Driving style identification method and system based on multi-source sensor data

Abstract

The invention discloses a driving style identification method and system based on multi-source sensor data. The method comprises the following steps: S1, acquiring the multi-source sensor data and preprocessing the multi-source sensor data; s2, constructing a symbol demand set and a symbol regulation compliance matrix to form a complete symbol library; s3, constructing a hierarchical rule tree; s4, constructing a variational Bayesian analyzable inference model; s5, updating variational Bayesian analyzable inference model parameters and a complete symbol library; and S6, online monitoring and real-time updating are carried out, and an updated driving style classification and risk assessment result is output. According to the method, the complete symbol library and the variational Bayesian analyzable inference model are constructed and are fully combined, so that the problem of realizing accurate recognition of the driving style under multiple constraint conditions is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of driving style recognition, and in particular to a driving style recognition method and system based on multi-source sensor data. Background Art

[0002] Existing driving style recognition technologies are mostly based on the comprehensive processing of vehicle driving status and environmental data. Onboard sensing devices collect vehicle speed, steering angle, accelerator and brake pedal signals, as well as road conditions and real-time traffic information. These systems typically aggregate this multi-channel data into a unified management module, then apply filtering or interpolation methods to correct for missing data and noise, allowing subsequent recognition models to operate on relatively clean input data.

[0003] Existing driving style recognition solutions typically rely on discrete processes for data processing and model inference, often lacking more refined, multi-layered representations. Because most systems only perform a one-time mapping of operational behaviors to onboard data, they struggle to timely identify different levels of driving intent or energy demands when handling complex driving conditions. Furthermore, some methods often remain at a static correction stage when responding to environmental changes and sensor anomalies, failing to flexibly adjust to temporary emergencies and high power demands. Furthermore, existing technologies often employ symbolic representations limited to limited operational labels, lacking the necessary adaptive constraints for energy management and traffic regulation compliance in real-world scenarios. This makes it difficult for systems to balance refined energy scheduling with multi-dimensional behavioral constraints.

[0004] Therefore, it is urgent to design a driving style recognition method and system based on multi-source sensor data to solve the problems existing in the above-mentioned existing technologies. Summary of the Invention

[0005] In response to the above-mentioned deficiencies or improvement needs of the prior art, the present invention provides a driving style recognition method based on multi-source sensor data. Its purpose is to solve the problem of accurately identifying driving styles under multiple constraints by constructing a complete symbol library and a variational Bayesian parsable inference model and fully combining the two.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions: A first aspect of the present invention provides a driving style recognition method based on multi-source sensor data, the method comprising the following steps: S1: Acquire multi-source sensor data and preprocess it to obtain high-quality sequences; S2: Based on the high-quality sequence, key features of multi-source sensor data are extracted, a symbol requirement set and a symbol regulation compliance matrix are constructed, and a complete symbol library is obtained; S3: Establish a mapping relationship between the complete symbol library and the high-quality sequence, perform symbol hierarchical division of the three requirements of safety, comfort, and economy, define the energy consumption coupling function, and construct a hierarchical rule tree; S4: Based on the hierarchical rule tree and combined with high-quality sequences, a multi-layer feature fusion function and probability prior distribution are established. Dynamic latent variables are introduced to characterize braking recovery and motor response. A block-by-block hierarchical approximate inference method is constructed to accelerate convergence, and a variational Bayesian analyzable inference model is obtained to output a joint posterior distribution. S5: Dynamically mapping the hierarchical rule tree to the joint posterior distribution to form a multi-layer rule loading function, establishing a hard threshold filtering boundary to filter high-risk operations, comparing the symbolic prior with the posterior, and updating the variational Bayesian analyzable inference model parameters and the complete symbolic library; S6: Monitor the vehicle's operating status online, update the variational Bayesian parsable inference model parameters and the complete symbol library in real time, and output updated driving style classification and risk assessment results.

[0007] As an embodiment of the present application, step S1 specifically includes: S11: Synchronously collect the speed of the vehicle during operation through the sensor group , acceleration , battery status , road conditions and driving operations , the collection period is unified by the time series Characterization; receive and summarize the output of each sensor through the data management unit to form a dimension of The initial observation matrix , the expression formula is as follows:

[0008] in, Indicates the Sensors at the time The original measurement value of S12: the initial observation matrix Perform time series alignment to obtain alignment matrix , the expression formula is as follows:

[0009] in, Represents the alignment function, through interpolation and interpolation, Mapping to a consistent moment , so that the initial observation matrix The rows correspond to the same , each column corresponds to the benchmark time ; S13: Suppress high-frequency random noise through sliding window filtering to form a smooth matrix , the expression formula is as follows:

[0010]

[0011] in, Indicates the Sensors at the time The smoothed corrected measurement value, Indicates the offset; S14: Yes and The difference between them is used for threshold judgment. If the threshold is met, the measurement value is marked as a mutation point and stored in the mutation point identification matrix. The expression formula is as follows:

[0012] in, represents the mutation determination threshold; Through scenario-based tagging, the recognition and retention of transient anomalies are improved, and the credibility of data in typical road conditions of electric vehicles is strengthened.

[0013] Combine the mutation point identification matrix and the smoothing matrix Make a comprehensive judgment based on vehicle operation and speed , acceleration , road conditions The correlation test at each moment The validity of the data is improved by deleting invalid data and updating the readings with reasonable correction space, and finally obtaining high-quality sequences. , .

[0014] As an embodiment of the present application, step S2 specifically includes: S21: Based on the high-quality sequence , extract vehicle speed , acceleration , battery status , road conditions and driving operations The key features of the symbol are combined with the design goals to build a symbol requirement set, and the symbol priority function is used to Quantify the key features. The specific formula is as follows:

[0015] in, Represents the index for different symbol candidates; Indicates high-quality sequences The index of different data channels in; Indicates at time Data Channel The output value of Indicates the total number of data channels, represents the sampling length, Indicates that the data channel ampersand The fusion weight of Express Perform nonlinear transformations to capture vehicle operation characteristics; S22: Based on the symbol requirement set constructed in step S21, select common driving behaviors and events such as turning, accelerating, decelerating, waiting at intersections, and speed limit areas, construct an initial symbol set, and establish a symbol reference table , to achieve fast mapping between symbolic index and semantic parsing, the formula is as follows:

[0016] in, Indicates the internal encoding of the basic symbol elements, Indicates the number of symbol types, Indicates the scalable subdivision level of each symbol; each symbol entry is numbered by the system and corresponds one-to-one with vehicle operation data, providing a semantic benchmark for energy management and driving style analysis.

[0017] S23: Introducing labels that represent micro-driving styles, adding descriptions of motor power response and braking recovery rate, and constructing energy consumption mapping functions , describing the relationship between energy consumption and symbolic labels, energy consumption mapping function The formula is as follows:

[0018] in, represents the symbol label index, Indicates a symbol, and Respectively represent the dimensions of contribution to energy consumption and energy recovery, and is the relevant weight, and is the energy management characteristic value in the data channel, and Indicates the total number of channels involved in energy input and output, and Represent the nonlinear functions for energy consumption and energy recovery respectively; S24: Introduce regulatory elements and establish a symbol compliance matrix, which is used to define each symbol based on speed limit, prohibited driving and road priority information. Assigning a regulatory compliance coefficient , and denote the symbol compliance as :

[0019] in, Symbols The corresponding regulatory compliance coefficient, Indicates the moment of union The correction amount of road supervision intensity, According to vehicle speed and road conditions Filter functions for determining violations or hidden dangers; S25: To achieve dynamic adaptation of the symbol library under different seasons or energy consumption constraints, construct the symbol adaptive update equation :

[0020] in, Indicates the initial symbol The basic attribute value when not affected by the environment, and is the adaptive tuning parameter, and Represent the time-varying variables of battery status and road conditions, and Represent the mapping functions for battery and road environment respectively, which are used to dynamically adjust the symbol attributes; S26: Based on the step-by-step progress of steps S21 to S25, a complete symbol library covering driving behavior, energy management, and traffic regulations is formed :

[0021] The complete symbol library The output includes basic symbol index, energy consumption conversion coefficient, compliance evaluation and adaptive management parameters.

