data, including time synchronization and spatial alignment.

[0044] Time synchronization: Using timing information from the BeiDou Navigation Satellite System to correct time labels for various remote sensing data.

[0045] Spatial registration: Using the control points measured by the BeiDou satellite navigation system positioning terminal as a reference, the coordinate system of different remote sensing data is unified, and high-precision spatial alignment between multi-source images is achieved through collinearity condition equations or image pyramid registration algorithms.

[0046] Remote sensing-assisted BeiDou positioning correction includes: utilizing existing remote sensing image products to acquire image features, thereby improving the robustness of BeiDou positioning, especially suitable for environments with severe signal obstruction. The technical solution is as follows:

[0047] ;

[0048] The BeiDou-calculated position is iteratively corrected using an optimizer (such as Levenberg-Marquardt). This indicates the acquisition of image features from existing remote sensing image products. This represents the matching target features extracted from existing navigation maps or prior models.

[0049] The spatial cognition and decision-making submodule includes an intelligent decision-making information extraction module, an intelligent decision-making knowledge fusion module, and an intelligent decision-making generation module.

[0050] The intelligent decision-making information extraction module is used to achieve intelligent inversion and dynamic decision-making of remote sensing data with the support of high-precision data, providing the system with interpretable, verifiable, and evolvable intelligent control information. Its functions include: ensuring remote sensing inversion accuracy, constructing spatiotemporal control points, solving residual minimum optimization problems, theoretical accuracy analysis, and model reinforcement learning. The intelligent decision-making information extraction module adopts a remote sensing inversion method based on control area network adjustment to solve the problems of "unreliability" and "lack of evolution" in remote sensing inversion results in complex environments. This method comprehensively incorporates remote sensing observation inversion models and environmental dynamic models, and uses ground control sampling points for constraints and optimization, specifically including the following steps:

[0051] Step 1: Modeling the unreliability of remote sensing signals; including:

[0052] When remote sensing signals (such as spectral radiance and backscattering coefficient) are used to retrieve surface physical quantities (such as soil moisture, leaf area index, and surface temperature), the uncertainty and accuracy may be uncontrollable due to factors such as atmospheric disturbance, terrain shading, and sensor noise.

[0053] Therefore, a dynamic model of the evolution of the inversion factor over time is constructed, and its general mathematical expression is as follows:

[0054] ;

[0055] in, : indicates the first A physical quantity to be inverted in time The measured physical quantity value; : Represents environmental factors (such as temperature, humidity, vegetation index); : Represents the dynamic evolution function between inverted objects, characterizing their physical evolution laws.

[0056] Step 2: Construct a spatiotemporal remote sensing control network; including:

[0057] To enhance the model's constraint, a spatiotemporal control network framework covering the target area was further constructed. This framework utilizes a multi-platform measurement system of a terrestrial IoT system for control point sampling. The terrestrial IoT system integrates the BeiDou Navigation Satellite System with various types of ground / low-altitude / water surface remote sensing and sensor equipment, including BeiDou satellite navigation system positioning terminals, lidar, UAV remote sensing platforms, buoys, etc. Each control point provides timestamped measured physical quantity data, forming a control point set.

[0058] ;

[0059] in, : Represents the longitude, latitude, and observation time of the control point; the control points form a three-dimensional sparse point cloud, used to correct the spatial and temporal consistency of remote sensing inversion results.

[0060] Step 3: Construct a remote sensing inversion model based on a spatiotemporal remote sensing control network, including:

[0061] A remote sensing observation inversion model is constructed using a dynamic model, introducing two residual terms:

[0062] ;

[0063] in: : Indicates the inversion value to be estimated; : Indicates that through remote sensing imagery The estimated value obtained from the inversion; : Represents the combination of the previous time value and the set of control points The derived dynamic estimate; : Represents the residual between the remote sensing estimate and the dynamic model estimate.

[0064] Construct the objective function: ;

[0065] in, This represents the adjustment coefficient, which controls the weight between the remote sensing estimate and the dynamic model estimate.

