A method and system for evaluating the accuracy of guide rails based on multi-sensor fusion

Abstract

This invention relates to the field of multi-sensor information fusion and intelligent manufacturing technology, and discloses a guide rail accuracy evaluation method and system based on multi-sensor fusion. The guide rail accuracy evaluation method based on multi-sensor fusion includes: constructing a multi-modal sensor value quantification model based on information entropy, calculating the sensor information contribution, and constructing a multi-sensor mutual information coupling model; constructing a self-optimizing sensor network configuration system, abstracting the multi-modal sensor system into a graph structure, and applying a graph reinforcement learning algorithm to generate the optimal sensor combination configuration; constructing a multi-timescale optimization framework and implementing adaptive data compression based on Huffman coding; constructing a dynamic Bayesian network model to fuse heterogeneous sensor data; and constructing a self-evolving evaluation strategy system to continuously improve the evaluation strategy through historical performance data. This invention, through information entropy-driven resource allocation, can dynamically allocate resources according to the sensor information value.

Description

Technical Field

[0001] This invention relates to the field of multi-sensor information fusion and intelligent manufacturing technology, and more specifically, to a guide rail accuracy evaluation method and system based on multi-sensor fusion. Background Technology

[0002] As a key component in high-end equipment manufacturing and precision machining, the accuracy assessment of guide rails is crucial for ensuring equipment performance and product quality. With the advancement of intelligent manufacturing and automation, guide rail accuracy assessment systems are gradually developing towards multi-sensor and multi-modal fusion. However, existing technologies still have many shortcomings in practical applications, mainly in the following aspects:

[0003] Existing guide rail accuracy assessment systems generally employ fixed sensor configurations, lacking the ability to dynamically adjust resource allocation based on actual working conditions and accuracy requirements. In high-precision scenarios, the system often needs to simultaneously activate all sensors and operate at high sampling rates, leading to increased energy and resource consumption. Conversely, in general accuracy scenarios, fixed configurations result in low resource utilization, making it difficult to achieve an effective balance between resources and accuracy. For example, some commercial guide rail testing equipment operates all sensors at full power in all working states, resulting in energy waste. As the number of sensors increases, the complexity of managing and maintaining traditional fixed-configuration sensor systems grows exponentially, making it difficult to meet the expansion needs of large-scale systems. When adding sensors or changing sensor types, it is usually necessary to manually redesign the data fusion strategy and reconfigure the system, a time-consuming process that affects the system's flexibility and maintainability. Existing technologies struggle to balance system optimization and real-time response. On one hand, while simple data fusion methods can meet certain real-time requirements, they are insufficient in terms of accuracy and robustness. On the other hand, while complex optimization algorithms can improve assessment accuracy, they incur high computational costs, making it difficult to meet the needs of online monitoring and real-time feedback. For example, some high-precision guide rail evaluation systems require several hours of computation time to complete a single evaluation, making real-time monitoring of the guide rail status impossible. Existing guide rail accuracy evaluation systems generally lack adaptive mechanisms, failing to automatically adjust evaluation strategies based on different working conditions, guide rail types, and accuracy requirements, resulting in poor system versatility and scalability. Whenever the application scenario changes, the system often needs to be redeveloped and reconfigured, increasing maintenance costs. Traditional methods struggle to quantify the actual contribution of different sensor data to accuracy evaluation, making it difficult to achieve intelligent resource allocation based on information value. This leads to a lack of scientific basis for resource allocation and management, failing to fully leverage the synergistic effects of various sensors and impacting overall evaluation results.

[0004] In summary, existing guide rail accuracy assessment technologies have several shortcomings that urgently need improvement in areas such as dynamic resource allocation, system scalability, real-time performance, adaptability, and information value quantification. There is an urgent need to propose a novel guide rail accuracy assessment method that can achieve multimodal information collaborative optimization, self-configuration, and intelligent resource management to meet the higher requirements for guide rail accuracy assessment in the fields of intelligent manufacturing and high-end equipment. Summary of the Invention

[0005] This invention provides a guide rail accuracy evaluation method and system based on multi-sensor fusion, which solves the technical problems in related technologies such as inefficient resource allocation, poor real-time performance and insufficient adaptive capability.

[0006] This invention provides a guide rail accuracy evaluation method based on multi-sensor fusion, comprising:

[0007] Construct a multimodal sensor value quantification model based on information entropy, calculate the sensor information contribution, and construct a multi-sensor mutual information coupling model;

[0008] Based on the output of the mutual information model, a self-optimizing sensor network configuration system is constructed. The multimodal sensor system is abstracted into a graph structure and a graph reinforcement learning algorithm is applied to generate the optimal sensor combination configuration.

[0009] Based on the optimal sensor combination configuration, multi-level dynamic resource allocation is realized, a multi-time-scale optimization framework is constructed, and adaptive data compression based on Huffman coding is achieved.

[0010] Based on compressed and optimized data, a dynamic Bayesian network model is constructed to fuse heterogeneous sensor data, enabling adaptive adjustment of the network structure.

[0011] Based on the adaptive adjustment results, a self-evolutionary evaluation strategy system is constructed, and the evaluation strategy is continuously improved through historical performance data.

[0012] Furthermore, the step of calculating the contribution of sensor information includes:

[0013] Obtain the probability distribution of guide rail accuracy parameters in historical datasets and calculate their entropy values;

[0014] Calculate the conditional entropy of the guide rail accuracy under known sensor data conditions;

[0015] The information contribution of the sensor is calculated by the difference between the entropy value and the conditional entropy.

[0016] Furthermore, the steps for constructing the sensor multi-sensor mutual information coupling model include:

[0017] Calculate the mutual information between sensor pairs to identify redundant and complementary sensor groups;

[0018] Calculate higher-order mutual information between sensors to capture complex nonlinear complementary relationships;

[0019] Establish a resource allocation function to link the contribution of sensor information with resource consumption.

[0020] Furthermore, the step of optimizing sensor configuration using the graph reinforcement learning algorithm includes:

[0021] The sensor topology network is represented as a graph structure, where nodes represent sensors and edges represent the relationships between sensors;

[0022] A graph neural network is constructed as a function approximator to realize node feature aggregation and node feature update;

[0023] A deep Q-learning framework is used for reinforcement learning training, and the sensor configuration is optimized by maximizing the balance between information gain and resource consumption through an objective function.

[0024] Furthermore, the multi-level dynamic resource allocation includes optimization at the following time scales:

[0025] Microsecond-level sensor adaptive sampling optimization;

[0026] Millisecond-level data processing strategy adjustments;

[0027] Second-level optimization of computing resource allocation;

[0028] Sensor network topology reconfiguration in minutes.