[0022] As an embodiment of the present application, step S3 specifically includes: S31: In the complete symbol library With high-quality sequences A mapping relationship is established between them, and the expression formula is as follows:

[0023] in, Represents a global consistency function, the complete symbol library Include symbol entries, high-quality sequences Include channels, Indicates the Symbols and channels The fusion weight of represents the mapping function, Indicates high-quality sequences In the channel ,time The measured value, Represents the complete symbol library Middle Internal encoding of symbols; S32: Based on the vehicle operation process, the three requirements of safety, comfort and economy are divided into symbol levels, and the priority of each level is respectively expressed as 、 、 Indicates that for each moment Tiered priority The measurement formula is as follows:

[0024] in, Represents a hierarchical index, Representation level The normalization factor of ; Indicates channel For this level The inhibition coefficient; S33: Define the energy consumption coupling function based on the vehicle's motor power and battery recovery characteristics Representation level At the moment The impact of energy consumption will be Embed high-level rules to balance the system between security and economy. The formula is as follows:

[0025] in, and Respectively represent the channel indexes of energy output and braking recovery, represents the output power weight, represents the recovery power weight, and Respectively represent the vehicle at time The power output channel and recovery channel values, Representation level Benchmark energy consumption compensation item; S34: Define an operation set for specific operations such as sudden acceleration and quick steering , symbolic conditions Make a judgment:

[0026] in, Indicates a driving operation. Representation and hierarchy The corresponding operation set, if , the operation is considered to be at the level Valid symbolic expressions under S35: Targeting levels The operational risk of high power or sudden braking is determined by building a rule matching function. , the formula is as follows:

[0027] in, Indicates time At the level The control interval under Indicates the coefficient for prioritizing the operation. Represents the coefficient of energy consumption penalty; The value of the sensor is determined to be greater than or less than zero, and excessive power or sudden braking behavior is identified in real time, and energy consumption restrictions are implemented; S36: The hierarchical priority , energy consumption coupling function , Operation symbol conditions , rule matching function Integrate to form a hierarchical rule tree containing high-level rules and operational subdivisions , the formula is as follows:

[0028] in, Representation level With a driving operation The corresponding relationship.

[0029] As an embodiment of the present application, step S4 specifically includes: S41: Based on the hierarchical rule tree, combined with high-quality sequences, energy consumption-related features and driving behavior classification factors are selected, mapped to a set of modelable indicators, and a multi-layer feature fusion function is established. , the rule constraints and data channels are comprehensively quantified, and the formula is expressed as follows:

[0030] in, represents the index of the modelable indicator, Indicates the first The activation strength of the rule, Represents energy consumption related characteristics, represents the driving behavior classification factor, Represents the fusion weight constant; S42: Order A random latent parameter set representing the driving style, based on a multi-layer feature fusion function The observed values construct a probability prior distribution , the prior distribution is constrained in the high-dimensional feature space through multiple integrals, and the formula is expressed as follows:

[0031] in, represents the random parameter space, represents the total number of fusion indicators, Representation and multi-layer feature fusion function The corresponding prior penalty coefficient; S43: Define a dynamic latent variable set for the braking recovery and motor response delay of the vehicle during operation ,make Represents the energy consumption coupling state and introduces the function The formula to describe the influence of dynamic latent variables on observations is as follows:

[0032] in, Indicates time The observation vector under represents the basic observation component without energy consumption dynamics, represents the latent variable coupling weight, Used to simulate the time series characteristics of braking recovery rate and delayed power response; S44: Construct a block-by-block hierarchical approximate inference method to accelerate convergence, treat the general variable block and the energy consumption coupling variable block separately, and define the overall posterior distribution The factorization pattern of is:

[0033] in, Indicates the A parameter set for a common block, Indicates the The hidden variable set of energy consumption coupling blocks, 、 Indicates the number of blocks; Using the variational expectation-maximization method and Continuous iterative updates; S45: Define scene focus functions for extreme road conditions or high power requirements , by increasing the local sampling frequency to refine the estimation accuracy of the corresponding latent variables, the formula is expressed as follows:

[0034] in, Indicates time In local sampling areas under extreme road conditions, represents a high-weight amplification factor, Reflects the degree of scene triggering, represents the key suppression coefficient, Represents a measure of energy limits or high load areas; S46: Combining the above steps, we can obtain a variational Bayesian analyzable inference model, which is based on a dynamic latent variable set. With random hidden parameter set The combined random variable As the core, through integration Input, using the combined joint posterior distribution To express the final inference result, the formula is as follows:

[0035] in, represents the likelihood probability distribution of the observed data, Kullback-Leibler divergence, used to measure the overall posterior distribution and the true distribution The difference between.

[0036] As an embodiment of the present application, step S5 specifically includes: S51: The hierarchical rule tree and the joint posterior distribution Create dynamic mappings for symbol labels With hidden variables Correlation function between Conduct comprehensive assessment to form a multi-layer rule loading function :

[0037] in, Indicates the index corresponding to the symbol rule in the hierarchical set, Load weights for symbolic rules, is the joint posterior distribution, the correlation function Rules for measuring symbols For latent variables The strength of the constraint; S52: Establishing hard threshold filtering boundaries for vehicle operation behavior , used to determine whether excessive or high-risk behavior violates layered safety requirements. When the value is greater than the given threshold, the rule check is immediately executed. The specific formula is as follows:

[0038] in, Represents the level, Indicates vehicle operation instructions. is the operation activation weight, Characterization Operation At the level The legality of For latent variables At the level The compensation factor, Represents the network hidden variable domain that satisfies the current level constraints; S53: By comparing the time and time The posterior distribution of is used to judge the transition confidence and define the distribution difference ,like If the value is greater than the specified threshold, it is determined that the driving style has changed significantly. The specific formula is as follows:

[0039] in, 、 Respectively indicate time and time The posterior distribution of represents the Kullback-Leibler divergence; S54: Prior symbol distribution The symbol probability obtained by actual inference For comparison:

[0040] Among them, if If the threshold is exceeded, a collaborative correction is triggered. By analyzing the cause of the difference, the symbol constraints are adjusted or driving risks are prompted, achieving a two-way update and iteration of the complete symbol library and the variational Bayesian parsable inference model. S55: Define multiple loss functions Represents the error metric of the variational Bayesian analytic inference model in different scene dimensions, and the error metric of the variational Bayesian analytic inference model parameters Perform dynamic correction, the formula is as follows:

[0041] in, and Represent the error component index of energy consumption coupling and driving style, is the learning step coefficient, Express gradient; S56: Define a warning decision function for extreme driving operations that cause the variational Bayesian analytic inference model to fail to converge , if the warning judgment function If the threshold is exceeded, an alert is issued and the final revision results of the variational Bayesian parseable inference model and the complete symbol library are output. The formula is as follows:

[0042] in, Indicates a high-risk area focused on extreme driving power or speeding situations, Represents the detection function corresponding to key factors such as braking and acceleration, is the trade-off coefficient.