[0066] Step 4, Jacobian matrix linearization and error propagation, including: [The text abruptly ends here, likely due to an incomplete sentence or a formatting error.] and Perform a first-order Taylor expansion to construct the Jacobian matrix J. Furthermore, the accuracy evaluation formula is as follows:

[0067] ;

[0068] in, The input is the covariance matrix of the observed data. This is calculated... The theoretical accuracy (i.e., standard deviation σ) can be used to perform confidence analysis on the inverted values ​​and serve as a constraint for intelligent decision-making.

[0069] The model self-evolution mechanism uses inversion accuracy and control error as loss functions. and Reinforcement learning is used to train the network so that it can be continuously updated and optimized during actual operation: the input is the current observation data, historical evolution data, and control point observation values; the output is the inversion results with controllable accuracy; the network parameters are dynamically adjusted according to the inversion error and theoretical accuracy; and multiple rounds of training are carried out by combining supervised learning and reinforcement learning mechanisms.

[0070] The intelligent decision-making knowledge fusion module embeds a geographic knowledge graph construction, a rule-based reasoning engine, and a knowledge update mechanism. The geographic knowledge graph construction function extracts "feature-attribute-relationship" triples from historical remote sensing imagery, sensor data, and manually labeled results to form semantic environmental information. The reasoning engine, based on graph neural networks and an ontology rule system, supports knowledge retrieval, causal analysis, and logical reasoning. Example: Based on anomaly in leaf area index + recent lack of precipitation, it infers "potential drought." The knowledge self-update mechanism automatically corrects knowledge edge weights and adds new nodes based on feedback from the learning layer, maintaining the dynamic evolution of the knowledge graph.

[0071] The intelligent decision generation module is used to extract rules and feature data, and achieves intelligent control output through model library management, strategy generation and decision evaluation.

[0072] The model library management integrates various task models based on deep neural networks, reinforcement learning, and Bayesian inference, and dynamically selects the optimal model according to the scenario.

[0073] Strategy generation includes: inputting remote sensing imagery of the target area, navigation trajectory, and semantic tags from a knowledge graph; outputting: a variable prescription map and a trajectory planning path. Example: Generating a triplet instruction table of "plot-operation type-operation parameters" for agricultural operations.

[0074] The decision evaluation includes: calculating multi-strategy scores based on the set objective function (minimizing operating costs, minimizing time, maximizing output, etc.), and using a multi-objective optimization algorithm to select the optimal decision scheme.

[0075] The execution and feedback module includes four sub-modules: decision execution, execution feedback collection, error analysis, model update, and knowledge transfer.

[0076] The decision execution submodule transforms the output high-level operational instructions into low-level execution instructions for specific platforms. Through task decomposition, equipment scheduling, and multi-agent collaborative control modules, it enables the coordinated operation of the embodied intelligent terminals. Task decomposition includes converting the global operational plan into deployable "atomic task units," such as breaking down "variable spraying" into "positioning flight → spray dosage setting → action triggering → flight path update." Equipment scheduling includes optimizing the allocation of heterogeneous equipment based on the current available resource status, employing a graph-optimized resource matching algorithm. Collaborative control includes supporting multi-platform collaborative operations, using a distributed multi-agent control protocol to synchronize operational plans and avoid conflicts.

[0077] The execution feedback acquisition submodule integrates remote sensing inversion monitoring values, execution logs of each work unit, and sensor status data to establish a multi-dimensional data feedback mechanism for the operation process. The error analysis submodule compares the deviation between the actual operation results and the expected operation instructions to construct a residual model.

[0078] ;

[0079] The influence weight of each stage is analyzed using the error propagation formula. : Represents the residual or execution error, which is the difference between the actual execution result and the expected execution target; : Indicates the actual execution result; : Indicates the planned value or target value.

[0080] The model update submodule updates the model parameters based on the residual size and distribution using federated learning and incremental training strategies to avoid overfitting and catastrophic forgetting.

[0081] The knowledge transfer submodule transfers the model structure and rule set trained in a certain region or device to a new scene, improving the adaptation efficiency and accuracy of the new scene through transfer learning strategies. It can maintain high-performance operation capabilities in different environments, different task objects, and different device platforms.

[0082] The present invention further provides specific embodiments, specifically a case study of Beidou smart unmanned farm application.