[0029] Furthermore, the steps for implementing adaptive data compression based on Huffman coding include:

[0030] Statistical analysis of the probability of occurrence of each value in the sensor data;

[0031] The negative logarithm of the probability of a data value occurring is used as its encoding length;

[0032] Construct a Huffman tree to assign a variable-length code to each data value;

[0033] The Huffman tree is dynamically updated as the data distribution changes.

[0034] Furthermore, the steps for constructing the dynamic Bayesian network model include:

[0035] Construct a three-layer network structure comprising a sensor layer, a feature layer, and a precision parameter layer;

[0036] Establish four types of dependency relationships: intra-layer synchronous dependency, inter-layer synchronous dependency, intra-layer time dependency, and inter-layer time dependency.

[0037] An adaptive adjustment algorithm for network structure is designed based on the principle of conditional information gain.

[0038] Furthermore, the steps for constructing the self-evolutionary evaluation strategy system include:

[0039] Establish a strategy library and performance evaluation mechanism to store evaluation strategies for different scenarios;

[0040] Implement a strategy optimization and evolution mechanism to continuously improve the evaluation strategy using historical performance data;

[0041] Design a scenario-similarity-based adaptive policy selection algorithm to select the most suitable evaluation policy from the policy library for the current scenario.

[0042] Furthermore, the resource allocation function quantifies the resource utilization efficiency of sensors by calculating the ratio of sensor information contribution to resource consumption, and prioritizes the allocation of high-efficiency sensors based on the resource efficiency ratio.

[0043] This invention provides a guide rail accuracy evaluation system based on multi-sensor fusion, used to perform the aforementioned guide rail accuracy evaluation method based on multi-sensor fusion, comprising:

[0044] The information entropy quantification module is used to construct a multimodal sensor value quantification model based on information entropy.

[0045] The sensor network configuration module is used to build a self-optimizing sensor network configuration system.

[0046] The resource allocation module is used to implement multi-level dynamic resource allocation;

[0047] The data fusion module is used to construct a dynamic Bayesian network model to fuse heterogeneous sensor data;

[0048] The evaluation strategy module is used to build a self-evolving evaluation strategy system.

[0049] The beneficial effects of this invention are as follows: through information entropy-driven resource allocation, the system can dynamically allocate resources according to the value of sensor information, reduce the use of redundant sensors, reduce system energy consumption, and at the same time increase sensor density in key areas to ensure evaluation accuracy.

[0050] Based on a self-optimizing sensor network and a self-evolving evaluation strategy, the system can automatically adjust the evaluation strategy according to different guide rail types, working conditions and accuracy requirements, improving its ability to adapt to changes in working conditions and supporting a unified evaluation platform for multiple types of guide rails.

[0051] Through multi-level dynamic resource allocation and a three-layer edge-fog-cloud computing architecture, the system improves data processing speed while maintaining comparable evaluation accuracy to traditional methods, achieving near real-time accuracy evaluation and meeting online monitoring needs.

[0052] The self-configuration mechanism reduces the complexity of sensor network management from the exponential growth of traditional systems to linear growth, shortens the integration time of new sensors from days to hours, reduces the need for manual intervention, and simplifies system maintenance and expansion.

[0053] Through strategy evolution and experience accumulation, system performance continues to improve with the extension of running time, and the improvement of evaluation efficiency shows a logarithmic growth trend, forming a virtuous cycle of evolution.

[0054] This invention exhibits emergent characteristics such as "dynamic equilibrium" and "information resonance," enabling it to automatically maintain an optimal balance between information and resources when operating conditions change, and to generate unexpected information value gains through specific sensor combinations. These characteristics cannot be achieved in traditional fixed configuration systems. Attached Figure Description

[0055] Figure 1 This is a flowchart of a guide rail accuracy evaluation method based on multi-sensor fusion in this invention;

[0056] Figure 2 It is a bar chart comparing the contribution of sensor information;

[0057] Figure 3 It is a tree diagram showing the distribution of mutual information among sensor combinations;

[0058] Figure 4 It is a grouped bar chart comparing optimization performance across multiple time scales;

[0059] Figure 5 This is a line graph showing the adaptive data compression effect based on Huffman coding;

[0060] Figure 6 This is a radar chart comparing the performance of multimodal information entropy optimization systems. Detailed Implementation

[0061] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, some features described in the examples may be combined in other examples.

[0062] At least one embodiment of the present invention discloses a guide rail accuracy evaluation method based on multi-sensor fusion, such as... Figure 1 As shown, it includes:

[0063] Step 1: Construct a multimodal sensor value quantification model based on information entropy, calculate the sensor information contribution, and construct a multi-sensor mutual information coupling model.

[0064] This step utilizes information entropy theory to construct a multimodal sensor data value quantification model, enabling accurate quantification of the contribution of different sensors and their combinations to the guide rail accuracy assessment; this step includes the following sub-steps.

[0065] Step 1.1: Calculate the contribution of sensor information;

[0066] For each sensor in the guide rail accuracy evaluation system The information contribution is calculated using the principle of information entropy, which is the amount by which the sensor contributes to reducing the uncertainty of guide rail accuracy assessment.

[0067] Before calculating the contribution of sensor information, the system first preprocesses the data from different sensors. Since position sensor data is typically expressed as displacement in millimeters or micrometers, acceleration sensor data in seconds per meter squared, temperature sensor data in degrees Celsius, and vibration sensor data may be expressed as frequency or amplitude, these data with different dimensions and numerical ranges need to be normalized to scale all sensor data to the same numerical range, ensuring the fairness and accuracy of the information contribution calculation.

[0068] Specifically, the contribution of sensor information is calculated as follows:

[0069] ;

[0070] in, Indicates sensor Information contribution The entropy value, representing the accuracy of the guide rail, signifies the uncertainty in the accuracy assessment of the guide rail. Indicates that in the case of a known sensor The conditional entropy of the guide rail accuracy under the given data conditions represents the uncertainty after using the sensor; the higher the information contribution, the greater the value of the sensor for accuracy assessment.

[0071] In practical applications, the system first obtains the guide rail accuracy parameters from the historical dataset. Calculate the entropy value of the probability distribution. Then for each sensor Calculate the conditional entropy after acquiring its data. The difference between the two is the information contribution of the sensor.

[0072] It should be understood that, in some implementations, the system may employ mutual information. It is used to replace information contribution, and its calculation method is the same, but from the perspective of information theory, it is interpreted as the mutual information between sensor data and accuracy parameters.