[0043] As an embodiment of the present application, step S6 specifically includes: S61: Filter out the vehicle's current main operating characteristics and combine them with the energy consumption safety range to form a basic constraint set and define the initial synthesis state , the formula is as follows:

[0044] in, Indicates the driving style label score initially identified by the vehicle, Indicates the proportion of remaining energy consumption, and is the adjustable weight coefficient; S62: Real-time sampling of vehicle speed, current output and pedal operation in a short period of time, and based on the joint posterior distribution Determine whether the vehicle has style deviation; if the driving parameters are monitored As the difference between them gradually increases, the style revision process is triggered, and the local posterior state is updated to adapt to the latest observation; S63: When high impact power or emergency braking abnormal operation is detected, a quick matching relationship is established with the complete symbol library, the corresponding symbol entry is found and the relevant cumulative number is recorded; if the abnormal entries accumulate rapidly within a short period of time, it is recorded as a high-risk operation trend and bound to the current operating characteristics of the vehicle; S64: For the high-risk operation trend, if the cumulative number of times exceeds a preset threshold, the situation is marked as a potential safety or energy consumption risk and an internal alarm signal is issued; S65: When external road conditions and driving style continue to change, the variational Bayesian parsable inference model and the complete symbol library are updated online and synchronously; S66: After completing multiple online revisions to the variational Bayesian parsable inference model and the complete symbol library, the vehicle's final energy consumption information and abnormal operation registration are integrated to produce the latest driving style classification and risk assessment results.

[0045] As an embodiment of the present application, the step S66 further includes the final discrimination score A simplified measurement is performed to directly determine the degree of deviation between the vehicle's current state and the target safety boundary. The formula is as follows:

[0046] in, Indicates the number of abnormal registrations. is the abnormality count penalty coefficient, if If it is lower than the safety threshold, it means that the vehicle behavior has deviated significantly and the intervention strategy needs to be strengthened appropriately. Otherwise, the normal driving style label and risk assessment result will be output.

[0047] The present invention also provides a driving style recognition system based on multi-source sensor data, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-mentioned driving style recognition method based on multi-source sensor data.

[0048] The beneficial effects of the present invention are: 1. This invention collects multi-source sensor data and ensures the integrity and effectiveness of observation information through unified time series management and mutation point smoothing. During the preprocessing process, data such as speed, acceleration, battery status, and road conditions are time-aligned and filtered with sliding windows. In addition, instantaneous anomalies are marked by threshold judgment. The resulting high-quality sequence meets physical rationality at every moment and provides a stable data foundation for subsequent symbol library creation and hierarchical reasoning.

[0049] 2. The present invention extracts features based on vehicle speed, acceleration, battery status, road conditions, and driving maneuvers, and then combines them with regulatory compliance and dynamic parameter adjustments to tightly bind actual operational behaviors to symbolic labels. This not only includes common indicators such as steering, acceleration, and waiting at intersections, but also further subdivides them based on motor power response and braking recovery rate, allowing the symbols to present the driver's micro-style characteristics at the semantic level. By monitoring changes in road conditions, seasons, and energy usage constraints, the present invention utilizes adaptive parameter adjustment equations to continuously modify symbol attributes, ensuring that they can still accurately reflect the vehicle status in high-load or energy-limited scenarios.

[0050] 3. The present invention establishes an overall framework that collaborates hierarchical parsing and variational Bayesian parsable inference models at the rule and reasoning levels, connects high-level abstract rules with detailed conditions for specific operations such as sudden acceleration or high-speed braking, and constructs a hierarchical rule tree to characterize the requirements and constraints at each level of safety, economy, and comfort. The variational Bayesian parsable inference model is then used to model the latent variables of the electric vehicle energy consumption coupling state and driving style, and block approximate inference is used to accelerate convergence. Through collaborative reasoning, the present invention can more flexibly capture high power output or emergency braking behavior when sudden working conditions occur, and immediately combine existing symbolic rules to perform short-term energy restrictions or risk warnings. The entire process is continuously iterated through online updates and local corrections to ensure that the precision of driving style recognition and the consistency of rule execution can be maintained under conditions of frequent and complex vehicle driving environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is a technical solution flow chart of a driving style recognition method based on multi-source sensor data provided in an embodiment of the present invention; Figure 2 A schematic diagram of a complete symbol library construction for a driving style recognition method based on multi-source sensor data provided in an embodiment of the present invention; Figure 3 Schematic diagram of constructing a variational Bayesian analyzable inference model for a driving style recognition method based on multi-source sensor data provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0052] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0053] Reference Figure 1-Figure 3 In a first aspect, the present invention provides a driving style recognition method based on multi-source sensor data, the method comprising the following steps: S1: Acquire multi-source sensor data and preprocess it to obtain high-quality sequences; S2: Based on the high-quality sequence, key features of multi-source sensor data are extracted, a symbol requirement set and a symbol regulation compliance matrix are constructed, and a complete symbol library is obtained; S3: Establish a mapping relationship between the complete symbol library and the high-quality sequence, perform symbol hierarchical division of the three requirements of safety, comfort, and economy, define the energy consumption coupling function, and construct a hierarchical rule tree; S4: Based on the hierarchical rule tree and combined with high-quality sequences, a multi-layer feature fusion function and probability prior distribution are established. Dynamic latent variables are introduced to characterize braking recovery and motor response. A block-by-block hierarchical approximate inference method is constructed to accelerate convergence, and a variational Bayesian analyzable inference model is obtained to output a joint posterior distribution. S5: Dynamically mapping the hierarchical rule tree to the joint posterior distribution to form a multi-layer rule loading function, establishing a hard threshold filtering boundary to filter high-risk operations, comparing the symbolic prior with the posterior, and updating the variational Bayesian analyzable inference model parameters and the complete symbolic library; S6: Monitor the vehicle's operating status online, update the variational Bayesian parsable inference model parameters and the complete symbol library in real time, and output updated driving style classification and risk assessment results.

[0054] Specifically, the present invention is applicable to various types of motor vehicles, including but not limited to electric vehicles, fuel vehicles, hybrid vehicles and self-driving vehicles, and realizes dynamic recognition and optimization of driving style through multi-source sensor data fusion and symbolic rule reasoning.

[0055] As an embodiment of the present application, step S1 specifically includes: S11: Synchronously collect the speed of the vehicle during operation through the sensor group , acceleration , battery status , road conditions and driving operations , the collection period is unified by the time series Characterization; receive and summarize the output of each sensor through the data management unit to form a dimension of The initial observation matrix , the expression formula is as follows:

[0056] in, Indicates the Sensors at the time The original measurement value of .

[0057] Specifically, the sensor group includes a vehicle speed sensor, an acceleration sensor, a voltage sensor, a current sensor, a road friction coefficient or pothole detection sensor, and a brake pressure sensor.

[0058] S12: the initial observation matrix Perform time series alignment to obtain alignment matrix , the expression formula is as follows:

[0059] in, Represents the alignment function, which is used to define the mapping relationship. Specifically, through interpolation and interpolation, Mapping to a consistent moment , so that the initial observation matrix The rows correspond to the same , each column corresponds to the benchmark time .

[0060] S13: Suppress high-frequency random noise through sliding window filtering to form a smooth matrix , the expression formula is as follows:

[0061]

[0062] in, Indicates the Sensors at the time The smoothed corrected measurement value, Indicates the offset.