[0083] The Beidou Smart Unmanned Farm, by integrating Beidou high-precision positioning and high-resolution remote sensing monitoring, achieves closed-loop intelligent management of the agricultural production process. As a model of modern agricultural intelligence, it possesses the characteristics of an embodied intelligent agent, exhibiting environmental perception, autonomous decision-making and execution, and iterative optimization. It is a typical embodied intelligent agent. (Example: Beidou Smart Unmanned Farm) Figure 2As shown, its core structure and operation process include six key links, including: a multi-band signal receiving module relying on the Beidou satellite navigation system to achieve centimeter-level real-time dynamic precision positioning; by integrating high-resolution satellite data and UAV multispectral and hyperspectral remote sensing data, providing a high-precision data foundation for subsequent environmental perception and spatial analysis; merging multi-source heterogeneous data from Beidou satellite positioning, high-resolution remote sensing, UAV aerial photography, and ground sensors (such as soil moisture and weather stations) to construct dynamic digital models of farm environment, agricultural machinery operation, and crop growth status; using remote sensing inversion models and machine learning methods to generate precise agricultural information products such as crop growth parameters, soil nutrient distribution, and pest and disease risks; integrating agricultural expert experience and agronomic knowledge such as crop physiological and ecological models into the decision support system, and establishing a dynamic simulation model of crop growth by combining historical and real-time data; the intelligent agent, through the variable prescription decision generation module, to achieve precise variable management based on spatial heterogeneity and differences in crop water and fertilizer requirements, automatically generating differentiated fertilization, irrigation, and pest and disease control plans, effectively improving resource utilization efficiency and agricultural product quality; and continuous iterative optimization based on operational feedback data and environmental change information. Through machine learning and reinforcement learning techniques, intelligent agents learn on their own and dynamically adjust their perception models and decision-making strategies to form a closed-loop optimization of the production process, thereby promoting the sustainable development of smart agriculture.

[0084] Three-dimensional data acquisition

[0085] The Beidou Smart Unmanned Farm System integrates three types of sensing methods—satellite remote sensing, UAV low-altitude remote sensing, and ground-based IoT sensors—to construct a comprehensive, high-precision, and fast-responding three-dimensional agricultural information perception system. This integrated air-space-ground multi-source information perception network follows a design logic of "complementary spatiotemporal resolution, mutual verification of data accuracy, and interlocking execution feedback." It achieves dynamic interaction and value-added fusion between satellite-led "strategic-level scanning," UAV-driven "tactical-level detailed investigation," and ground-based sensor-supported "cellular-level feedback." This forms the foundation for precise agricultural machinery operations, real-time crop status monitoring, and variable management decision support. It not only provides multi-scale, multi-dimensional information on crop physiological states and soil environment but also offers powerful data support for intelligent regulation and early warning throughout the entire agricultural production process.

[0086] At the "space" level, the BeiDou Navigation Satellite System and Gaofen Remote Sensing Satellites form a collaborative division of labor of "positioning + monitoring": the BeiDou system, relying on real-time dynamic carrier phase differential technology, achieves centimeter-level positioning accuracy, providing a highly reliable foundation for the path planning, operation execution, and trajectory correction of unmanned agricultural machinery; while the Gaofen Remote Sensing Satellites have sub-meter-level panchromatic and meter-level multispectral observation capabilities, and can collect spectral, structural, and geometric feature data of crops in large areas of farmland at high frequency. Through panchromatic-multispectral image fusion and remote sensing inversion algorithm processing, key physiological information such as crop chlorophyll content, water stress level, and signs of pests and diseases is extracted, and deeply linked with the BeiDou positioning system, ultimately forming a complete operational closed loop from problem identification to precise spatial positioning.

[0087] At the airspace level, unmanned aerial vehicle (UAV) systems focus on high-precision, highly flexible near-field perception tasks. Compared to satellite systems, UAVs offer significant advantages such as centimeter-level perception accuracy, hourly response speed, and low-cost deployment, making them particularly suitable for the needs of precision agriculture scenarios, including detailed crop diagnosis at the field scale and rapid response to sudden agricultural emergencies. When a high-resolution satellite detects an anomaly, the UAV can quickly take off and, using its onboard payloads such as hyperspectral cameras, infrared thermal imagers, SAR (synthetic aperture radar), and LiDAR (lidar radar), conduct detailed verification of the target area. Figure 3 As shown in A, this type of dedicated agricultural monitoring drone is equipped with multiple types of payloads, including multispectral (8 channels: red, green, blue, 2 red edges, 2 near-infrared), thermal infrared, SAR, and LiDAR, which can simultaneously collect multi-dimensional data and complete the inversion analysis of agricultural parameters.

