Cognitive flashing system and method based on neural symbols

Abstract

The invention relates to a cognitive flashing system and method based on neural symbols, and the system comprises a multi-modal sensing module which is used for obtaining multi-modal data of a vehicle, and the multi-modal data comprises state data, software feature data and context data; the neural symbol fusion module is used for outputting a first flashing task schedule and a second flashing task schedule through a symbol inference engine and a neural network based on the multi-modal data; the flash task scheduling comprises flash target and flash resource allocation; the fusion module is used for fusing the first flashing task scheduling and the second flashing task scheduling through an attention mechanism and outputting a third flashing task scheduling; and the decision generation module is used for executing the flashing operation based on the third flashing task scheduling. By fusing symbolic reasoning and deep learning, the problems of stiffness and black box of a traditional flashing decision system are solved, and the adaptivity, safety and decision efficiency of vehicle software OTA flashing are remarkably improved.

Description

Technical Field

[0001] This invention belongs to the field of intelligent connected vehicle software technology, specifically relating to a cognitive writing system and method based on neural symbols. Background Technology

[0002] With the rapid development of intelligent connected vehicles, over-the-air (OTA) software download technology has become a key technology in the automotive electronics field. Currently, vehicle software rewriting decisions mainly adopt the following two technical approaches: 1. Rule-based decision-making systems: These systems make decisions through predefined rewriting rules and procedures. This system relies on a rule base built from expert experience, ensuring the interpretability of the decision-making process, but suffers from poor flexibility and insufficient adaptability. 2. Machine learning-based decision-making systems: These systems learn decision patterns from historical data through neural network models. This method has strong pattern recognition capabilities, but the decision-making process lacks interpretability and is highly dependent on the quality and quantity of data.

[0003] The existing technology has the following main technical defects: 1. Decision rigidity problem: Rule-based systems cannot adapt to complex dynamic environments. When new scenarios not covered by predefined rules arise, the system cannot make effective decisions.

[0004] 2. Black-box decision-making problem: Although neural network-based systems have strong learning capabilities, their decision-making process is not transparent, making it difficult to meet the high safety and reliability requirements of the automotive industry.

[0005] 3. Lagging knowledge updates: Both systems struggle to achieve continuous knowledge evolution and require manual intervention to update and maintain rules or models.

[0006] 4. Insufficient utilization of multi-source information: Existing systems fail to effectively integrate multimodal information such as vehicle status, software characteristics, and environmental context, resulting in limited comprehensiveness and accuracy of decision-making. Summary of the Invention

[0007] To address the problems raised in the background art, a first aspect of the present invention provides a cognitive writing system based on neural symbols, comprising: a multimodal perception module for acquiring multimodal data of a vehicle, the multimodal data including state data, software feature data, and context data; a neural symbol fusion module for outputting a first writing task schedule and a second writing task schedule based on the multimodal data through a symbolic reasoning engine and a neural network, respectively; the writing task schedule including a writing target and a writing resource allocation; a fusion module for fusing the first and second writing task schedules through an attention mechanism and outputting a third writing task schedule; a decision generation module for performing a writing operation based on the third writing task schedule; and a feedback optimization module for evaluating the writing effect using the multimodal data after the writing operation; and optimizing the rule base of the symbolic reasoning engine and the model parameters of the neural network based on the writing effect.

[0008] In some embodiments of the present invention, the neural symbol fusion module includes: a symbolic reasoning engine, used to output a first writing task schedule based on symbols through forward chain reasoning based on a preset predicate logic rule base and multimodal data; and a neural network, used to output a second writing task schedule based on multimodal data through a Transformer encoder and attention mechanism.

[0009] In some embodiments of the present invention, the fusion module includes: an analysis unit, configured to dynamically adjust the weights of the first and second brush task scheduling based on scene features; and a fusion unit, configured to weightedly fuse the first and second brush task scheduling based on the weights, and output a third brush task scheduling.