[0063] S14: Yes and The difference between them is used for threshold judgment. If the threshold is met, the measurement value is marked as a mutation point and stored in the mutation point identification matrix. The expression formula is as follows:

[0064] in, represents the mutation determination threshold; Specifically, when a mutation point is detected, it is not always discarded as a fault. Instead, a scenario annotation is made for the driving environment at that moment based on the typical operating conditions of the vehicle, such as regenerative braking, uphill braking, sudden acceleration and starting, and stopping / starting at intersections. In this way, instantaneous fluctuations that truly belong to the vehicle's normal rapid response or energy recovery process can be identified and retained, rather than simply deleting the scenario markers. Therefore, through the above-mentioned scenario markers, the recognition and retention of instantaneous abnormalities of vehicles, especially electric vehicles, are improved, and the credibility of the data in typical road conditions of electric vehicles is enhanced.

[0065] Combine the mutation point identification matrix and the smoothing matrix Make a comprehensive judgment, delete invalid data and update the readings with reasonable correction space, and adjust the readings according to vehicle operation. and speed , acceleration , road conditions The correlation test at each moment The effectiveness of the final high-quality sequence , .

[0066] The comprehensive judgment is to compare the mutation point mark generated in the previous step with the value after smoothing the sliding window: if a mutation point is marked as a typical scene and its numerical deviation is within the acceptable correction range, the original reading is replaced by the smoothed value; if the mutation point does not meet any scene labels and deviates too much from the smoothed values of the surrounding moments, it is judged as failure or noise and the data is directly discarded.

[0067] In addition, according to the vehicle operation and speed , acceleration , road conditions The correlation test at each moment The effectiveness of the system is specifically as follows: at each moment, the degree of match between the driving operation (accelerator, brake, steering) and the speed / acceleration / road conditions at that moment is checked. For example, deceleration or negative acceleration should occur when there is braking input; the change in accelerator pedal depth should correspond to positive acceleration; on flat roads, the vehicle speed fluctuation should not be too large; if the consistency between the command and the physical feedback is higher than the preset threshold, the data at that moment is considered valid; if the consistency is insufficient, it is considered that there is an anomaly at that point, and the upper-level rules are eliminated again.

[0068] Specifically, the present invention takes the processing of multi-source sensor data during vehicle driving as the starting point, and ensures the integrity and effectiveness of observation information through unified time series management and mutation point smoothing; during the processing, the speed, acceleration, battery status, road conditions and other data are time-aligned and filtered with sliding windows, and instantaneous anomalies are marked in combination with threshold judgment; the high-quality sequence formed in this way meets physical rationality at every moment, and provides a stable data foundation for subsequent symbol creation and hierarchical reasoning. Compared with conventional single or simplified preprocessing methods, this technical solution also specifically introduces a scenario-based labeling mechanism to further evaluate the credibility of the detected potential mutation data, ensuring that the identification of temporary anomalies in typical road conditions is not simply filtered, thereby improving the retention of extreme vehicle operations.

[0069] As an embodiment of the present application, step S2 specifically includes: S21: Based on the high-quality sequence , extract vehicle speed , acceleration , battery status , road conditions and driving operations The key features of the symbol are combined with the design goals to build a symbol requirement set, and the symbol priority function is used to Quantify the key features. The specific formula is as follows:

[0070] in, Represents the index for different symbol candidates; Indicates high-quality sequences The index of different data channels in; Indicates at time Data Channel The output value of Indicates the total number of data channels, represents the sampling length, Indicates that the data channel ampersand The fusion weight of Express Nonlinear transformations are performed to capture vehicle operation characteristics.

[0071] The design goal is to closely align the extraction of key features and the construction of a symbol requirement set with the application requirements that the system will subsequently address, ensuring that the selected features can both maximize the reflection of the vehicle's operating status and transient anomalies under typical operating conditions and meet the requirements for real-time computing, low false alarms, and stable and reliable system performance. Specifically, this includes: anomaly recognition accuracy: ensuring that the constructed symbols can prominently reflect transient vehicle anomalies or signs of failure; scenario credibility: symbols are required to stably extract valid information under typical road conditions and operating modes; real-time performance and computational overhead: the selected features must be easy to calculate and update quickly to meet online operation requirements; and system reliability: balancing false alarm and missed alarm rates to ensure the reliable execution of subsequent diagnostic or control strategies.

[0072] S22: Based on the symbol requirement set constructed in step S21, select common driving behaviors and events such as turning, accelerating, decelerating, waiting at intersections, and speed limit areas, construct an initial symbol set, and establish a symbol reference table , to achieve fast mapping between symbolic index and semantic parsing, the formula is as follows:

[0073] in, Indicates the internal encoding of the basic symbol elements, Indicates the number of symbol types, Indicates the scalable subdivision level of each symbol, which is used to accurately distinguish between operations such as steering, acceleration, and deceleration in the system.

[0074] S23: Next, we introduce labels that represent micro-driving styles, such as short-term high-output conditions, add descriptions of motor power response and braking recovery rate, and construct an energy consumption mapping function. , describing the relationship between energy consumption and symbolic labels, energy consumption mapping function The formula is as follows:

[0075] in, represents the symbol label index, Representation symbol; and Respectively represent the dimensions of contribution to energy consumption and energy recovery, and is the relevant weight, and is the energy management characteristic value in the data channel, and Indicates the total number of channels involved in energy input and output, and Represent the nonlinear functions for energy consumption and energy recovery respectively; Through the energy consumption mapping function Reflect energy consumption changes at the symbolic level, providing a measurable basis for hierarchical reasoning.

[0076] S24: Then introduce regulatory elements and establish a symbol compliance matrix to help the system identify the matching between driving behavior and road regulations. Assigning a regulatory compliance coefficient , and denote the symbol compliance as :

[0077] in, Symbols The corresponding regulatory compliance coefficient, Indicates the moment of union The correction amount of road supervision intensity, According to vehicle speed and road conditions Filter functions for determining violations or hidden dangers; The symbol compliance matrix is used to determine whether driving behavior complies with traffic management requirements, and illegal operations are prompted or blocked based on the compliance level. According to the rule verification, if the symbol does not comply with the rules, it will be corrected by fallback: the physical range, logical consistency and cross-variable correlation of the quantized symbol are checked; if failure occurs, the symbol at that moment is temporarily restored to the sliding window filter value, and its fusion weight or threshold is adjusted; regeneration and verification: requantization and re-verification are carried out, and it is repeated for a maximum of a fixed number of iterations. If it still does not comply, the symbol at that moment is discarded.

[0078] S25: Different seasons or special energy consumption restrictions will cause frequent changes in vehicle status and road demand. Therefore, in order to achieve dynamic adaptation of the symbol library under different seasons or special energy consumption restrictions, a symbol adaptive update equation is constructed. :

[0079] in, Indicates the initial symbol The basic attribute value when not affected by the environment, and is the adaptive tuning parameter, and Represent the time-varying variables of battery status and road conditions, and They represent the mapping functions for the battery and road environment, respectively, and are used to dynamically adjust symbol attributes.

[0080] S26: Based on the step-by-step progress of steps S21 to S25, a complete symbol library covering driving behavior, energy management, and traffic regulations is formed :

[0081] The complete symbol library The output includes basic symbol index, energy consumption conversion coefficient, compliance evaluation and adaptive management parameters, providing semantic priors for control strategies and reasoning algorithms.

[0082] Specifically, this application builds a complete symbol library for driving behavior and energy management, extracts features of vehicle speed, acceleration and battery parameters, and combines regulatory compliance with dynamic parameter adjustment items to tightly bind actual operating behaviors with symbol labels. This not only includes common signs such as turning, acceleration, and waiting at intersections, but also further subdivides motor power response and braking recovery rate, so that the symbols can present the driver's micro-style characteristics at the semantic level. Compared with the traditional approach of relying solely on static symbol judgment, this solution will monitor changes in roads, seasons and energy usage constraints, and use adaptive parameter adjustment equations to modify symbol attributes at any time, so that it can still accurately reflect the vehicle status in high-load or energy-limited scenarios.