[0088] At the "geographic" level, the sensing system primarily consists of field IoT sensors, including soil moisture, conductivity, temperature, and pH monitoring equipment, as well as meteorological microenvironment stations. These sensors use high-frequency data collection at the second to minute level to construct a continuous monitoring chain from the root layer to the canopy to the atmospheric boundary layer, becoming the farm's "sensory nerve endings," such as... Figure 3 Figures B, C, and D are shown. Figure B shows the integrated weather station, Figure C shows the soil stratification multi-parameter detector (battery-powered), and Figure D shows the soil stratification multi-parameter detector (solar-powered). These devices, combined with remote sensing capabilities, achieve soil water / humidity / temperature / fertility error ≤5%, growth and yield prediction error ≤15%, and meteorological data error ≤6%, and support 7-day forecasts.

[0089] Multi-source sensing data fusion

[0090] To comprehensively perceive the status of farmland crops and the environment, the BeiDou Smart Unmanned Farm System integrates multi-source heterogeneous data, including satellite remote sensing (such as Gaofen-2 and Gaofen-6), UAV multispectral imagery, hyperspectral imagery, BeiDou high-precision positioning data, ground sensor observation data (such as soil moisture, temperature, and conductivity), agricultural machinery sensor data, and operation log information. This module realizes the entire process of information processing from data acquisition to information reconstruction and from raw observation to fusion representation, significantly improving the accuracy and continuity of agricultural machinery operation and crop status perception, and constructing a high spatiotemporal resolution, dynamically updatable farmland perception information model.

[0091] To address the challenges of more dynamic and complex environments in agricultural settings, which demand higher standards for map timeliness, multi-source data, and adaptation to agricultural characteristics, this paper proposes a method for constructing full-scene agricultural maps based on multi-source sensing information. This method integrates UAV remote sensing imagery, digital elevation models (DEMs), and ground-based lidar point cloud data. Unified modeling and multi-dimensional information integration are performed using geographic information platforms such as ArcGIS to create a semantic map that supports agricultural machinery navigation and environmental understanding. Furthermore, the system incorporates a perception data-driven behavioral intelligence algorithm to achieve dynamic obstacle avoidance and trajectory planning in complex scenarios such as farm roads and fields. A local planner based on the Frenet coordinate system can respond in real-time to information on obstacles ahead, oncoming vehicles, and terrain boundaries. Combined with target detection results provided by the perception layer, it dynamically adjusts the navigation path, improving the flexibility and safety of the operation process.

[0092] To address the robust perception requirements of unmanned agricultural machinery operating in various scenarios and under complex weather conditions, a multi-sensor perception system integrating LiDAR, millimeter-wave radar, and monocular vision was constructed. Through time synchronization and coordinate alignment, various detection results were unified into the BEV space, and target-level fusion was achieved using intersection-union-ratio (IUGR) matching, improving obstacle recognition accuracy and dynamic adaptability. A lightweight LiDAR point cloud perception algorithm was designed, combining columnar coding and 2D feature extraction, balancing edge deployment efficiency and 3D detection accuracy. Experiments show that the system possesses good real-time performance and environmental adaptability, significantly improving the perception stability and operational safety of unmanned agricultural machinery in unstructured field scenarios.

[0093] To address data heterogeneity and resolution differences, a multi-level algorithm integrating physical model priors and deep features is introduced to improve the consistency and discriminative power of expressions for tasks such as crop condition identification, growth monitoring, and prescription generation. By utilizing BeiDou time synchronization and the WGS84 coordinate system, temporal and spatial alignment of remote sensing satellites, UAVs, and ground equipment is achieved, constructing a unified spatiotemporal reference framework for the system. Experiments show that this fusion strategy exhibits good adaptability and stability in complex agricultural environments, providing reliable support for unmanned farm perception systems.