[0010] Furthermore, the analysis unit includes: an identification unit for identifying the security and complexity of the current scene; and an adjustment unit for adjusting the weights of the first and second write task scheduling based on the identification results of the identification unit: if it is a security constraint scene, a higher weight is assigned to the symbolic inference engine; if it is a pattern recognition scene, a higher weight is assigned to the neural network; and if it is a mixed scene, the weights are dynamically allocated according to the ratio of the symbolic inference engine and the neural network.

[0011] In some embodiments of the present invention, the decision generation module includes: a determination unit, configured to determine the dependencies and resource allocation of the writing target; and an execution unit, configured to perform a writing operation based on the dependencies and resource allocation.

[0012] In some embodiments of the present invention, the feedback optimization module includes: an evaluation unit, used to evaluate the writing effect through multimodal data after the writing operation; the writing effect includes user satisfaction, writing success rate, average time consumption and cost; a first optimization unit, used to optimize the rule base and rule confidence of the symbolic inference engine according to the writing effect; and a second optimization unit, used to optimize the model parameters of the neural network according to the writing effect.

[0013] In some embodiments of the present invention, the multimodal perception module includes: a data acquisition unit for acquiring multimodal data of the vehicle; and a mapping unit for normalizing the multimodal data and mapping the normalized multimodal data to a multidimensional vector space.

[0014] A second aspect of the present invention provides a method for writing data based on the neural symbol-based cognitive writing system of the first aspect, comprising: acquiring multimodal data of a vehicle, the multimodal data including state data, software feature data, and context data; based on the multimodal data, outputting a first writing task schedule and a second writing task schedule through a symbolic inference engine and a neural network, respectively; the writing task schedule including a writing target and a writing resource allocation; fusing the first writing task schedule and the second writing task schedule through an attention mechanism and outputting a third writing task schedule; performing a writing operation based on the third writing task schedule; evaluating the writing effect through the multimodal data after the writing operation; and optimizing the rule base of the symbolic inference engine and the model parameters of the neural network according to the writing effect.

[0015] A third aspect of the present invention provides an electronic device comprising: one or more processors; and a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the writing method of the cognitive writing system based on neural symbols provided in the second aspect of the present invention.

[0016] In a fourth aspect, the present invention provides a computer-readable medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the writing method of the cognitive writing system based on neural symbols provided in the second aspect of the present invention.

[0017] The beneficial effects of this invention are: This invention, through neuro-symbolic fusion, simultaneously achieves the powerful pattern recognition capabilities of deep learning and the logical rigor and interpretability of symbolic reasoning, effectively solving the black-box problem of traditional AI systems. The system's adaptive capabilities are significantly enhanced: it can handle the vast majority of undefined scenarios and dynamically adjust decision-making strategies in real time, supporting online learning and continuous knowledge evolution. Decision efficiency is greatly improved: the parallel processing architecture significantly reduces overall decision response time and increases system throughput. Closed-loop feedback and continuous knowledge evolution: a complete closed loop is established from decision execution to effect evaluation, and then to the automatic optimization of the symbolic rule base and neural network model. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the basic structure of a cognitive writing system based on neural symbols in some embodiments of the present invention. Figure 2 This is a schematic diagram of the specific structure of a cognitive writing system based on neural symbols in some embodiments of the present invention; Figure 3 This is a schematic diagram illustrating the basic principle of a neural symbol-based cognitive writing system in some embodiments of the present invention; Figure 4 This is a schematic diagram illustrating the working principle of the decision generation module of the neural symbol-based cognitive writing system in some embodiments of the present invention. Figure 5 This is a basic flowchart of the writing method of the cognitive writing system based on neural symbols in some embodiments of the present invention; Figure 6 This is a schematic diagram of the structure of an electronic device in some embodiments of the present invention. Detailed Implementation

[0019] The principles and features of the present invention are described below with reference to the accompanying drawings. The examples given are only for explaining the present invention and are not intended to limit the scope of the present invention.