[0083] As an embodiment of the present application, step S3 specifically includes: S31: In the complete symbol library With high-quality sequences A mapping relationship is established between them, and a global measurement of symbolic features and data consistency is performed to form a comprehensive input suitable for subsequent hierarchical analysis. The expression formula is as follows:

[0084] in, Represents the global consistency function, used to measure Each symbol in the high-quality sequence The degree of coupling between Include symbol entries; high-quality sequences Include channels; Indicates the Symbols and channels The fusion weight of Represents the mapping function, which is used to calculate the degree of association between symbols and data; Indicates high-quality sequences In the channel ,time The measured value of Represents the complete symbol library Middle Internal encoding of symbols; By calculation , evaluate the degree of coupling between symbols and data, and screen out symbol entries with high consistency.

[0085] S32: Based on the vehicle operation process, the three requirements of safety, comfort and economy are divided into symbol levels, and the priority of each level is respectively expressed as 、 、 Indicates that for each moment Tiered priority To measure, Represents the hierarchical index, and the formula is as follows:

[0086] in, Representation level The normalization factor of ; Indicates channel For this level The suppression coefficient is used to control the fluctuation of the priority of each level under specific operating conditions. The larger the value, the higher the priority of the level.

[0087] S33: Define the energy consumption coupling function based on the motor power and battery recovery characteristics unique to electric vehicles Representation level At the moment The impact of energy consumption will be Embed high-level rules to balance the system between security and economy. The formula is as follows:

[0088] in, and Respectively represent the channel indexes of energy output and braking recovery, represents the output power weight, represents the recovery power weight, and Respectively represent the vehicle at time The power output channel and recovery channel values, Representation level The baseline energy consumption compensation item is used to keep the overall energy consumption stable; Through the energy consumption coupling function and layer priority Jointly guide high-level rule priorities and comprehensively balance security and economic demands.

[0089] S34: Based on the high-level rules and energy consumption coupling results in the previous step, define an operation set for specific operations such as sudden acceleration and quick steering , symbolic conditions Make a judgment:

[0090] in, Indicates a driving operation. Representation and hierarchy The corresponding operation set, if , the operation is considered to be at the level Valid symbolic expressions under The downward refinement approach enables hierarchical rules to be implemented in actual operating scenarios, combining high-level abstract requirements with vehicle dynamics processes, and solving the problem of insufficient precision in rule segmentation in existing technologies.

[0091] S35: Targeting levels The operational risk of high power or sudden braking is determined by building a rule matching function. Once a large current shock or a sudden increase in braking occurs, short-term energy limitation is triggered. The formula is as follows:

[0092] in, Indicates time At the level The control interval under Indicates the coefficient for prioritizing the operation. Represents the coefficient of energy consumption penalty; The value of is greater than zero or less than zero, and excessive power or sudden braking behavior is identified in real time and energy consumption restrictions are implemented to overcome the blind spot problem under external environmental fluctuations and realize refined energy management and control.

[0093] S36: The hierarchical priority in the above step is , energy consumption coupling function , Operation symbol conditions , rule matching function Integrate to form a hierarchical rule tree containing high-level rules and operational subdivisions , the formula is as follows:

[0094] in, Representation level With a driving operation The corresponding relationship, hierarchical rule tree Output the prior constraints for subsequent variational Bayesian inferential analytical model reasoning in a graph-structured manner.

[0095] As an embodiment of the present application, step S4 specifically includes: S41: Based on the hierarchical rule tree, combined with high-quality sequences, energy consumption-related features and driving behavior classification factors are selected, mapped to a set of modelable indicators, and a multi-layer feature fusion function is established. , which is used to comprehensively quantify rule constraints and data channels. The formula is as follows:

[0096] in, represents the index of the modelable indicator, Indicates the first The activation strength of the rule, Represents energy consumption related characteristics, represents the driving behavior classification factor, Represents the fusion weight constant.

[0097] S42: Based on the historical driving records of step S1 and the hierarchical rule tree of step S3, A random latent parameter set representing the driving style, based on a multi-layer feature fusion function The observed values construct a probability prior distribution , the prior distribution is constrained in the high-dimensional feature space through multiple integrals, and the formula is expressed as follows:

[0098] in, represents the random parameter space, represents the total number of fusion indicators, Representation and multi-layer feature fusion function The corresponding prior penalty coefficient is used when the network is initialized, while ensuring the diversity of vehicle behavior. Normalization is performed so that the initial distribution can reflect the characteristics of driving style while also providing sufficient discrimination to support variational inference.

[0099] S43: Define a dynamic latent variable set for the vehicle's braking recovery and motor response delay during operation ,make Represents the energy consumption coupling state, dynamic hidden variable set By using a random hidden parameter set Combine to form a joint random variable , improve the extended characterization of energy consumption dimension, introduce function The formula to describe the influence of dynamic latent variables on observations is as follows:

[0100] in, Indicates time The observation vector under represents the basic observation component without energy consumption dynamics, represents the latent variable coupling weight, Used to simulate the timing characteristics of braking recovery rate and delayed power response.

[0101] S44: For joint random variables As the dimension increases and the traditional inference speed is limited, a block-by-block approximate inference method is constructed to accelerate convergence. The general variable block and the energy consumption coupling variable block are processed separately to define the overall posterior distribution. The factorization pattern of is:

[0102] in, Indicates the A parameter set for a common block, Indicates the The hidden variable set of the energy consumption coupling variable block, 、 Indicates the number of blocks, Indicates the Universal Block The variational posterior distribution approximation factor of , Indicates the Energy consumption coupling variable block The variational posterior distribution approximation factor of ; Using the variational expectation-maximization method and Continuous iterative updates.

[0103] S45: Define the scene focus function for extreme road conditions or high power requirements detected during the inference process , by increasing the local sampling frequency to refine the estimation accuracy of the corresponding latent variables, the formula is expressed as follows:

[0104] in, Indicates time In local sampling areas under extreme road conditions, represents a high-weight amplification factor, Reflects the degree of scene triggering, represents the key suppression coefficient, Represents a measure of energy limits or high load areas; Whenever the scene focus function When it is greater than the set threshold, it enters the specific scene focus mode, increases the sampling density in the corresponding latent variable dimension and performs further refined inference. The set threshold is an empirical hyperparameter. and Normalization, then in history The sample is determined by quantiles (such as the 95% quantile).

[0105] S46: Combining the above steps, a variational Bayesian analyzable inference model is obtained that considers the coupling relationship between driving operation and energy consumption while retaining multi-layer rule constraints. The model is based on the joint random variable As the core, through integration Input, lays an effective prior for subsequent real-time style recognition and collaborative correction, and uses the combined joint posterior distribution To express the final inference result, the formula is as follows:

[0106] in, represents the likelihood probability distribution of the observed data, Kullback-Leibler divergence, used to measure the overall posterior distribution With prior distribution The difference between.

[0107] As an embodiment of the present application, step S5 specifically includes: S51: The hierarchical rule tree and the joint posterior distribution Create dynamic mappings for symbol labels With hidden variables Correlation function between Conduct comprehensive assessment to form a multi-layer rule loading function :

[0108] in, Indicates the index corresponding to the symbol rule in the hierarchical set, Load weights for symbolic labels, is the joint posterior distribution, the correlation function For metric symbol labels For latent variables The strength of the constraint; Loading functions through multiple layers of rules Obtain a coupled expression of symbolic rules and network inference process to ensure that both vehicle dynamics and hierarchical semantics are taken into account.