[0094] Precise monitoring of agricultural parameters

[0095] This module relies on satellite remote sensing, drone monitoring and ground-based IoT equipment to construct a multi-source sensor data fusion system through standardized processing and fusion of multi-scale and multi-modal data. It completes the three-dimensional real-time inversion of key agricultural parameters such as soil (moisture, temperature, nutrients) and crops (emergence rate, growth, nitrogen content, etc.), ultimately supporting precision agricultural production decisions and promoting the integrated development of farm sensing networks and agricultural diagnosis systems.

[0096] Soil parameter inversion utilizes multi-source data fusion from satellite, UAV remote sensing, and ground sensors, combined with machine learning models, to achieve dynamic monitoring of soil moisture, temperature, and nutrients. For soil moisture monitoring, microwave remote sensing enables all-weather dynamic tracking; thermal infrared bands are used to assist in identifying water stress conditions, and the combination of these two methods significantly improves monitoring accuracy. For soil temperature monitoring, thermal infrared and passive microwave remote sensing technologies are used, simultaneously coupled with ground sensor data and climate models, to achieve high-frequency dynamic monitoring at the farmland scale. For soil nutrient inversion, hyperspectral remote sensing technology is employed to capture the spectral characteristics of soil nutrients. Combined with machine learning algorithms and crop growth model assimilation processes, nutrient changes are dynamically tracked, providing data support for precision fertilization.

[0097] Crop parameter inversion uses technologies such as high-resolution remote sensing and UAV multispectral imaging to assess growth status and pests and diseases, guiding agricultural management: high-resolution remote sensing and UAV multispectral imaging are used for leaf area index, chlorophyll, and nitrogen content inversion; machine learning improves the intelligence and prediction accuracy of growth models; UAV imagery and ground sensors monitor pests and diseases in real time, enabling rapid identification and early warning; yield prediction couples remote sensing features with crop growth models, using methods such as random forests for accurate forecasting, optimizing agricultural production.

[0098] Meteorological parameter inversion acquires key climate variables such as wind speed, temperature, humidity, precipitation, and solar radiation by fusing satellite remote sensing and ground meteorological data. Multi-source fusion technology enhances the spatiotemporal resolution of meteorological data, supporting agricultural modeling and risk assessment. Meteorological information serves as a core input for crop growth simulation, pest and disease early warning, and yield prediction. Combined with crop and soil models, it helps smart farms adapt to climate change and optimize management, ultimately improving the stability and resilience of agricultural production.

[0099] Intelligent decision management

[0100] Intelligent decision-making, based on multi-source agricultural sentiment data, provides precise guidance for all stages of tillage, sowing, field management, and harvesting. The core of efficient production management in unmanned farms lies in: Tillage combines ground observation and field sampling methods, using elevation data such as DEM and soil nutrient data to divide fields; sowing intelligently determines the sowing period and density based on meteorological conditions and crop needs, relying on BeiDou positioning and intelligent seeders for precise operation; field management dynamically adjusts water, fertilizer, and pesticide application plans based on crop growth stages and soil moisture, combining plant protection remote sensing, spore monitoring, and AI recognition systems to construct a pest and disease early warning and response mechanism. Figure 4 As shown in the figure, Figure A represents the corn emergence rate, Figure B represents the total nitrogen content of the soil, Figure C represents the field water holding capacity, and Figure D represents the wheat growth. Based on this, agricultural statistics use remote sensing inversion and IoT data to conduct cross-temporal and multi-dimensional agricultural parameter aggregation and trend analysis, serving crop growth diagnosis and yield prediction. Agricultural machinery management relies on digital twin and 3D modeling technologies to complete operation path optimization, status monitoring, and collaborative scheduling, further improving operation efficiency and resource utilization.

[0101] Variable prescription maps, as the core execution carrier of intelligent decision-making, integrate soil, crop, and meteorological information to generate spatial zoning schemes for operations such as sowing, fertilization, and irrigation, directly driving the precise execution of unmanned agricultural machinery. For example... Figure 6 As shown, Figure A represents the fertilizer prescription, and Figure B represents the irrigation prescription. It supports multiple formats including TIFF, Shapefile, and ISO-XML, and is evolving towards autonomous generation and real-time updates, becoming the intelligent hub for "on-demand supply" in farms. Through full-scene digital reconstruction and an IoT sensing system, integrating crop models, agricultural machinery, and sensing terminals, it ultimately achieves visualization, interactivity, and controllability throughout the entire "cultivation, management, and harvesting" process, propelling unmanned farms from automated operation to intelligent control.