[0073] like Figure 2 As shown, different types of sensors contribute differently to the accuracy assessment of the guide rail. The figure reveals that position and force sensors contribute the most information, while temperature sensors contribute relatively less. This result validates the rationality and effectiveness of using information entropy theory to quantify the value of different sensors in this application.

[0074] Step 1.2: Construct a multi-sensor mutual information coupling model;

[0075] To quantify the information redundancy or complementarity among multiple sensors, this application constructs a multi-sensor mutual information coupling model. This model identifies redundant sensor groups and complementary sensor groups by calculating the mutual information between sensor combinations. The specific mutual information calculation is as follows:

[0076] ;

[0077] in, Indicates sensor and sensors Regarding guide rail accuracy mutual information, Indicates sensor For guide rail accuracy Information contribution Indicates sensor For guide rail accuracy Information contribution Indicates sensor and sensors Combined effect on guide rail accuracy Information contribution;

[0078] when When, it indicates the sensor and sensors There is information redundancy; when When, it indicates the sensor and sensors They are complementary, and using them in combination can generate additional information gain.

[0079] Through this mutual information model, the system can identify and quantify the synergistic effects between sensors, providing a foundation for subsequent optimization of sensor configuration.

[0080] It should be noted that in some implementations, the system can also calculate higher-order mutual information between three or more sensors to capture complex nonlinear complementary relationships, for example:

[0081] ;

[0082] in, Indicates sensor ,sensor and sensors Regarding guide rail accuracy Third-order mutual information, Indicates sensor and sensors Regarding guide rail accuracy mutual information, Indicates sensor and sensors As a whole with sensors Regarding the accuracy of the guide rail mutual information, Indicates sensor and sensors Regarding guide rail accuracy mutual information, Indicates sensor and sensors Regarding guide rail accuracy Mutual information; third-order mutual information can be captured by sensors. ,sensor and sensors The complex nonlinear relationship between them revealed a synergistic effect that second-order mutual information could not identify.

[0083] like Figure 3 As shown, the distribution of mutual information among different sensor combinations is illustrated. The area represents the information value of each sensor and its combined effect. It can be intuitively seen that some sensor combinations (such as acceleration + vibration) have high complementarity, and the information value they generate is greater than the sum of their individual contributions. This verifies the "information resonance" phenomenon mentioned in this application, that is, the information value generated by a specific sensor combination far exceeds the sum of their individual contributions.

[0084] Step 1.3: Establish a model to balance information entropy and resource efficiency;

[0085] By linking information value with resource consumption, a resource allocation function is established.

[0086] Before establishing the resource allocation function, the system preprocesses various heterogeneous resource indicators. Since information contribution is a dimensionless probability value typically between 0 and 1, computational resource consumption may be expressed as CPU utilization percentage, memory usage in MB, or processing time in milliseconds, and energy consumption is usually measured in watts or joules. These indicators have different dimensions and numerical ranges; directly calculating ratios can lead to some indicators being over-amplified or under-amplified. Therefore, it is necessary to standardize each resource consumption indicator, converting it into dimensionless values ​​of the same scale, in order to perform effective resource efficiency ratio calculations.

[0087] The resource allocation function is implemented as follows: First, the information contribution of each sensor to the accuracy assessment is obtained. Then, its computational resource consumption and energy consumption are calculated separately, and weighting coefficients are set according to the actual application scenario. The information contribution is used as the numerator, and the weighted sum of resource consumption is used as the denominator. The two are divided to obtain the resource efficiency ratio. The weighting coefficients can be adjusted according to the importance attached to computational resources, energy, and time. Finally, the system prioritizes the sensors based on this ratio to achieve optimal resource allocation.

[0088] The system evaluates the resource efficiency of sensors by calculating the ratio of sensor information contribution to resource consumption. Specifically, it first obtains the sensor's information contribution to the guide rail accuracy assessment as the numerator; then, it obtains the sensor's computational resource consumption and energy consumption, multiplies them by corresponding weighting coefficients, and adds them together as the denominator; finally, it divides the numerator by the denominator to obtain the sensor's resource efficiency ratio. The weighting coefficients can be adjusted according to the relative importance of computational resources and energy in the application scenario. The higher the resource efficiency ratio, the greater the information value provided by the sensor per unit of resource consumption, and the system will prioritize the allocation of sensors with high resource efficiency.

[0089] This resource allocation function quantifies the resource utilization efficiency of sensors by calculating the information contribution per unit of resource consumption, providing a quantitative basis for optimal resource allocation. The system can rank the resource efficiency of each sensor according to this function and prioritize the allocation of high-efficiency sensors.

[0090] Alternatively, in some implementations, the resource allocation function can be extended to take into account time costs.

[0091] When calculating resource efficiency, the system considers not only computational resource consumption and energy consumption, but also the time required to acquire and process sensor data as an additional factor. Specifically, in the denominator of resource consumption, in addition to multiplying computational resource consumption by the first weighting coefficient and energy consumption by the second weighting coefficient, time cost is multiplied by the third weighting coefficient. The third weighting coefficient is adjusted according to the importance of the time factor in the application scenario, and can be appropriately increased in scenarios with high real-time requirements.

[0092] Step 2: Based on the output of the mutual information model, construct a self-optimizing sensor network configuration system, abstract the multimodal sensor system into a graph structure, and apply a graph reinforcement learning algorithm to generate the optimal sensor combination configuration;

[0093] Based on the information entropy quantification model established in step 1, this step constructs a self-optimizing sensor network configuration system to realize the automated optimization configuration of the sensor network; this step includes the following sub-steps.

[0094] Step 2.1, construct the sensor topology network model;

[0095] Abstracting a multimodal sensor system into a graph structure:

[0096] ;

[0097] in This diagram illustrates the structure of a multimodal sensor system. Represents a set of sensor nodes. Represents the set of edges representing relationships between sensors;

[0098] The weights of the edges are determined by the degree of information complementarity or redundancy calculated by the mutual information model. Through this topology model, the system can intuitively represent the sensor network structure and the relationships between sensors, providing a basis for network optimization.

[0099] In practice, the system first initializes the sensor topology graph, treating each sensor as a node in the graph. Then, based on the mutual information value calculated in step 1.2, it establishes edge connections between nodes, with the edge weight set to the absolute value of the mutual information. Edges with positive mutual information (indicating redundancy) are marked in red, and edges with negative mutual information (indicating complementarity) are marked in green, facilitating the system's visual analysis and optimization decisions.