[0020] Example 1 refer to Figure 1 and Figure 2In a first aspect, the present invention provides a cognitive writing system 1 based on neural symbols, comprising: a multimodal perception module 11 for acquiring multimodal data of a vehicle, the multimodal data including state data, software feature data, and context data; a neural symbol fusion module 12 for outputting a first writing task schedule and a second writing task schedule based on the multimodal data through a symbolic reasoning engine and a neural network, respectively; the writing task schedule including writing targets and writing resource allocation; a fusion module 13 for fusing the first writing task schedule and the second writing task schedule through an attention mechanism and outputting a third writing task schedule; a decision generation module 14 for performing a writing operation based on the third writing task schedule; and a feedback optimization module 15 for evaluating the writing effect through the multimodal data after the writing operation; and optimizing the rule base of the symbolic reasoning engine and the model parameters of the neural network according to the writing effect.

[0021] refer to Figure 3 In some embodiments of the present invention, the multimodal perception module 11 includes: a collection unit for collecting multimodal data of the vehicle; and a mapping unit for normalizing the multimodal data and mapping the normalized multimodal data to a multidimensional vector space.

[0022] Specifically, the multimodal perception module collects vehicle status data such as CPU load, memory usage, power supply voltage, and network quality (packet loss rate, latency) from the ECU at a frequency of 100Hz via the CAN bus, and applies Kalman filtering for smoothing and noise reduction. It parses the binary file of the software to be flashed, statically extracting the version number, digital signature, and dependency library list; and dynamically monitors its runtime API call sequence and resource usage patterns. It obtains the vehicle's geographical location and network signal strength (RSRP / RSSI) through GPS, network modules, etc., and analyzes user historical behavior data to identify operational preferences. All the above numerical parameters are normalized to the [0,1] interval, and a 256-dimensional multimodal feature vector is constructed for use by the neural network. Simultaneously, key parameters (such as vehicle_speed=0) are converted into logical facts for symbolic reasoning.

[0023] In some embodiments of the present invention, the neural symbol fusion module 12 includes: The symbolic reasoning engine is used to output the first write task schedule based on symbols through forward chain reasoning, based on a pre-defined predicate logic rule base and multimodal data. Specifically, the symbolic reasoning engine loads a first-order predicate logic rule base (including security constraints, system dependencies, and business logic rules). Logical facts are input into the engine, and forward chain reasoning is executed. For example, it verifies whether the rule `safe_to_flash(Software):-vehicle_speed(0), battery_level(Level), Level>0.3` is satisfied. It outputs a symbolic decision suggestion (such as "Allow flashing") and calculates its confidence score C_symbol (based on rule matching degree and historical accuracy).

[0024] A neural network is used to output a second brush task scheduler based on multimodal data, through a Transformer encoder and attention mechanism.

[0025] Specifically, the neural network (path) inputs a 256-dimensional feature vector into an 8-layer Transformer encoder. Feature weights are calculated using a self-attention mechanism, and after a nonlinear transformation, a decision probability distribution is output (e.g., [allowed: 0.75, delayed: 0.20, forbidden: 0.05]). Then, the neural confidence C_neural (based on model calibration and prediction consistency) is calculated.

[0026] In some embodiments of the present invention, the fusion module 13 includes: an analysis unit, configured to dynamically adjust the weights of the first and second brush task scheduling based on scene features; and a fusion unit, configured to weightedly fuse the first and second brush task scheduling based on the weights, and output a third brush task scheduling.

[0027] Furthermore, the analysis unit includes: an identification unit, used to identify the security and complexity of the current scene; Specifically, the analysis unit analyzes scene features in real time. When decision-making heavily relies on hard safety constraints, it tends to favor symbolic pathways; when it relies on complex pattern recognition, it tends to favor neural networks.

[0028] The adjustment unit is used to adjust the weights of the first and second writing task scheduling based on the recognition results of the recognition unit: if it is a security constraint scenario, a higher weight is assigned to the symbolic inference engine; if it is a pattern recognition scenario, a higher weight is assigned to the neural network; if it is a mixed scenario, the weights are dynamically allocated according to the ratio of the symbolic inference engine and the neural network.

[0029] The weight calculation formula is: W_symbol = C_symbol / (C_symbol + C_neural). W_neural = 1 - W_symbol; In the formula, W_symbol and W_neural represent the weights of the symbolic inference engine and the neural network, respectively, and C_symbol and C_neural represent the confidence levels of the symbolic inference engine and the neural network, respectively.