[0102] Unmanned operation of agricultural machinery throughout the entire process

[0103] Unmanned operation of agricultural machinery throughout the entire process is the core support for achieving efficient operation of smart farms, covering three major aspects: reliable navigation and positioning, precise operation based on variables, and operation quality monitoring.

[0104] In terms of navigation and positioning, relying on high-precision semantic maps, multi-source fusion positioning, and path planning algorithms, a continuous navigation system covering the entire scene of "hangar-farm road-field" is constructed. This system integrates GNSS, IMU, and LiDAR to enable agricultural machinery to operate autonomously in the field, plan its own paths, park automatically, and work collaboratively. High-precision farmland maps are obtained through UAV oblique photography, combined with a 3D curved surface path planning algorithm to improve the coverage and smoothness of agricultural machinery paths.

[0105] In terms of variable-rate operations, unmanned aerial vehicles (UAVs) utilize intelligent control systems to achieve precise control over tillage, sowing, fertilization, and pesticide application. Intelligent tillage implements integrate deep tillage shovels, rotary tillage mechanisms, and compaction devices, automatically adjusting depth and power to optimize soil fertility. Variable-rate application systems for seeds, fertilizers, and pesticides improve operational accuracy and resource utilization through real-time control of fertilization speed, seeding rate, and spray boom height. For example... Figure 6 The diagram shows the operations of an unmanned farm. Diagram A shows unmanned deep plowing, diagram B shows variable fertilizer spreading, diagram C shows unmanned deep loosening, and diagram D shows unmanned harvesting.

[0106] In terms of operation quality monitoring, a real-time operation quality assessment model is constructed by combining attitude sensor data and operation status data to monitor key indicators such as tillage depth, seeding rate, and pesticide distribution, ensuring the stability and consistency of operation quality of unmanned aerial vehicles (UAVs) in complex farmland environments. Through operation process perception and retrospective analysis, data support is provided for intelligent management and operation optimization of agricultural machinery, promoting the development of unmanned operations towards high quality and high intelligence.

[0107] System Iterative Optimization

[0108] Autonomous feedback and iterative optimization in unmanned farms is a key link in realizing intelligent closed-loop control of agricultural operations. It relies on multi-source sensing data, operation feedback information and self-learning algorithms to promote the transformation from "passive response" to "active evolution". Based on the feedback of crop sentiment perception, operation result monitoring and management decision-making, a closed-loop path of "perception-decision-execution-evaluation-re-optimization" is constructed to continuously improve the efficiency of unmanned operations and the accuracy of agronomic regulation.

[0109] At the operational feedback level, real-time data collection and spatial mapping of operational quality indicators such as sowing depth, fertilization uniformity, and pesticide coverage, combined with crop growth, abnormal growth areas, and environmental changes, enable precise correction of variable prescription maps and operational parameters. At the intelligent learning level, algorithms such as reinforcement learning are introduced to construct a dynamic optimization framework for agricultural machinery operation strategies and agricultural model parameters based on historical operational data and crop yield performance. By adaptively adjusting sowing strategies, path planning, and task scheduling, the unmanned farm acquires environmental adaptability and cross-seasonal learning capabilities. At the global optimization level, big data analysis and digital twin technology are integrated to construct a virtual-real integrated dynamic farm model. Virtual operational simulation verifies the rationality of actual parameter settings and dynamically adjusts spatiotemporal scales, achieving a unified approach to intelligent operations and agricultural situation response.