[0100] Step 2.2: Apply graph reinforcement learning algorithm to optimize sensor configuration;

[0101] The algorithm applies graph reinforcement learning to optimize the sensor topology network and automatically generates the optimal sensor combination configuration. This algorithm models the sensor configuration problem as a Markov decision process, with the objective function being:

[0102] ;

[0103] in Let be the objective function, representing the expected cumulative return; Let be the expectation operator, representing the expected value over all possible state transitions; The summation symbol; This is the termination time step of the decision-making process; This is a discount factor used to balance the weights of current and future rewards; For the first The information gain provided by the combination of time sensors; Weighting coefficients to balance information gain and resource consumption; For the first Resource consumption at any given moment.

[0104] The system finds the optimal balance between information gain and resource consumption by maximizing this objective function.

[0105] The specific implementation of the graph reinforcement learning algorithm is as follows.

[0106] First, the system represents the sensor topology network as a graph structure:

[0107] ;

[0108] in A graph structure representation of the sensor topology network; Let be the set of nodes, representing all sensor nodes; Let be the set of edges, representing the relationship edges between sensors; This is a node feature matrix, where each row contains attributes such as sensor type, location, accuracy, and energy consumption.

[0109] Before constructing the node feature matrix, the system preprocesses data for various sensor attributes. Since the sensor types are categorical data (e.g., position sensors, accelerometers, temperature sensors, etc.), sensor positions are continuous numerical data possibly represented in coordinate form, sensor accuracy is typically expressed as a percentage or absolute error, and sensor energy consumption is expressed in power units, the categorical data needs to be encoded and converted. Non-numerical data such as sensor type needs to be converted into numerical data, for example, using one-hot encoding or label encoding. Simultaneously, due to the significant differences in the numerical ranges of each attribute, normalization processing is required to ensure the stability and convergence of the graph neural network training.

[0110] Then, the system constructs a graph neural network (GNN) as a function approximator, whose structure includes multiple layers of graph convolutional layers and fully connected layers; the information transmission mechanism of the graph convolutional layers is as follows.

[0111] Node feature aggregation: For nodes From its neighboring nodes Information gathering:

[0112] ;

[0113] in For the node To the node Information messages; For message functions, define how to aggregate information; For the first Layer nodes eigenvectors; For the first Layer nodes eigenvectors; For nodes and Edge characteristics between them.

[0114] message function The specific implementation involves integrating the features of the current node, the features of its neighboring nodes, and the edge features between the two nodes to generate a message for information transmission. Typically, this integration can be achieved by concatenating the features and inputting them into a multilayer perceptron or linear transform, outputting a new set of feature vectors for subsequent node state updates.

[0115] Node feature update: Updating node features based on aggregated information:

[0116] ;

[0117] in For the first Layer nodes eigenvectors; For feature update function; For the first Layer nodes eigenvectors; The summation symbol; For nodes The set of neighboring nodes; For the node To the node The information message.

[0118] Feature update function The specific implementation involves integrating the features of the current node with the messages passed from its neighboring nodes. This is typically done through methods such as concatenation, weighted summation, or gating. The integrated result is then input into a nonlinear transformation (such as an activation function) to obtain the feature representation of the node in the new layer.

[0119] Based on the constructed graph neural network, the algorithm uses the Deep Q-Network (DQN) framework for reinforcement learning training; the specific steps are as follows.

[0120] State representation: The current sensor network configuration state is represented as a node feature matrix, which includes parameters such as the on / off state and sampling frequency of each sensor.

[0121] Action definition: The action space includes:

[0122] Activate / deactivate specific sensors;

[0123] Adjust the sensor sampling frequency;

[0124] Adjust the data sharing configuration between sensors.

[0125] Reward Calculation: Rewards are calculated based on the current information gain and resource consumption.

[0126] ;

[0127] in For the first The reward value at any given moment; For the first Information gain at any given moment; These are weighting coefficients used to balance information gain and resource consumption. For the first Real-time normalized resource consumption.

[0128] Experience replay: Maintaining the experience buffer Storing and transferring samples , used for batch training, where This is the current state; Actions to be taken; For the reward obtained; The next state after the action is performed; This is a sample of a single state transition.

[0129] Target network: Using the target network Stable training, regularly from the main network Copy parameters.

[0130] Loss function: Minimize the temporal difference (TD) error.

[0131] ;

[0132] in The loss function; For the experience buffer Mid-sampling The expected value of the sample; As a reward; Discount factor; For all possible actions Take the maximum value; For the target network to the next state and actions Q-value estimation; The main network's current state and actions Q-value estimation.

[0133] Training process:

[0134] Initialize the main network and target network ;

[0135] For each training cycle:

[0136] according to - Greedy strategy for selecting actions;

[0137] Perform the action, observe the reward and the next state;

[0138] Store experience in a buffer ;

[0139] from Training is performed using small batches of sampled data.

[0140] Calculate the TD objective and loss function;

[0141] Update the main network parameters;

[0142] Every Update the target network every step. The target network synchronization period.

[0143] Through repeated training and iteration, the system gradually learns the optimal sensor configuration strategy and can automatically generate the optimal sensor combination according to different working conditions and accuracy requirements. In practical applications, the system can be trained in a simulation environment in advance, and then the trained model can be deployed to the actual environment and fine-tuned online to adapt to the characteristics of the actual environment.

[0144] Alternatively, in some implementations, the system may use a policy gradient method (such as DDPG, PPO) instead of DQN to handle the continuous action space and achieve finer sensor parameter adjustment.

[0145] Step 2.3: Implement adaptive reconstruction of the sensor network;

[0146] According to one embodiment of this application, the system constructs an adaptive reconfiguration mechanism for the sensor network, enabling the system to automatically adjust the network structure according to changes in the environment, changes in task requirements, or changes in sensor state; the mechanism includes three parts: trigger condition detection, reconfiguration scheme generation, and smooth transition implementation.

[0147] Triggering conditions include: changes in operating conditions (such as guide rail load or speed changes exceeding a threshold), changes in accuracy requirements (such as increased or decreased accuracy requirements), sensor failure or performance degradation, and the addition of new sensors to the system; when the system detects these conditions, it will initiate the network reconfiguration process.

[0148] The reconstruction scheme is generated based on the current sensor state and the latest information entropy quantization results. The graph reinforcement learning algorithm in step 2.2 is applied to quickly generate a sensor configuration scheme that adapts to the new situation. To ensure system stability, a smooth transition mechanism is introduced and a gradual adjustment strategy is adopted to avoid the impact of sudden configuration changes on the system.