[0030] refer to Figure 4 In some embodiments of the present invention, the decision generation module 14 includes: a determination unit, used to determine the dependency relationship and resource allocation of the writing target; and an execution unit, used to perform the writing operation based on the dependency relationship and resource allocation.

[0031] Specifically, the cognitive decision generator formulates a detailed flashing plan (including the target ECU, resource allocation, and rollback scheme) based on the final decision. It executes the flashing operation and monitors data transmission and resource usage status throughout the process, triggering a handling mechanism in case of anomalies.

[0032] Specifically, it also includes a feedback learning and optimization unit to evaluate the effectiveness of the decision from both technical (success rate, time consumption) and business (user experience) dimensions. Based on the evaluation results, it optimizes the system by: revising or expanding the symbol rule base; incrementally training the neural network using new data; and adjusting the parameters in the dynamic weight calculation model to optimize weight allocation in similar scenarios in the future.

[0033] In embodiments of the present invention, data acquisition and preprocessing ≤ 5ms, dual-path parallel processing ≤ 20ms, decision generation ≤ 15ms, and overall system response time ≤ 40ms. Real-time performance: Data acquisition and preprocessing ≤ 5ms, dual-path parallel processing ≤ 20ms, decision generation ≤ 15ms, and overall system response time ≤ 40ms. Accuracy: Decision accuracy ≥ 96.5%, risk identification accuracy ≥ 95%. Resources: Memory usage ≤ 8GB, CPU utilization ≤ 70%, supports operating temperature from -40℃ to 85℃.

[0034] Continue to refer to Figure 4 For example, it details the decision-making logic and multi-path processing mechanism of active modulation. The complete process of active modulation is presented in the form of a standard flowchart; diamonds represent decision points, and rectangles represent processing steps; colors distinguish different states (normal, successful, abnormal); it reflects the closed-loop control logic from monitoring to modulation; based on the activity threshold, the modulation party (adjustment party) is selected; adjustment methods include resource reallocation, priority adjustment, and function degradation.

[0035] The attention fusion mechanism based on dynamic confidence is understandable: it pioneered a fusion method that dynamically allocates symbolic and neural pathway weights based on real-time scene confidence, achieving an intelligent balance between "logical rigor" and "pattern recognition capability." The parallel processing architecture for vehicle OTA with neural symbols and neural networks is designed to allow symbolic reasoning and neural networks to execute in parallel and independently at the hardware level, significantly improving processing efficiency and system response speed. The closed-loop feedback and continuous knowledge evolution system establishes a complete closed loop from decision execution to effect evaluation, and then to the automatic optimization of the symbolic rule base and neural network model, enabling the system to continuously improve itself. The unified representation and collaborative processing mechanism for multimodal perception data proposes a method to construct unified feature vectors and logical facts from multi-source heterogeneous data such as vehicle state, software features, and environmental context, providing a solid foundation for dual-pathway processing.

[0036] In some embodiments of the present invention, the feedback optimization module 15 is used to evaluate the brushing effect through multimodal data after the brushing operation; and optimize the rule base of the symbolic reasoning engine and the model parameters of the neural network based on the brushing effect.

[0037] Furthermore, the feedback optimization module 15 includes: The evaluation unit is used to evaluate the brushing effect through multimodal data after the brushing operation; the brushing effect includes user satisfaction, brushing success rate, average time consumption and cost; Specifically, the technical effectiveness evaluation includes: Decision accuracy: Number of successful decisions / Total number of decisions; System performance impact: Performance comparison before and after flashing; Resource utilization efficiency: CPU, memory, and network utilization analysis; Business effectiveness evaluation: User satisfaction: User feedback collected through the vehicle system; Service quality indicators: Flashing success rate, average time consumption statistics; Economic benefit analysis: Cost savings calculation.

[0038] The first optimization unit is used to optimize the rule base and rule confidence of the symbolic reasoning engine based on the brushing effect. Specifically, the optimization of the symbol rule base includes root cause analysis of failure cases and correction of erroneous rules; extraction of successful case patterns and expansion of the rule base; and dynamic adjustment of rule confidence.