[0110] The Beidou Smart Unmanned Farm is a typical practical embodiment of embodied intelligent technology. It constructs an intelligent feedback closed-loop system encompassing "Beidou remote sensing signal access—sensory fusion—information product generation—agronomical knowledge integration—variable decision-making—intelligent operation execution—feedback iterative optimization." This not only improves the precision and efficiency of the entire agricultural production process but also demonstrates significant ecological benefits in water conservation, fertilizer reduction, pesticide reduction, and carbon sequestration, reflecting the high degree of integration of digitalization, intelligence, and greening in the agricultural field. In the future, with the continuous and in-depth integration and evolution of cutting-edge technologies such as artificial intelligence, big data models, blockchain, and digital twins, the Beidou Smart Unmanned Farm will propel agriculture into a new development stage of "data-defined yield, intelligent-led management, and system-synergistic efficiency" across a wider range of regions and in greater depths, becoming a core engine for the high-quality development of smart agriculture in my country and a key benchmark for global agricultural technology innovation.

[0111] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of the present invention can be implemented using various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0112] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0113] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0114] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0115] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the invention.

[0116] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A Beidou and remote sensing integrated space-time wearable intelligent system, characterized in that, The system comprises an intelligent sensing module, a data fusion processing module, an intelligent decision module, and an execution and feedback module; the intelligent sensing module is configured to acquire positioning data and remote sensing data; the data fusion processing module is configured to fuse and process the positioning data and the remote sensing data acquired by the intelligent sensing module to obtain space-time data of a target region; the intelligent decision module is configured to generate a decision instruction for the target region based on the space-time data; and the execution and feedback module is configured to execute corresponding operations according to the decision instruction and to feed back the results. The intelligent decision module comprises a space cognition and decision sub-module, which comprises an intelligent decision information extraction module, the intelligent decision information extraction module introduces a remote sensing observation inversion model and an environmental dynamics model, and is constrained and optimized by ground control points, and specifically comprises the following steps: Step 1: modeling of the untrustworthiness of remote sensing signals; comprising: a dynamics model for the evolution of inversion factors over time is constructed, and the general mathematical expression is: ； wherein, : represents the measured physical quantity value of the first : represents the measured physical quantity value of the first : represents the measured physical quantity value of the first : represents the measured physical quantity value of the first : represents the measured physical quantity value of the first Step 2: constructing a space-time remote sensing control network; comprising: each control point provides measured physical quantity data with a time stamp to form a control point set: ； wherein : indicates the longitude, latitude and observation time of the control point; Step 3: constructing a remote sensing inversion model based on the space-time remote sensing control network, comprising: a remote sensing observation inversion model is constructed using the dynamics model, and two residual terms are introduced: ； wherein: represents an estimated inversion value; represents a remote sensing image estimated by inversion; represents a dynamic estimate value derived by the previous time value and the control point set estimated by inversion; represents the residual between the remote sensing estimate value and the dynamic model estimate value; Constructing the objective function: ; wherein, represents a regulation coefficient, controlling the weight between the remote sensing estimate and the dynamic model estimate; Step 4, to and a first order Taylor expansion, construct the Jacobian matrix J, further, the accuracy evaluation formula is: ； wherein, is the covariance matrix of the input observation data, and the inversion value is analyzed for confidence by calculating the standard deviation σ of 2.The Beidou and remote sensing integrated space-time wearable intelligent system according to claim 1, characterized in that, The intelligent sensing module acquires positioning data through the Beidou satellite navigation system, acquires remote sensing signals and remote sensing image data through satellite remote sensing systems and aerial remote sensing systems, and acquires visual images and navigation trajectories through ground Internet of Things systems. 3.The Beidou and remote sensing fusion spatiotemporal wearable intelligent system according to claim 1, characterized in that, The data fusion processing module is configured to perform geometric correction, space-time reference unification and alignment of remote sensing images; the geometric correction algorithm for the remote sensing images is specifically: The control points measured by the Beidou satellite navigation system positioning terminal are used as ground control points, and the ground control points are uniformly distributed in the remote sensing image region; A first-order to third-order polynomial model is selected to establish the mapping relationship between the remote sensing image coordinates (x, y) and the ground coordinates (X, Y), as follows: ； ； , are model coefficients; By constructing the observation equation set, the model coefficients are solved by the least squares method to minimize the control point coordinate fitting error; the conversion relationship is used to perform geometric correction on the remote sensing image, and a digital orthophoto map is generated through bilinear interpolation or cubic convolution resampling; The space-time reference unification and alignment specifically comprises time synchronization and space alignment, and remote sensing assisted Beidou positioning correction; The time synchronization is to correct the time labels of various types of remote sensing data by using the time information of the Beidou satellite navigation system; The space registration is to unify the coordinate systems of different remote sensing data by taking the control points measured by the Beidou satellite navigation system positioning terminal as the reference datum, so as to realize the space alignment among multi-source images; The remote sensing assisted Beidou positioning correction adopts the following formula: ; wherein, is the error, represents the existing remote sensing image product to obtain image features, represents the matching target features extracted from the existing navigation map or prior model, and the correction of Beidou positioning is completed by optimizing .