[0149] In addition, according to another embodiment of this application, the system can build a preset reconstruction scheme library, pre-calculate the optimal configuration scheme for common scenario changes, and directly load the corresponding scheme when the triggering conditions are met, thereby further improving the reconstruction speed.

[0150] Step 3: Based on the optimal sensor combination configuration, realize multi-level dynamic resource allocation, construct a multi-timescale optimization framework, and realize adaptive data compression based on Huffman coding;

[0151] Based on the sensor value quantification model and self-optimizing network configuration established in the first two steps, this step realizes multi-level dynamic allocation of system resources to ensure maximum evaluation accuracy under limited resource conditions; this step includes the following sub-steps.

[0152] Step 3.1: Construct a multi-timescale optimization framework;

[0153] According to one embodiment of this application, the system constructs a multi-time-scale optimization framework to optimize resource allocation at different time granularities, thereby achieving efficient utilization of system resources; the framework includes:

[0154] Microsecond level: Sensor adaptive sampling optimization, real-time adjustment of sampling frequency;

[0155] Millisecond level: Adjusting data processing strategies, dynamically selecting processing algorithms and precision;

[0156] Second-level: Optimized allocation of computing resources, dynamically scheduling tasks among processing units;

[0157] Minute-level: Sensor network topology reconstruction, adjusting sensor combination configuration.

[0158] The optimization at each level works in concert. Lower-level optimization is constrained by the configuration of higher-level optimization, while higher-level optimization adjusts based on feedback from lower-level optimization, forming a closed-loop optimization system. For example, when the system detects an abnormality in the accuracy of the guide rail in a certain area, microsecond-level optimization will immediately increase the sampling frequency of the sensors in that area, millisecond-level optimization will allocate more accurate processing algorithms to the data in that area, second-level optimization will allocate more computing resources for data analysis in that area, and minute-level optimization may reconstruct the sensor network and increase the number or types of sensors in that area.

[0159] Optionally, in some implementations, the system can add hourly and daily optimization levels, respectively, for historical data pattern analysis and long-term optimization strategy adjustment, to cope with changes on a larger time scale.

[0160] like Figure 4 As shown in the figure, the performance of the multi-timescale optimization framework proposed in this application is demonstrated across different performance metrics. It is clear from the figure that microsecond-level optimization performs best in real-time performance, while minute-level optimization has significant advantages in accuracy and resource efficiency. This result verifies the performance improvement effect of different timescale optimizations on various aspects of the system, proving that the multi-timescale optimization framework can achieve comprehensive performance enhancement of the system.

[0161] Step 3.2: Implement adaptive data compression based on Huffman coding;

[0162] By applying the principle of information entropy, adaptive compression of sensor data is achieved, improving data transmission and storage efficiency; specifically, the encoding length is dynamically allocated to different sensor data based on information entropy.

[0163] The system first calculates the probability of occurrence of each value in the sensor data. Then, for each data value, it calculates the negative logarithm of its probability (base 2) as its encoding length. In this way, high-frequency data values ​​are assigned shorter codes, and low-frequency data values ​​are assigned longer codes, thereby minimizing the overall encoding length. For example, if a sensor data value has a probability of occurrence of 0.5, its encoding length is approximately 1 bit; if the probability is 0.25, the encoding length is approximately 2 bits; if the probability is 0.125, the encoding length is approximately 3 bits, and so on.

[0164] The system first calculates the probability distribution of data from each sensor, then constructs a Huffman tree and assigns a variable-length code to each data value. As the data distribution changes, the system dynamically updates the Huffman tree to maintain optimal compression efficiency. This adaptive compression method based on information entropy can reduce data transmission volume, reduce network bandwidth and storage space requirements, while ensuring lossless compression and not affecting evaluation accuracy.

[0165] It should be noted that in some implementations, the system can perform hierarchical compression based on the importance of the data, using lossless compression to ensure accuracy for critical data and lossy compression to further improve the compression ratio for less important data. In addition, the system can combine predictive coding technology, using time series models to predict data values ​​and transmitting only the prediction error to further improve compression efficiency.

[0166] like Figure 5 As shown, the adaptive data compression method based on Huffman coding proposed in this application demonstrates its compression performance on different types of sensor data. Compared with traditional fixed-length coding (compression rate of 0), all sensor data achieves a higher compression rate, with optical sensor data showing the best compression performance, retaining only 38% of the original data. This result fully verifies the advantages of the adaptive compression method proposed in this application in improving data transmission and storage efficiency.

[0167] Step 3.3: Construct a three-layer computing architecture of edge-fog-cloud;

[0168] According to another embodiment of this application, the system constructs a three-layer computing architecture of edge-fog-cloud, and dynamically allocates computing tasks to the most suitable computing layer according to task computing needs and timeliness requirements.

[0169] Edge layer: Deployed at sensor nodes or nearby processing units, it handles tasks with high real-time requirements and low computational load, such as data filtering and simple feature extraction.

[0170] Fog layer: Deployed in field-level processing units to handle tasks with moderate computational load and timeliness requirements, such as preliminary fusion analysis and short-term trend prediction;

[0171] Cloud layer: Deployed on a central server or cloud platform, it handles complex tasks that require a large amount of computation and are not real-time, such as deep model training and historical data mining.

[0172] The specific implementation of the computational task allocation mechanism is as follows.

[0173] The system performs each computational task Define a three-dimensional feature vector:

[0174] ;

[0175] in For the task eigenvectors; For the task The computational complexity (the amount of computational resources required). For the task Real-time requirements (maximum allowable latency); For the task Data dependencies (the amount of data required to be accessed).

[0176] Before defining the task feature vector, the system preprocesses the data across three dimensions. Since computational complexity may be expressed as the number of floating-point operations or algorithm time complexity, real-time requirements are expressed as latency thresholds in time units (milliseconds, seconds), and data dependencies are expressed as data size (KB, MB, GB), the data in these three dimensions have completely different dimensions and numerical ranges. Therefore, standardization is required to convert them into values ​​of the same scale, ensuring that each dimension has an equal weight when calculating the matching score.

[0177] Task feature extraction function , , The specific implementation involves calculating the computational resources required by each task, the maximum allowed latency, and the amount of data the task needs to access. These values ​​are obtained by analyzing parameters such as the task type, input / output data size, and real-time requirements, and are used for subsequent task allocation and scheduling.

[0178] At the same time, for each computing layer Define resource characteristic vectors:

[0179] ;

[0180] in For the computation layer Resource characteristic vector; For the computation layer Computational power; For the computation layer Response time; For the computation layer Available bandwidth.