[0039] The second optimization unit is used to optimize the model parameters of the neural network based on the brushing effect.

[0040] Specifically, incremental learning involves fine-tuning model parameters using new data; online learning allows for real-time model adjustments to adapt to new scenarios; model version management maintains multiple versions and supports rollback. Fusion parameter tuning adjusts weight calculation parameters based on performance feedback; optimizes confidence calculation methods; and adjusts scene classification thresholds. In particular, a long-term system maintenance performance monitoring system is in place: Establish key performance indicator dashboards, set performance alarm thresholds, and generate performance analysis reports regularly; Improve adaptive capabilities: Enhanced learning mechanisms for new scenarios, optimized anomaly handling capabilities, and continuous improvement of user experience are ensured by technical parameters; Real-time performance is guaranteed: Data acquisition and preprocessing: ≤5ms; Dual-path parallel processing: ≤20ms; Decision generation and optimization: ≤15ms; End-to-end response time: ≤40ms; Accuracy indicators: Decision accuracy: ≥96.5%; Risk identification accuracy: ≥95%; System availability: ≥99.9%; Resource constraints: Memory usage: ≤8GB; CPU utilization: ≤70%; Storage space: ≤256GB; Power consumption limit: ≤25W.

[0041] Example 2 refer to Figure 5 A second aspect of the present invention provides a method for writing data based on the neural symbol-based cognitive writing system of the first aspect, comprising: S100. acquiring multimodal data of a vehicle, wherein the multimodal data includes state data, software feature data, and context data; S200. based on the multimodal data, outputting a first writing task schedule and a second writing task schedule through a symbolic reasoning engine and a neural network, respectively; wherein the writing task schedule includes a writing target and a writing resource allocation; S300. fusing the first writing task schedule and the second writing task schedule through an attention mechanism, and outputting a third writing task schedule; S400. performing a writing operation based on the third writing task schedule; S500. evaluating the writing effect through the multimodal data after the writing operation; and optimizing the rule base of the symbolic reasoning engine and the model parameters of the neural network according to the writing effect.

[0042] Furthermore, in step S300, the process of fusing the first and second brush task schedules through an attention mechanism and outputting the third brush task schedule includes: dynamically adjusting the weights of the first and second brush task schedules based on scene features; and weightedly fusing the first and second brush task schedules based on the weights and outputting the third brush task schedule.

[0043] In one specific embodiment of the present invention, the following steps are included: Step S1: System Initialization and Parameter Configuration Step S11: Hardware resource pre-allocation: Allocate 128MB of double-buffered memory area for perception data caching to prevent data read / write conflicts; reserve 64MB of dedicated memory area for policy calculation to ensure real-time computing needs; allocate 256MB of working memory for the execution process to support parallel processing tasks; reserve CPU cores 2 and 3 specifically for handling write-related computing tasks; allocate 512MB of GPU video memory for neural network model inference; set network communication to the highest priority to ensure real-time data transmission. Step S12: Software Module Loading and Initialization: Load the security assessment model, complete weight initialization and inference verification tests; load the network prediction model, establish an environmental state prediction baseline; load the strategy generation model, initialize genetic algorithm parameters; establish CAN bus data stream connection, configure VCU, BMS, and gateway communication parameters; initialize environmental sensor data stream, set signal strength and location information acquisition frequency; establish user behavior data channel, load historical preference database.

[0044] Step S20: Synchronous acquisition and preprocessing of multimodal data: Step S21: Accurately acquire vehicle status data. Synchronously acquire the following parameters via CAN bus at a frequency of 100Hz: CPU load rate of each ECU, accuracy ±1%; memory usage rate, sampling interval 50ms; power supply voltage and fluctuation range, accuracy ±0.1V; network communication quality: packet loss rate (accuracy 0.1%), transmission delay (accuracy 1ms), available bandwidth; apply Kalman filtering algorithm to smooth the raw data: set process noise covariance Q=0.01; set observation noise covariance R=1.0; perform 5 iterations of filtering to eliminate sensor noise.