[0109] S52: Establishing hard threshold filtering boundaries for vehicle operation behavior , used to determine whether excessive or high-risk behavior violates layered safety requirements. When the value is greater than the given threshold (95% quantile), the rule check is immediately executed. The specific formula is as follows:

[0110] in, Represents the level, Indicates a driving operation. is the operation activation weight, Representation symbolic conditions, For latent variables At the level The compensation factor, Represents the network hidden variable domain that satisfies the current level constraints.

[0111] S53: By comparing the time and time The posterior distribution of is used to judge the transition confidence and define the distribution difference ,like If the value is greater than the specified threshold (95% percentile), it is determined that the driving style has changed significantly, and the monitoring process will change direction, combining amplitude and persistence. The specific formula is as follows:

[0112] in, 、 Respectively indicate time and time The posterior distribution of represents the Kullback-Leibler divergence; S54: When there is a large difference between the posterior inference and the symbol prior, it is determined whether the driver is performing high-risk operations or the rule tolerance deviation will be increased by the prior symbol probability. The symbol probability obtained by actual inference For comparison:

[0113] in, Represents the distance between the prior symbol probability vector and the actual inferred posterior symbol probability vector. If If the threshold (95th percentile) is exceeded, a coordinated correction is triggered. By analyzing the cause of the difference, the symbol constraints are adjusted or driving risk is indicated, achieving a two-way update and iteration of the complete symbol library and the variational Bayesian parsable inference model. S55: Define multiple loss functions Represents the error metric of the variational Bayesian analytic inference model in different scene dimensions, and the error metric of the variational Bayesian analytic inference model parameters Perform dynamic correction, the formula is as follows:

[0114] in, and Represent the error component index of energy consumption coupling and driving style respectively, is the learning step coefficient, Express gradient; When the deviation is serious or the external environment fluctuates violently, the variational Bayesian can analytically infer the model parameters. Make dynamic corrections and continuous corrections It can maintain the accuracy of coupling energy consumption calculation with driving behavior in high-speed scenarios and changing road conditions.

[0115] Specifically, this application establishes an overall framework of hierarchical parsing and variational Bayesian collaboration at the rule and reasoning level, connects high-level abstract rules with detailed conditions for specific operations such as sudden acceleration or high-speed braking, and generates a hierarchical rule tree to represent the needs and constraints of various levels of safety, economy, and comfort. It then uses a variational Bayesian network to model the latent variables of the coupled state of electric vehicle energy consumption and driving style, and accelerates convergence by block approximate inference. Compared with the method of using only probabilistic models or only symbolic classification, the collaborative reasoning method can more flexibly capture high power output or emergency braking behavior when sudden working conditions occur, and immediately combine existing symbolic rules for short-term energy restriction or risk warnings. The entire process is continuously iterated through online updates and local corrections to ensure that the system can maintain the precision of driving style recognition and the consistency of rule execution under frequent and complex vehicle driving environments.

[0116] S56: If the variational Bayesian analytic inference model fails to converge due to extreme driving operations, a risk warning will be triggered and the final updated model parameters will be Write the posterior distribution back to the system for downstream modules to query; define the warning judgment function , if the warning judgment function If the threshold is exceeded, an alert is issued and the final revision results of the variational Bayesian parseable inference model and the complete symbol library are output. The formula is as follows:

[0117] in, Indicates a high-risk area focused on extreme driving power or speeding situations, Respectively represent the detection functions corresponding to braking, acceleration, and actual wheel-end power, is the trade-off coefficient; when If the threshold (95% percentile) is exceeded, a warning will be issued and the final revision results of the variational Bayesian parsable inference model and the complete symbol library will be output, providing a direct basis for subsequent real-time style identification and risk intervention.

[0118] As an embodiment of the present application, on the basis of ensuring the coordination of multi-layer rules and variational Bayesian posterior distribution, real-time online updating is performed, and step S6 specifically includes: S61: After completing the multi-layer rule loading and risk correction in the above steps, the vehicle's current main operating characteristics are screened out, including real-time vehicle speed, acceleration, steering angle, and recent abnormal operations. These characteristics are combined with the energy consumption safety range to form a basic constraint set to guide real-time judgment and update in subsequent short cycles.

[0119] To strengthen the unified constraint, define the initial synthesis state , including the weighted combination of driving style weight and energy consumption remaining metric, the formula is expressed as follows:

[0120] in, Indicates the driving style label score initially identified by the vehicle, Indicates the proportion of remaining energy consumption, and is the adjustable weight coefficient; S62: Real-time sampling of vehicle speed, current output and pedal operation in a short period of time, and based on the joint posterior distribution Determine whether the vehicle has style deviation; if the driving parameters are monitored As the difference between them increases, the style revision process is triggered, and the local posterior state is updated to adapt to the latest observation; by retaining the initial synthesis state The constraints of the short-term observations are integrated with the historical accumulated information, taking into account both instantaneous anomalies and the continuity of the overall behavior.

[0121] S63: When high impact power or emergency braking sudden abnormal operation is monitored, a quick matching relationship is established with the complete symbol library, the corresponding symbol entry is found and the relevant cumulative number is registered; if the abnormal entries accumulate quickly in a short period of time, it is recorded as a high-risk operation trend and bound to the current operating characteristics of the vehicle.

[0122] S64: For the high-risk operation trend, if the cumulative number exceeds a preset threshold, the situation will be marked as a potential safety or energy consumption risk and an internal alarm signal will be issued; combined with the actual vehicle acceleration and electrical fault warning information, it will be evaluated whether the vehicle needs to be immediately speed-limited, torque-reduced or warned, so as to intervene in high energy consumption and safety hazards at an early stage and improve driving reliability.

[0123] The pre-set threshold is an empirical threshold given according to the severity of the event: Mild high-risk operation: ≥5 times within 10 minutes; Moderately high-risk operation: ≥3 times within 10 minutes; Severe high-risk operation: ≥2 times within 30 minutes.

[0124] S65: When external road conditions and driving styles continue to change, the variational Bayesian parsable inference model and the complete symbol library are corrected online and synchronously; ensuring that the system maintains accuracy under high load or extreme environments; setting the online update rate and using the latest sampling of vehicle status and abnormal operations to fine-tune the key latent variables of the variational Bayesian parsable inference model and the weights of the complete symbol library, allowing them to iterate with fluctuations in monitoring data.

[0125] S66: After completing multiple online revisions to the variational Bayesian parsable inference model and the complete symbol library, the vehicle's final energy consumption information and abnormal operation registration are integrated to produce the latest driving style classification and risk assessment results.

[0126] As an embodiment of the present application, the step S66 further includes the final discrimination score A simplified measurement is performed to directly determine the degree of deviation between the vehicle's current state and the target safety boundary. The formula is as follows:

[0127] in, Indicates the number of abnormal registrations. is the abnormality count penalty coefficient, if If it is lower than the safety threshold, it means that there is a large deviation in the vehicle behavior and the intervention strategy needs to be moderately strengthened. Otherwise, the normal driving style label and risk assessment results will be output to provide dynamic support for downstream adaptive driving strategies and energy consumption management.

[0128] The present invention also provides a driving style recognition system based on multi-source sensor data, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-mentioned driving style recognition method based on multi-source sensor data.