[0181] Before defining the resource characteristic vector of the computing layer, the system preprocesses the resource characteristic data. Since computing power may be expressed in FLOPS, number of cores, or processing speed, response time is latency in milliseconds, and available bandwidth is network transmission rate in Mbps or Gbps, these indicators have different dimensions and numerical ranges. When matching computing tasks with the computing layer, it is necessary to normalize each resource characteristic to prevent indicators with large numerical ranges from dominating the matching results.

[0182] Computation layer resource characteristic extraction function , , The specific implementation involves separately calculating the computing power, response time, and available bandwidth of each computing layer. These parameters are obtained by detecting information such as the hardware configuration, network status, and current load of the computing nodes, and are used to evaluate the resource status of each computing layer.

[0183] The system calculates the matching score between the task and the computing layer based on the task feature vector and the resource characteristics of each computing layer.

[0184] The system determines the optimal execution layer for a computational task by calculating a comprehensive matching score. Specifically, it first calculates the ratio of the computational layer's computing power to the task's computational complexity and multiplies it by a first weighting coefficient. Then, it calculates the ratio of the task's real-time requirements to the computational layer's response time and multiplies it by a second weighting coefficient. Next, it calculates the ratio of the computational layer's bandwidth to the task's data dependency and multiplies it by a third weighting coefficient. Finally, it adds these three factors together to obtain the matching score between the task and the computational layer. The weighting coefficients can be adjusted according to the application scenario. The system selects the computational layer with the highest matching score to execute the corresponding computational task and considers the current load of each layer through a load balancing factor to ensure reasonable resource allocation.

[0185] The system dynamically determines the task allocation strategy based on task characteristics, current resource status, and network conditions. For example, for real-time monitoring of guide rail accuracy, basic data filtering and feature extraction are completed at the edge layer, multi-sensor data fusion and preliminary evaluation are completed at the fog layer, and complex accuracy trend analysis and prediction are completed at the cloud layer, forming a hierarchical and collaborative computing system.

[0186] In addition, the system employs an asynchronous communication mechanism between layers, allowing lower layers to work independently while higher layers are still processing, ensuring real-time responsiveness. The system also implements a task migration mechanism, which allows tasks to be dynamically migrated to other layers for processing when network connections are interrupted or resources are insufficient, ensuring continuous system operation.

[0187] Step 4: Based on the compressed and optimized data, construct a dynamic Bayesian network model to fuse heterogeneous sensor data and achieve adaptive adjustment of the network structure;

[0188] To address the challenge of fusion of heterogeneous sensor data, this step constructs a dynamic Bayesian network model to effectively integrate data from different types of sensors; this step includes the following sub-steps.

[0189] Step 4.1: Establish a dynamic Bayesian network structure;

[0190] According to one embodiment of this application, the system constructs a Dynamic Bayesian Network (DBN) structure adapted to guide rail accuracy evaluation. The network consists of three layers: a sensor layer, a feature layer, and an accuracy parameter layer. The sensor layer nodes represent the raw data from various sensors, the feature layer nodes represent the features extracted from the raw data, and the accuracy parameter layer nodes represent various indicators of guide rail accuracy.

[0191] Before constructing the dynamic Bayesian network, the system preprocesses heterogeneous data from different sensors. Since continuous sensor data (such as temperature, displacement, and velocity) have different numerical ranges and units, discrete sensor data (such as switch states and mode selections) require encoding, and time-series feature data may have different sampling frequencies and time scales, to ensure that the Bayesian network can effectively learn the conditional dependencies between variables, all input data need to be standardized, and categorical variables need to be appropriately encoded to suit probabilistic inference calculations.

[0192] In the network structure, the edges represent the conditional dependencies between nodes, and the system learns the conditional probability distribution of these dependencies based on historical data. Unlike static Bayesian networks, dynamic Bayesian networks add a time dimension, enabling them to model temporal dependencies and making them more suitable for dynamic evaluation of guide rail accuracy over time.

[0193] The specific implementation of dynamic Bayesian networks is as follows.

[0194] First, define the network node type:

[0195] Sensor node set:

[0196] ;

[0197] in, Represents a set of sensor nodes. , , They represent the 1st, 2nd, and 3rd respectively. One sensor node, This represents the total number of sensor nodes.

[0198] Feature node set:

[0199] ;

[0200] in, Represents the set of feature nodes. , , They represent the 1st, 2nd, and 3rd respectively. Each feature node This represents the total number of feature nodes.

[0201] Precision parameter node set:

[0202] ;

[0203] in, Represents the set of precision parameter nodes. , , They represent the 1st, 2nd, and 3rd respectively. Precision parameter nodes This represents the total number of precision parameter nodes.

[0204] Next, for each time slice Establish the types of dependency relationships between nodes:

[0205] Intra-layer temporal dependency: Dependencies between nodes within the same layer, such as features. and Conditional dependencies within the same time slice;

[0206] Inter-layer temporal dependency: Dependencies between nodes in different layers, such as sensors. With features The mapping relationship;

[0207] Intra-layer temporal dependency: The dependency relationship between the same node in adjacent time slices, such as... and The temporal relationship;

[0208] Inter-layer time dependency: The dependency relationship between nodes in different layers in adjacent time slices, such as... and The predictive relationship.

[0209] For each dependency, the system needs to learn its conditional probability distribution; for example, for the mapping from the feature layer to the precision layer, the system needs to learn the conditional probability distribution. ,in Indicates precision parameters The set of parent nodes may include feature nodes and precision parameter nodes from the previous time step; It is a conditional probability function.

[0210] The initial network structure is set based on expert knowledge and includes known sensor-feature-accuracy dependencies. Subsequently, the system continuously optimizes the network topology through structural learning algorithms to discover potential dependencies.

[0211] In a practical guide rail accuracy evaluation system, the sensor layer may include nodes such as position sensors, acceleration sensors, and temperature sensors; the feature layer may include nodes such as average value, standard deviation, spectral characteristics, and time domain characteristics; and the accuracy parameter layer may include nodes such as straightness, parallelism, and flatness.

[0212] Optionally, in some implementations, the system can construct a hierarchical Bayesian network, establishing multiple sub-networks at different abstraction levels to address accuracy evaluation problems at different scales.

[0213] Step 4.2: Implement adaptive adjustment of the network structure;

[0214] Based on the principle of conditional information gain, this application designs an adaptive adjustment algorithm for Bayesian network structures, enabling the network to automatically adjust its structure according to data distribution and changes in operating conditions; the adjustment mechanism includes:

[0215] Calculate the conditional information gain for each potential edge (node ​​pair);

[0216] Add connection edges to node pairs whose information gain exceeds the threshold;

[0217] Prune existing edges whose information gain is below the threshold;

[0218] Applying network simplification algorithms ensures a streamlined and efficient structure.