[0045] Step S22: Deep feature extraction using software: Static feature extraction: Parse the ELF header structure of the software binary file to extract the version number and build timestamp; verify that the digital signature uses the RSA-2048 algorithm; analyze the dependency library list and build a version compatibility matrix; extract security level identifiers and permission requirement lists; Dynamic behavior monitoring: Monitor API call sequences and record call frequency and time distribution; analyze memory usage patterns and detect memory leak risks; record exception handling logs and classify and statistically analyze the frequency of error codes; monitor CPU usage fluctuations and identify abnormal peak patterns; Step S23: Comprehensive environmental context perception; External environment perception: GPS positioning data update frequency 1Hz, accuracy ±3 meters; 4G / 5G signal strength (RSRP) sampling frequency 10Hz, threshold range -110dBm to -60dBm; WiFi signal strength (RSSI) sampling frequency 5Hz, threshold range -90dBm to -30dBm; Weather condition recognition, including temperature, humidity, and rainfall; Road type recognition, distinguishing between highways, urban roads, and rural roads; User behavior analysis: Analyze driving data from the past 30 days to identify patterns of rapid acceleration and sudden braking; Statistical analysis of the usage frequency of each functional module to establish a user preference profile; Analyze historical data writing records to identify user acceptance thresholds; Step S24: Data Standardization and Feature Engineering Numerical parameter normalization: Using the min-max normalization method, all parameters are mapped to the [0,1] interval; outliers are removed and interpolated using the 3σ principle; a 256-dimensional multimodal feature vector is constructed: vehicle state features: 32 dimensions (CPU, memory, voltage, network, etc.); software features: 128 dimensions (version, signature, dependency relationships, behavior patterns, etc.); environmental context features: 96 dimensions (location, signal, weather, user behavior, etc.); logical fact assertion generation: continuous variables are discretized into logical assertions, such as "battery_level(high)" and "network_quality(good)"; a mapping table between fact assertions and original data is established.

[0046] First, the system performs layer normalization on the 256-dimensional feature vector and adds positional encoding to preserve feature sequence information. Then, the Transformer encoder performs forward propagation: it consists of 8 encoder layers, each composed of a multi-head self-attention mechanism (8 heads, attention dimension 64), a feedforward neural network (hidden layer dimension 2048), residual connections, and layer normalization, using the GELU activation function instead of ReLU. In the output layer processing stage, the final encoded output is subjected to global average pooling, then passes through a fully connected layer to output the decision probability distribution, and a Softmax function is used to ensure the probability sum is 1. The comprehensive confidence evaluation includes model calibration (adjusting the output distribution through temperature scaling), prediction consistency (based on the inverse variance of the last 10 predictions), and feature importance (feature contribution based on attention weights). The final confidence calculation formula is C_neural = 0.5 × model calibration + 0.3 × prediction consistency + 0.2 × feature importance.

[0047] Step S3: Step S31: Dynamic Weight Calculation: The system analyzes scene features in real time. When decision-making heavily relies on hard safety constraints, it tends to favor symbolic pathways; when relying on complex pattern recognition, it tends to favor neural pathways. The weight calculation formula is: W_symbol=C_symbol / (C_symbol+C_neural); W_neural=1-W_symbol.

[0048] Step S32: Weighted Fusion and Output: Perform linear weighted fusion of the output vectors from the two paths according to their weights: Final_Vector = W_symbol Symbolic_Vector+W_neural Neural_Vector. The category with the highest probability is ultimately selected as the decision output.

[0049] It should be noted that attention fusion and intelligent decision-making are involved. In the dynamic weight adaptive calculation, the system analyzes scene features in real time, identifies safety-critical features (such as hard constraints like vehicle speed, voltage, and temperature) and pattern recognition features (such as soft constraints like network trends and user behavior), and uses a random forest classifier for scene classification.