[0129] By fully combining symbolic driving behavior representation with variational Bayesian inference strategies, the present invention enables the system to retain the differentiated requirements of multi-layer rules for safety, economy, and comfort, while also tracking energy consumption factors in real time under different working conditions and making rapid judgments on abnormal operations, thereby solving the technical problem of how to accurately identify driving styles under multiple constraints.

[0130] The above descriptions are merely some preferred embodiments of the present disclosure and illustrate the underlying technical principles. Those skilled in the art should understand that the scope of the invention encompassed by the embodiments of the present disclosure is not limited to technical solutions formed by specific combinations of the aforementioned technical features. It also encompasses other technical solutions formed by any combination of the aforementioned technical features or their equivalents, without departing from the aforementioned inventive concept. For example, a technical solution formed by replacing the aforementioned features with (but not limited to) technical features with similar functions disclosed in the embodiments of the present disclosure.

Claims

1. A driving style recognition method based on multi-source sensor data, characterized in that: The method comprises the following steps: S1: Acquire multi-source sensor data and preprocess it to obtain high-quality sequences; S2: Based on the high-quality sequence, key features of multi-source sensor data are extracted, a symbol requirement set and a symbol regulation compliance matrix are constructed, and a complete symbol library is obtained; S3: Establish a mapping relationship between the complete symbol library and the high-quality sequence, perform symbol hierarchical division of the three requirements of safety, comfort, and economy, define the energy consumption coupling function, and construct a hierarchical rule tree; S4: Based on the hierarchical rule tree and combined with high-quality sequences, a multi-layer feature fusion function and probability prior distribution are established. Dynamic latent variables are introduced to characterize braking recovery and motor response. A block-by-block hierarchical approximate inference method is constructed to accelerate convergence, and a variational Bayesian analyzable inference model is obtained to output a joint posterior distribution. S5: Dynamically mapping the hierarchical rule tree to the joint posterior distribution to form a multi-layer rule loading function, establishing a hard threshold filtering boundary to filter high-risk operations, comparing the symbolic prior with the posterior, and updating the variational Bayesian analyzable inference model parameters and the complete symbolic library; S6: Monitor the vehicle's operating status online, update the variational Bayesian parsable inference model parameters and the complete symbol library in real time, and output updated driving style classification and risk assessment results.

2. The driving style recognition method based on multi-source sensor data according to claim 1, characterized in that: The step S1 specifically includes: S11: Synchronously collect the speed of the vehicle during operation through the sensor group , acceleration , battery status , road conditions and driving operations , the collection period is unified by the time series Characterization; receive and summarize the output of each sensor through the data management unit to form a dimension of The initial observation matrix , the expression formula is as follows: in, Indicates the Sensors at the time The original measurement value of S12: the initial observation matrix Perform time series alignment to obtain alignment matrix , the expression formula is as follows: in, Represents the alignment function, through interpolation and interpolation, Mapping to a consistent moment ; S13: Suppress high-frequency random noise through sliding window filtering to form a smooth matrix , the expression formula is as follows: in, Indicates the Sensors at the time The smoothed corrected measurement value, Indicates the offset; S14: Yes and The difference between them is used for threshold judgment. If the threshold is met, the measurement value is marked as a mutation point and stored in the mutation point identification matrix. The expression formula is as follows: in, represents the mutation determination threshold; Combine the mutation point identification matrix and the smoothing matrix Make a comprehensive judgment based on vehicle operation and speed , acceleration , road conditions The correlation test at each moment The validity of the data is improved by deleting invalid data and updating the readings with reasonable correction space, and finally obtaining high-quality sequences. , .

3. The driving style recognition method based on multi-source sensor data according to claim 2, characterized in that: The step S2 specifically includes: S21: Based on the high-quality sequence , extract vehicle speed , acceleration , battery status , road conditions and driving operations The key features of the symbol requirement set are constructed by the symbol priority function Quantify the key features. The specific formula is as follows: in, Represents the index for different symbol candidates; Indicates high-quality sequences The index of different data channels in; Indicates at time Data channel The output value of Indicates the total number of data channels, represents the sampling length, Indicates that the data channel ampersand The fusion weight of Express Perform nonlinear transformations to capture vehicle operation characteristics; S22: Based on the symbol requirement set constructed in step S21, select common driving behaviors and events such as turning, accelerating, decelerating, waiting at intersections, and speed limit areas, construct an initial symbol set, and establish a symbol reference table , to achieve fast mapping between symbolic index and semantic parsing, the formula is as follows: in, Indicates the internal encoding of the basic symbol elements, Indicates the number of symbol types, Indicates the degree of subdivision to which each symbol can be expanded; S23: Introducing labels that represent micro-driving styles, adding descriptions of motor power response and braking recovery rate, and constructing energy consumption mapping functions , describing the relationship between energy consumption and symbolic labels, energy consumption mapping function The formula is as follows: in, represents the symbol label index, Indicates a symbol, and Respectively represent the dimensions of contribution to energy consumption and energy recovery, and is the relevant weight, and is the energy management characteristic value in the data channel, and Indicates the total number of channels involved in energy input and output, and Represent the nonlinear functions for energy consumption and energy recovery respectively; S24: Introduce regulatory elements and establish a symbol compliance matrix, which is used to define each symbol based on speed limit, prohibited driving and road priority information. Assigning a regulatory compliance coefficient , and denote the symbol compliance as : in, Symbols The corresponding regulatory compliance coefficient, Indicates the moment of union The correction amount of road supervision intensity, According to vehicle speed and road conditions Filter functions for determining violations or hidden dangers; S25: Constructing symbolic adaptive update equations , to achieve dynamic adaptation of the symbol library under different seasons or energy consumption restrictions. The expression formula is as follows: in, Indicates the initial symbol The basic attribute value when not affected by the environment, and is the adaptive tuning parameter, and Represent the time-varying variables of battery status and road conditions, and Represent the mapping functions for battery and road environment respectively, which are used to dynamically adjust the symbol attributes; S26: Based on the step-by-step progress of steps S21 to S25, a complete symbol library covering driving behavior, energy management, and traffic regulations is formed : The complete symbol library The output includes basic symbol index, energy consumption conversion coefficient, compliance evaluation and adaptive management parameters.

4. The driving style recognition method based on multi-source sensor data according to claim 3, characterized in that: The step S3 specifically includes: S31: In the complete symbol library With high-quality sequences A mapping relationship is established between them, and the expression formula is as follows: in, Represents a global consistency function, the complete symbol library Include symbol entries, high-quality sequences Include channels, Indicates the Symbols and channels The fusion weight of represents the mapping function, Indicates high-quality sequences In the channel ,time The measured value, Represents the complete symbol library Middle Internal encoding of symbols; S32: Based on the vehicle operation process, the three requirements of safety, comfort and economy are divided into symbol levels, and the priority of each level is respectively expressed as 、 、 Indicates that for each moment Tiered priority The measurement formula is as follows: in, Represents a hierarchical index, Representation level The normalization factor of ; Indicates channel For this level The inhibition coefficient; S33: Define the energy consumption coupling function based on the vehicle's motor power and battery recovery characteristics Representation level At the moment The impact of energy consumption will be Embed high-level rules to balance the system between security and economy. The formula is as follows: in, and Respectively represent the channel indexes of energy output and braking recovery, represents the output power weight, represents the recovery power weight, and Respectively represent the vehicle at time The power output channel and recovery channel values, Representation level Benchmark energy consumption compensation item; S34: Define an operation set for specific operations such as sudden acceleration and quick steering , symbolic conditions Make a judgment: in, Indicates a driving operation. Representation and hierarchy The corresponding operation set, if , the operation is considered to be at the level Valid symbolic expressions under S35: Targeting levels The operational risk of high power or sudden braking is determined by building a rule matching function. , the formula is as follows: in, Indicates time At the level The control interval under Indicates the coefficient for prioritizing the operation. Represents the coefficient of energy consumption penalty; The value of the sensor is determined to be greater than or less than zero, and excessive power or sudden braking behavior is identified in real time, and energy consumption restrictions are implemented; S36: The hierarchical priority , energy consumption coupling function , Operation symbol conditions , rule matching function Integrate to form a hierarchical rule tree containing high-level rules and operational subdivisions , the formula is as follows: in, Representation level With a driving operation The corresponding relationship.