[0219] The method for calculating conditional information gain is as follows.

[0220] For nodes and and the established edge set Conditional information gain is expressed as:

[0221] ;

[0222] in For a given edge set Under the condition, node For nodes Information gain; For a given edge set under conditions Conditional entropy; In the known and under conditions Conditional entropy; This is a conditional symbol.

[0223] when At that time, and Add an edge between them; when an edge already exists At that time, prune the branches on that side, among which The preset threshold for adding edges, This is the preset threshold for pruning the branches.

[0224] The structure learning algorithm employs a scoring-based approach, using the Bayesian information criterion as the scoring function:

[0225] ;

[0226] in Scoring based on Bayesian information criteria; For network structure; For observation datasets; For logarithmic operators; In network structure and parameter estimation Data The likelihood function value; These are parameter estimates; This refers to the number of model parameters; This represents the data sample size.

[0227] The BIC criterion balances model fit and complexity, preventing overfitting.

[0228] The system periodically executes the above process to continuously optimize the network structure. When a change in operating conditions is detected, the system triggers an additional structure learning process to quickly adapt to the new operating conditions. This adaptive adjustment mechanism enables the Bayesian network to continuously evolve with environmental changes and maintain optimal fusion performance.

[0229] In specific applications of guide rail accuracy evaluation, the system may find that the correlation between certain sensors and accuracy parameters is enhanced under certain operating conditions (such as the increased importance of the accelerometer during high-speed operation), and automatically adjust the network structure to strengthen these correlations; at the same time, weaken those connections whose correlation is reduced under the current operating conditions (such as the reduced importance of the temperature sensor during stable operation).

[0230] It should be understood that in some implementations, the system may employ an incremental structure learning algorithm to perform online structure adjustment on newly arrived data, avoiding full retraining and improving system response speed.

[0231] Step 4.3: Integrate heterogeneous sensor data for accuracy assessment;

[0232] According to another embodiment of this application, the system utilizes a constructed dynamic Bayesian network model to fuse data from different types of sensors and infer the guide rail accuracy parameters; the specific fusion process includes:

[0233] Data preprocessing: Normalization, noise reduction, and feature extraction are performed on the data from each sensor.

[0234] Evidence input: The processed sensor data is input as evidence into the Bayesian network;

[0235] Probabilistic inference: Using algorithms such as junctiontree or variational inference, the posterior probability distribution of precision parameters is calculated;

[0236] Results integration: Based on the posterior distribution, the point estimates and confidence intervals of the precision parameters are output.

[0237] The probabilistic inference framework of Bayesian networks can effectively handle the uncertainty of sensor data. When some sensor data is missing or there is noise, the system can still make reasonable inferences based on other sensor data. At the same time, the framework can also provide the confidence level of the evaluation results, providing a reliable reference for decision-making.

[0238] In addition, the system also adopts a dynamic weight adjustment strategy, which dynamically adjusts the weight of the sensors in the fusion process according to the reliability of the sensors under different operating conditions, thereby further improving the fusion accuracy. For example, when the temperature changes drastically, the system will reduce the weight of temperature-sensitive sensor data and increase the weight of temperature-compensated sensors.

[0239] Step 5: Based on the adaptive adjustment results, construct a self-evolutionary evaluation strategy system and continuously improve the evaluation strategy using historical performance data;

[0240] Construct a self-evolving evaluation strategy system so that the accuracy evaluation method can be continuously optimized and evolved through experience accumulation; this step includes the following sub-steps.

[0241] Step 5.1: Establish a strategy library and performance evaluation mechanism;

[0242] According to one embodiment of this application, the system constructs an evaluation strategy library to store evaluation strategies for different scenarios, guide rail types and accuracy requirements; each strategy includes information such as sensor configuration, resource allocation scheme, and fusion model parameters; at the same time, a performance evaluation mechanism is established to quantitatively evaluate the effect of the strategy from multiple dimensions such as evaluation accuracy, resource consumption, and time efficiency.

[0243] The system records the actual performance of the strategy during each evaluation process, including the deviation between the evaluation results and the true values ​​(by comparing with high-precision offline measurements), resource usage, completion time, and other indicators. It calculates the comprehensive performance score and stores the score in association with the strategy for subsequent strategy selection and optimization.

[0244] It should be noted that the performance evaluation adopts a multi-objective evaluation framework, which dynamically adjusts the weight of each dimension indicator according to the different needs of the application scenario. For example, in real-time monitoring scenarios, the weight of time efficiency will be increased; while in high-precision detection scenarios, the weight of evaluating accuracy will be higher.

[0245] Step 5.2: Implement strategy optimization and evolution mechanisms;

[0246] This application designs a strategy optimization and evolution mechanism to continuously improve the evaluation strategy through historical performance data; the mechanism mainly includes the following algorithms.

[0247] Strategy extraction: Extracting high-performance strategies and their applicable conditions from historical records;

[0248] Strategy generalization: Analyze the commonalities of high-performance strategies to form a general strategy template;

[0249] Policy Mutation: Perform small random mutations on existing policies to explore potential improvement opportunities;

[0250] Strategy cross: Combining the strengths of different strategies to generate new strategies;

[0251] Strategy validation: Verify the effectiveness of the new strategy through simulation or small-scale testing.

[0252] The system executes the above process regularly, continuously updating and expanding the strategy library; as the system runs longer, the experience and knowledge accumulated in the strategy library will also increase, and the system performance will continue to improve.

[0253] Optionally, in some implementations, the system can employ a genetic algorithm framework to manage the evolution of the strategy library; each strategy is represented as an "individual", high-performance strategies are retained through the survival of the fittest principle, and new strategies are generated through mutation and crossover operations to achieve the optimized evolution of the strategy library.

[0254] Step 5.3: Implement adaptive strategy selection based on scene similarity;

[0255] When the system faces a new evaluation task, it needs to select or generate the most suitable evaluation strategy from the strategy library for the current scenario. To this end, according to another embodiment of this application, the system designs a strategy adaptive selection algorithm based on scenario similarity.

[0256] Feature extraction: Extract the feature vector of the current scene (including guide rail type, operating parameters, accuracy requirements, etc.);

[0257] Similarity calculation: Calculate the similarity between the current scene and each historical scene in the strategy library;

[0258] Strategy selection: Based on similarity and historical performance, select the basic strategy that is most suitable for the current scenario;

[0259] Strategy Adjustment: Fine-tune the basic strategy according to the differences in the scenario to generate the final execution strategy.