[0050] Step S4: Dynamic weight allocation is adjusted according to scenario type: In safety-critical scenarios, W_symbol=0.9, W_neural=0.1; In pattern recognition scenarios, W_symbol=0.2, W_neural=0.8; in mixed scenarios, confidence ratio allocation is used, i.e., W_symbol=C_symbol / (C_symbol+C_neural), W_neural=1-W_symbol. Step 402 performs weighted fusion for precise execution: First, the output vector is standardized, and the symbolic output is vectorized to [allow → 1.0, delay → 0.0, prohibit → 0.0]. The neural output is calibrated using temperature scaling to ensure that the vectors are in the same metric space; then, linear weighted fusion is performed, and the following calculations are made: Final_Vector = W_symbol × Symbolic_Vector + W_neural × Neural_Vector, and evaluate the fusion quality (calculate decision determinism through entropy); the final decision generates the class with the highest probability, and the output format is {decision:"allow|delay|deny",confidence:0.xx,timestamp:"yyyy-MM-ddHH:mm:ss"}.

[0051] Step S5: Covers decision execution and full-process monitoring.

[0052] Step S51, intelligent generation of flashing plan, includes determining the target ECU list (based on dependency graph topology sorting, identifying parallel flashing opportunities and avoiding conflicts), optimizing resource allocation (reserving 30% of CPU computing power, time-sharing memory reuse, and ensuring network bandwidth), and formulating rollback plan (target recovery time < 5 seconds, and contingency plan for abnormal handling).

[0053] Step S52. Detailed monitoring of the execution process includes real-time status tracking (data transmission progress, system resource usage, key parameter monitoring) and anomaly detection and handling (threshold alarms such as voltage <12.2V or temperature >85℃, pattern recognition, automatic recovery).

[0054] Step S60 involves feedback learning and continuous optimization.

[0055] Step S61. Multi-dimensional effect evaluation includes technical effects (decision accuracy, system performance impact, resource utilization efficiency) and business effects (user satisfaction, service quality indicators, economic benefit analysis).

[0056] Step S62. Intelligent knowledge base update includes symbol rule base optimization (root cause analysis of failure cases, pattern extraction of success cases, and adjustment of rule confidence), neural network model update (incremental learning, online learning, and version management), and fusion parameter tuning (weight calculation parameters, confidence method, and scene classification threshold).

[0057] Step S63. Long-term system maintenance includes performance monitoring system (key indicator dashboard, alarm threshold, performance report) and adaptive capability improvement (new scenario learning, anomaly handling optimization, user experience improvement).

[0058] In terms of technical parameter assurance, the real-time requirements are: data acquisition and preprocessing ≤5ms, dual-path parallel processing ≤20ms, decision generation and optimization ≤15ms, and end-to-end response time ≤40ms; the accuracy indicators include decision accuracy ≥96.5%, risk identification accuracy ≥95%, and system availability ≥99.9%; the resource constraints are: memory usage ≤8GB, CPU utilization ≤70%, storage space ≤256GB, and power consumption limit ≤25W.

[0059] Example 3 refer to Figure 6 In a third aspect, the present invention provides an electronic device comprising: one or more processors; and a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the writing method of the neural symbol-based cognitive writing system of the second aspect of the present invention.

[0060] Electronic device 500 may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) 501, which can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) 502 or a program loaded from storage device 508 into random access memory (RAM) 503. The RAM 503 also stores various programs and data required for the operation of electronic device 500. The processing unit 501, ROM 502, and RAM 503 are interconnected via bus 504. An input / output (I / O) interface 505 is also connected to bus 504.

[0061] Typically, the following devices can be connected to I / O interface 505: input devices 506 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 507 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 508 including, for example, hard disks; and communication devices 509. Communication device 509 allows electronic device 500 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 6 An electronic device 500 with various devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively. Figure 6 Each box shown can represent a device or multiple devices as needed.