5. The driving style recognition method based on multi-source sensor data according to claim 4, characterized in that: The step S4 specifically includes: S41: Based on the hierarchical rule tree, combined with high-quality sequences, energy consumption-related features and driving behavior classification factors are selected, mapped to a set of modelable indicators, and a multi-layer feature fusion function is established. , the rule constraints and data channels are comprehensively quantified, and the formula is expressed as follows: in, represents the index of the modelable indicator, Indicates the first The activation strength of the rule, Represents energy consumption related characteristics, represents the driving behavior classification factor, Represents the fusion weight constant; S42: Order A random latent parameter set representing the driving style, based on a multi-layer feature fusion function The observed values construct a probability prior distribution , the prior distribution is constrained in the high-dimensional feature space through multiple integrals, and the formula is expressed as follows: in, represents the random parameter space, represents the total number of fusion indicators, Representation and multi-layer feature fusion function The corresponding prior penalty coefficient; S43: Define a dynamic latent variable set for the braking recovery and motor response delay of the vehicle during operation ,make Represents the energy consumption coupling state and introduces the function The formula to describe the influence of dynamic latent variables on observations is as follows: in, Indicates time The observation vector under represents the basic observation component without energy consumption dynamics, represents the latent variable coupling weight, Used to simulate the time series characteristics of braking recovery rate and delayed power response; S44: Construct a block-by-block hierarchical approximate inference method to accelerate convergence, treat the general variable block and the energy consumption coupling variable block separately, and define the overall posterior distribution The factorization pattern of is: in, Indicates the A parameter set for a common block, Indicates the The hidden variable set of energy consumption coupling blocks, 、 Indicates the number of blocks; Using the variational expectation-maximization method and Continuous iterative updates; S45: Define scene focus functions for extreme road conditions or high power requirements , by increasing the local sampling frequency to refine the estimation accuracy of the corresponding latent variables, the formula is expressed as follows: in, Indicates time In local sampling areas under extreme road conditions, represents a high-weight amplification factor, Reflects the degree of scene triggering, represents the key suppression coefficient, Represents a measure of energy limits or high load areas; S46: Combining the above steps, we can obtain a variational Bayesian analyzable inference model, which is based on a dynamic latent variable set. With random hidden parameter set The combined random variable As the core, through integration Input, using the combined joint posterior distribution To express the final inference result, the formula is as follows: in, represents the likelihood probability distribution of the observed data, Kullback-Leibler divergence, used to measure the overall posterior distribution With prior distribution The difference between.

6. The driving style recognition method based on multi-source sensor data according to claim 5, characterized in that: The step S5 specifically includes: S51: The hierarchical rule tree and the joint posterior distribution Create dynamic mappings for symbol labels With hidden variables Correlation function between Conduct comprehensive assessment to form a multi-layer rule loading function : in, Indicates the index corresponding to the symbol rule in the hierarchical set, Load weights for symbolic rules, is the joint posterior distribution, the correlation function Rules for measuring symbols For latent variables The strength of the constraint; S52: Establishing hard threshold filtering boundaries for vehicle operation behavior , used to determine whether excessive or high-risk behavior violates layered safety requirements. When the value is greater than the given threshold, the rule check is immediately executed. The specific formula is as follows: in, represents a hierarchical semantic index, Indicates vehicle operation instructions. is the operation activation weight, Representation symbolic conditions, For latent variables At the level The compensation factor, Represents the network hidden variable domain that satisfies the current level constraints; S53: By comparing the time and time The posterior distribution of is used to judge the transition confidence and define the distribution difference ,like If the value is greater than the specified threshold, it is determined that the driving style has changed significantly. The specific formula is as follows: in, 、 Respectively indicate time and time The posterior distribution of represents the Kullback-Leibler divergence; S54: Prior symbol distribution The symbol probability obtained by actual inference For comparison: Among them, if If the threshold is exceeded, a collaborative correction is triggered. By analyzing the cause of the difference, the symbol constraints are adjusted or driving risks are prompted, achieving a two-way update and iteration of the complete symbol library and the variational Bayesian parsable inference model. S55: Define multiple loss functions Represents the error metric of the variational Bayesian analytic inference model in different scene dimensions, and the error metric of the variational Bayesian analytic inference model parameters Perform dynamic correction, the formula is as follows: in, and Represent the error component index of energy consumption coupling and driving style respectively, is the learning step coefficient, Express gradient; S56: Define a warning decision function for extreme driving operations that cause the variational Bayesian analytic inference model to fail to converge , if the warning judgment function If the threshold is exceeded, an alert is issued and the final revision results of the variational Bayesian parseable inference model and the complete symbol library are output. The formula is as follows: in, Indicates a high-risk area focused on extreme driving power or speeding situations, represents the detection function corresponding to braking, acceleration, and actual wheel-end power, is the trade-off coefficient.

7. The driving style recognition method based on multi-source sensor data according to claim 6, characterized in that: The step S6 specifically includes: S61: Filter out the vehicle's current main operating characteristics and combine them with the energy consumption safety range to form a basic constraint set and define the initial synthesis state , the formula is as follows: in, Indicates the driving style label score initially identified by the vehicle, Indicates the proportion of remaining energy consumption, and is the adjustable weight coefficient; S62: Real-time sampling of vehicle speed, current output and pedal operation in a short period of time, and based on the joint posterior distribution Determine whether the vehicle has style deviation; if the driving parameters are monitored As the difference between them gradually increases, the style revision process is triggered, and the local posterior state is updated to adapt to the latest observation; S63: When high impact power or emergency braking abnormal operation is detected, a quick matching relationship is established with the complete symbol library, the corresponding symbol entry is found and the relevant cumulative number is recorded; if the abnormal entries accumulate rapidly within a short period of time, it is recorded as a high-risk operation trend and bound to the current operating characteristics of the vehicle; S64: For the high-risk operation trend, if the cumulative number of times exceeds a preset threshold, the situation is marked as a potential safety or energy consumption risk and an internal alarm signal is issued; S65: When external road conditions and driving style continue to change, the variational Bayesian parsable inference model and the complete symbol library are updated online and synchronously; S66: After completing multiple online revisions to the variational Bayesian parsable inference model and the complete symbol library, the vehicle's final energy consumption information and abnormal operation registration are integrated to produce the latest driving style classification and risk assessment results.

8. The driving style recognition method based on multi-source sensor data according to claim 7, characterized in that: The step S66 also includes the final discrimination score A simplified measurement is performed to directly determine the degree of deviation between the vehicle's current state and the target safety boundary. The formula is as follows: in, Indicates the number of abnormal registrations. is the abnormality count penalty coefficient, if If it is lower than the safety threshold, it means that the vehicle behavior has deviated significantly and the intervention strategy needs to be strengthened. Otherwise, the normal driving style label and risk assessment result will be output.

9. A driving style recognition system based on multi-source sensor data, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the driving style recognition method based on multi-source sensor data according to any one of claims 1 to 8.

Images