[0260] This similarity-based strategy selection mechanism can make full use of historical experience to quickly generate high-quality strategies for new scenarios, while avoiding strategy search from scratch and improving system response speed.

[0261] In addition, the system also implements a strategy recommendation function. When users manually configure the system, it can recommend the most suitable strategy parameters based on the characteristics of the current scenario to assist users in decision-making. In practical applications, the system can integrate user choices and feedback into the strategy optimization process through interactive learning, thereby achieving human-machine collaborative evolution.

[0262] like Figure 6 As shown in the radar chart, this comprehensive comparison of the multimodal information entropy collaborative optimization method proposed in this application with traditional methods in five key performance indicators. It is clear that the optimization system of this application improves resource utilization efficiency, system adaptability, and autonomous learning ability compared to traditional methods, while reducing management complexity, which is entirely consistent with the above description of technical effects. This chart visually demonstrates the comprehensive advantages of the proposed method over traditional methods in multiple dimensions.

[0263] A guide rail accuracy evaluation system based on multi-sensor fusion, used to execute the aforementioned guide rail accuracy evaluation method based on multi-sensor fusion, includes:

[0264] The information entropy quantification module is used to construct a multimodal sensor value quantification model based on information entropy.

[0265] The sensor network configuration module is used to build a self-optimizing sensor network configuration system.

[0266] The resource allocation module is used to implement multi-level dynamic resource allocation;

[0267] The data fusion module is used to construct a dynamic Bayesian network model to fuse heterogeneous sensor data;

[0268] The evaluation strategy module is used to build a self-evolving evaluation strategy system.

[0269] The embodiments of the present invention have been described above. However, the embodiments are not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make more equivalent embodiments under the guidance of the present embodiments, and all of them are within the protection scope of the present embodiments.

Claims

1. A method for evaluating the accuracy of guide rails based on multi-sensor fusion, characterized in that, include: Construct a multimodal sensor value quantification model based on information entropy, calculate the sensor information contribution, and construct a multi-sensor mutual information coupling model; Based on the output of the mutual information coupling model, a self-optimizing sensor network configuration system is constructed. The multimodal sensor system is abstracted into a graph structure and the graph reinforcement learning algorithm is applied to generate the optimal sensor combination configuration. Based on the optimal sensor combination configuration, multi-level dynamic resource allocation is realized, a multi-time-scale optimization framework is constructed, and adaptive data compression based on Huffman coding is achieved. Based on compressed and optimized data, a dynamic Bayesian network model is constructed to fuse heterogeneous sensor data, enabling adaptive adjustment of the network structure. The steps to construct a dynamic Bayesian network model include: Construct a three-layer network structure comprising a sensor layer, a feature layer, and a precision parameter layer; Establish four types of dependency relationships: intra-layer synchronous dependency, inter-layer synchronous dependency, intra-layer time dependency, and inter-layer time dependency. An adaptive adjustment algorithm for network structure is designed based on the principle of conditional information gain. Based on the adaptive adjustment results, a self-evolutionary evaluation strategy system is constructed, and the evaluation strategy is continuously improved through historical performance data.

2. The guide rail accuracy evaluation method based on multi-sensor fusion according to claim 1, characterized in that, The steps for calculating the contribution of sensor information include: Obtain the probability distribution of guide rail accuracy parameters in historical datasets and calculate their entropy values; Calculate the conditional entropy of the guide rail accuracy under known sensor data conditions; The information contribution of the sensor is calculated by the difference between the entropy value and the conditional entropy.

3. The guide rail accuracy evaluation method based on multi-sensor fusion according to claim 1, characterized in that, The steps for constructing the multi-sensor mutual information coupling model include: Calculate the mutual information between sensor pairs to identify redundant and complementary sensor groups; Calculate higher-order mutual information between sensors to capture complex nonlinear complementary relationships; Establish a resource allocation function to link the contribution of sensor information with resource consumption.

4. The guide rail accuracy evaluation method based on multi-sensor fusion according to claim 1, characterized in that, The steps for optimizing sensor configuration using the applied graph reinforcement learning algorithm include: The sensor topology network is represented as a graph structure, where nodes represent sensors and edges represent the relationships between sensors; A graph neural network is constructed as a function approximator to realize node feature aggregation and node feature update; A deep Q-learning framework is used for reinforcement learning training, and the sensor configuration is optimized by maximizing the balance between information gain and resource consumption through an objective function.

5. The guide rail accuracy evaluation method based on multi-sensor fusion according to claim 1, characterized in that, The multi-level dynamic resource allocation includes optimization at the following time scales: Microsecond-level sensor adaptive sampling optimization; Millisecond-level data processing strategy adjustments; Second-level optimization of computing resource allocation; Sensor network topology reconfiguration in minutes.

6. The guide rail accuracy evaluation method based on multi-sensor fusion according to claim 1, characterized in that, The steps for implementing adaptive data compression based on Huffman coding include: Statistical analysis of the probability of occurrence of each value in the sensor data; The negative logarithm of the probability of a data value occurring is used as its encoding length; Construct a Huffman tree to assign a variable-length code to each data value; The Huffman tree is dynamically updated as the data distribution changes.

7. The guide rail accuracy evaluation method based on multi-sensor fusion according to claim 1, characterized in that, The steps for constructing the self-evolution evaluation strategy system include: Establish a strategy library and performance evaluation mechanism to store evaluation strategies for different scenarios; Implement a strategy optimization and evolution mechanism to continuously improve the evaluation strategy using historical performance data; Design a scenario-similarity-based adaptive policy selection algorithm to select the most suitable evaluation policy from the policy library for the current scenario.

8. The guide rail accuracy evaluation method based on multi-sensor fusion according to claim 3, characterized in that, The resource allocation function quantifies the resource utilization efficiency of sensors by calculating the ratio of sensor information contribution to resource consumption, and prioritizes the allocation of high-efficiency sensors based on the resource efficiency ratio.

9. A guide rail accuracy evaluation system based on multi-sensor fusion, characterized in that, A method for evaluating guide rail accuracy based on multi-sensor fusion as described in any one of claims 1-8 includes: The information entropy quantification module is used to construct a multimodal sensor value quantification model based on information entropy. The sensor network configuration module is used to build a self-optimizing sensor network configuration system. The resource allocation module is used to implement multi-level dynamic resource allocation; The data fusion module is used to construct a dynamic Bayesian network model to fuse heterogeneous sensor data; The evaluation strategy module is used to build a self-evolving evaluation strategy system.

Images