[0062] Specifically, according to embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 509, or installed from a storage device 508, or installed from a ROM 502. When the computer program is executed by a processing device 501, it performs the functions defined in the methods of embodiments of this disclosure. It should be noted that the computer-readable medium described in embodiments of this disclosure can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In embodiments of this disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in connection with an instruction execution system, apparatus, or device. In embodiments of this disclosure, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. Program code contained on a computer-readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0063] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device. The aforementioned computer-readable medium carries one or more computer programs, which, when executed by the electronic device, cause the electronic device to: Computer program code for performing the operations of embodiments of this disclosure can be written in one or more programming languages ​​or a combination thereof. Programming languages ​​include object-oriented programming languages—such as Java, Smalltalk, C++, and Python—and conventional procedural programming languages—such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0064] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0065] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A cognitive writing system based on neural symbols, characterized in that, include: A multimodal perception module is used to acquire multimodal data of the vehicle, including state data, software feature data, and context data; The neural symbol fusion module is used to output a first brushing task schedule and a second brushing task schedule based on multimodal data through a symbolic reasoning engine and a neural network, respectively; the brushing task schedule includes brushing targets and brushing resource allocation; The fusion module is used to fuse the first and second flush task schedules through an attention mechanism and output the third flush task schedule. The decision generation module is used to execute the write operation based on the third write task scheduling. The feedback optimization module is used to evaluate the brushing effect using multimodal data after the brushing operation; based on the brushing effect, it optimizes the rule base of the symbolic reasoning engine and the model parameters of the neural network.

2. The cognitive writing system based on neural symbols according to claim 1, characterized in that, The neural symbol fusion module includes: The symbolic reasoning engine is used to output the first write task schedule based on symbols through forward chain reasoning, based on a pre-defined predicate logic rule base and multimodal data. A neural network is used to output a second brush task scheduler based on multimodal data, through a Transformer encoder and attention mechanism.

3. The cognitive writing system based on neural symbols according to claim 1, characterized in that, The fusion module includes: The analysis unit is used to dynamically adjust the weights of the first and second write task scheduling based on scene characteristics. The fusion unit is used to weight and fuse the first and second write task schedules based on the weights, and output the third write task schedule.

4. The cognitive writing system based on neural symbols according to claim 3, characterized in that, The analysis unit includes: The identification unit is used to identify the security and complexity of the current scene; The adjustment unit is used to adjust the weights of the first and second writing task scheduling based on the recognition results of the recognition unit: if it is a security constraint scenario, a higher weight is assigned to the symbolic inference engine; if it is a pattern recognition scenario, a higher weight is assigned to the neural network; if it is a mixed scenario, the weights are dynamically allocated according to the ratio of the symbolic inference engine and the neural network.

5. The cognitive writing system based on neural symbols according to claim 1, characterized in that, The decision generation module includes: The determination unit is used to determine the dependencies and resource allocation of the target to be written; An execution unit is used to perform a flush operation based on the aforementioned dependencies and resource allocation.

6. The cognitive writing system based on neural symbols according to claim 1, characterized in that, The feedback optimization module includes: The evaluation unit is used to evaluate the brushing effect through multimodal data after the brushing operation; the brushing effect includes user satisfaction, brushing success rate, average time consumption and cost; The first optimization unit is used to optimize the rule base and rule confidence of the symbolic reasoning engine based on the brushing effect. The second optimization unit is used to optimize the model parameters of the neural network based on the brushing effect.

7. A method for writing in a cognitive writing system based on neural symbols as described in claim 1, characterized in that, include: Acquire multimodal data of the vehicle, including state data, software feature data, and context data; Based on multimodal data, a first and a second write task schedule are output through a symbolic reasoning engine and a neural network, respectively; the write task schedule includes write targets and write resource allocation. The first and second flushing task schedules are merged through an attention mechanism, and the third flushing task schedule is output. Based on the third-party write task scheduling, the write operation is executed. The brushing effect is evaluated using multimodal data after the brushing operation; based on the brushing effect, the rule base of the symbolic reasoning engine and the model parameters of the neural network are optimized.

8. The writing method of the cognitive writing system based on neural symbols according to claim 7, characterized in that, The process of fusing the first and second flushing task schedules through an attention mechanism and outputting the third flushing task schedule includes: Based on scenario characteristics, the weights of the first and second write task scheduling are dynamically adjusted. Based on the weights, the first and second write task schedules are weighted and merged, and the third write task schedule is output.

9. An electronic device, comprising: One or more processors; A storage device for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the writing method of the cognitive writing system based on neural symbols as described in any one of claims 7 to 8.

10. A computer-readable medium having a computer program stored thereon, wherein, When the computer program is executed by the processor, it implements the writing method of the cognitive writing system based on neural symbols as described in any one of claims 7 to 8.

Images