4. The Beidou and remote sensing integrated space-time wearable intelligent system according to claim 1, characterized in that, The intelligent decision module further comprises a positioning and navigation control sub-module; The positioning and navigation control sub-module is configured to perform geometric correction and image registration on the remote sensing image based on the positioning data and the ground control point information, eliminate the inertial drift error, and generate three-dimensional pose data of a target object. The space cognition and decision-making submodule takes remote sensing images, visual images, and navigation trajectories as inputs, uses a deep convolutional neural network to realize ground feature classification, state recognition, environment understanding, and risk judgment of the target area, further generates task execution strategies and work path planning results with situational response capabilities, and outputs decision-making prescription maps, target priority rankings, and work instruction sets.

5. The Beidou and remote sensing integrated space-time wearable intelligent system according to claim 4, characterized in that, The execution and feedback module is used to control the following devices: aerial work units including aerial work aviation remote sensing aircraft and aerial work unmanned aerial vehicles, which realize aerial variable work based on the decision-making prescription map; ground work units including unmanned agricultural machines, tools, and unmanned ships for ground production work, which realize ground and water surface production and work of unmanned agricultural machines, tools, and unmanned ships based on the decision-making prescription map; Indoor work units realize indoor intelligent work and manufacturing according to the decision-making prescription map and work instructions through unmanned work vehicles and unmanned manufacturing equipment; underwater work units include underwater work robots, which realize underwater work based on the decision-making prescription map. 6.The Beidou and remote sensing fusion space-time wearable intelligent system according to claim 4, characterized in that, The space cognition and decision-making submodule also includes an intelligent decision-making knowledge fusion module and an intelligent decision-making generation module.

7. The Beidou and remote sensing integrated space-time wearable intelligent system according to claim 6, characterized in that, The intelligent decision-making knowledge fusion module embeds a geographic knowledge graph construction, a rule reasoning engine, and a knowledge updating mechanism, extracts "ground feature-attribute-relation" triples from historical remote sensing images, sensor data, and artificial annotation results to construct a geographic knowledge graph, forms semantic environmental information, and uses a graph neural network and an ontology rule system to support knowledge retrieval, causal analysis, and logical reasoning. The knowledge self-updating mechanism automatically corrects knowledge edge weights and adds new nodes based on learning layer feedback information to maintain the dynamic evolution of the knowledge graph. 8.The Beidou and remote sensing fusion space-time wearable intelligent system according to claim 6, characterized in that, The intelligent decision-making generation module realizes intelligent control output through model library management, strategy generation, and decision evaluation; the model library management is used to dynamically select the optimal model according to the scene; the strategy generation includes inputting target area remote sensing images, navigation trajectories, and knowledge graph semantic labels and outputting variable prescription maps and trajectory planning paths; the decision evaluation includes calculating multi-strategy scores according to the set target function and selecting the optimal decision scheme using a multi-objective optimization algorithm. 9.The Beidou and remote sensing fusion space-time wearable intelligent system according to claim 1, characterized in that, The execution and feedback module includes four submodules: decision execution, execution feedback collection, error analysis, model updating, and knowledge migration; The decision execution submodule converts the output work instructions into bottom-layer execution instructions, and the execution feedback collection submodule integrates remote sensing inversion monitoring values, execution logs of each work unit, and sensor state data to establish a multidimensional data backflow mechanism for the work process. An error analysis submodule compares the deviation between the actual operation result and the expected operation instruction to construct a residual error model: ; wherein, : represents a residual error; : represents an actual execution result; : represents a planned value or a target value; a model updating submodule updates model parameters based on a residual error size and distribution; a knowledge transfer submodule is used to transfer a trained model structure and a rule set to a new scenario.

Images