A typical vehicle operating condition construction method and device and a storage medium

Abstract

This application discloses a method, apparatus, and storage medium for constructing typical vehicle operating conditions, which can construct more representative and adaptable typical vehicle operating conditions. The method for constructing typical vehicle operating conditions includes: generating typical vehicle operating conditions based on a preset double-chain Markov model; wherein the preset double-chain Markov model is constructed through the following steps: classifying collected driving behavior data based on the road operating environment to obtain multi-dimensional driving behavior data; weighting the multi-dimensional driving behavior data based on traffic flow to generate a weighted database; dividing the weighted database into kinematic states to obtain kinematic states, and dividing it into speed ranges to obtain vehicle speed states; and constructing a double-chain Markov model based on the kinematic states and the vehicle speed states.

Description

Technical Field

[0001] This application relates to the field of intelligent transportation technology, specifically to a method, device, and storage medium for constructing typical vehicle operating conditions. Background Technology

[0002] In the fields of automotive engineering and transportation, the accurate construction of vehicle operating conditions is crucial for many aspects, including vehicle performance evaluation, energy consumption analysis, emissions testing, and the development of intelligent transportation systems. Traditional methods for constructing typical operating conditions are primarily based on experimental testing or adherence to relevant regulations. These traditional standard operating conditions played a vital role in specific historical periods and traffic environments; however, with the continuous evolution of the traffic environment, their limitations have become increasingly apparent. Current traffic environments are becoming increasingly complex, and driving modes are showing a trend towards diversification. Traditional standard operating conditions generally suffer from the problem of limited testing scenarios. Especially in the current context of increasingly severe urban traffic congestion and highly dynamic vehicle operation, traditional standard operating conditions can no longer accurately reflect the vehicle's operating state under actual traffic scenarios and cannot meet the needs of modern transportation research and applications. In recent years, the widespread adoption of in-vehicle intelligent terminals and the maturity of big data collection technologies have provided new ideas and methods for operational condition modeling. However, most existing data-driven operational condition modeling methods focus on cluster analysis, typical sequence extraction, and speed distribution reconstruction. These methods typically assume that all types of data samples have equal statistical contributions, neglecting the spatial heterogeneity of traffic operations and the complexity of road structures. This results in operational condition models that often suffer from insufficient representativeness and poor transferability in practical applications. Therefore, a new method for constructing typical vehicle operating conditions is urgently needed to overcome the shortcomings of traditional methods and existing data-driven approaches. Summary of the Invention

[0003] To address the aforementioned technical problems, this application is proposed. Embodiments of this application provide a method, apparatus, and storage medium for constructing typical vehicle operating conditions, which can construct more representative and adaptable typical vehicle operating conditions.

[0004] According to a first aspect of this application, a method for constructing typical vehicle operating conditions is provided, comprising: generating typical vehicle operating conditions based on a preset double-chain Markov model; wherein the preset double-chain Markov model is constructed through the following steps: classifying collected driving behavior data based on the road operating environment to obtain multi-dimensional driving behavior data; weighting the multi-dimensional driving behavior data based on traffic flow to generate a weighted database; dividing the weighted database into kinematic states to obtain kinematic states, and dividing it into speed ranges to obtain vehicle speed states; and constructing a double-chain Markov model based on the kinematic states and the vehicle speed states.

[0005] As one possible implementation, multiple driving segments are extracted from the weighted database, and sub-segments of a preset duration are extracted from the driving segments to form an input dataset; wherein, the sub-segments include vehicle speed data and kinematic features; the input dataset is input into a preset self-organizing map neural network, and unsupervised clustering is performed on the kinematic features of the sub-segments to divide driving behavior into multiple kinematic states; the vehicle speed data of the sub-segments is divided according to a preset speed range to obtain multiple vehicle speed states; wherein, based on the kinematic states and the vehicle speed states, a double-chain Markov model is constructed, including: associating and pairing the kinematic states and vehicle speed states at the same time stamp to form joint state pairs; based on the joint state pairs in the time series, statistically analyzing the state transition patterns, and generating a joint transition probability matrix.

[0006] As one possible implementation, based on joint state pairs in a time series, statistical state transition patterns are analyzed to generate a joint transition probability matrix, including: counting the transitions of all adjacent joint state pairs in the historical sequence; and estimating the joint probability of transitioning from the current joint state to the next joint state based on the transition counts; wherein the joint probability is mathematically represented as the product of a first probability component and a second probability component; the first probability component represents the transition probability between kinematic states, and the second probability component represents the transition probability between vehicle speed states given the next kinematic state.

[0007] As one possible implementation, a typical vehicle operating condition is generated based on a preset double-chain Markov model, including: based on the preset double-chain Markov model, starting from the initial kinematic state and vehicle speed state, and based on the joint transition probability matrix, simultaneously generating a kinematic state sequence and a vehicle speed state sequence; and generating a typical vehicle operating condition based on the kinematic state sequence and the vehicle speed state sequence.

[0008] As one possible implementation, the method for constructing typical vehicle operating conditions further includes: determining a target driving distance based on preset road labels as a termination condition for generating typical vehicle operating conditions; setting the initial kinematic state to idle and the initial vehicle speed state to 0 as the initial operating condition state; wherein, based on the preset double-chain Markov model, starting from the initial kinematic state and vehicle speed state, and based on the joint transition probability matrix, synchronously generating a kinematic state sequence and a vehicle speed state sequence includes: starting from the initial operating condition state, extracting random numbers from a uniformly distributed preset dataset; searching for an index position that satisfies a preset condition in the joint transition probability matrix based on the random numbers; determining the next operating condition state based on the index position; and adding the next operating condition state to the kinematic state sequence and the vehicle speed state sequence.

[0009] As one possible implementation, the classification of collected driving behavior data based on the road operating environment to obtain multi-dimensional driving behavior data includes: classifying the collected driving behavior data into road type dimension, time period dimension, and region dimension based on the road operating environment; and forming sample space combination units based on the combination of road type dimension, time period dimension, and region dimension.

[0010] As one possible implementation, multi-dimensional driving behavior data is weighted based on traffic flow to generate a weighted database. This includes: weighting sample space combination units that meet preset requirements based on traffic flow to generate the weighted database; wherein the sample space combination units that meet the preset requirements include: the actual collection volume of the sample space combination unit is greater than or equal to a preset target collection volume; or the traffic flow deviation rate of the sample space combination unit is less than a preset deviation rate; wherein the traffic flow deviation rate refers to the deviation between the actual sample proportion of the sample space combination unit and the corresponding target traffic flow proportion.

[0011] As one possible implementation, the method for constructing typical vehicle operating conditions includes: statistically analyzing traffic flow data for each sample space combination unit; wherein the traffic flow data represents the total number of vehicles passing through the sample space combination unit per unit time; wherein weighting multi-dimensional driving behavior data based on traffic flow to generate a weighted database includes: assigning a weight value to the corresponding sample space combination unit according to the traffic flow data of each sample space combination unit; wherein the weight value is proportional to the traffic flow data.

[0012] According to a second aspect of this application, a typical vehicle operating condition construction device is provided, comprising: a generation module for generating typical vehicle operating conditions based on a preset double-chain Markov model; a classification module for classifying collected driving behavior data based on the road operating environment to obtain multi-dimensional driving behavior data; a weighting module for weighting the multi-dimensional driving behavior data based on traffic flow to generate a weighted database; a partitioning module for partitioning the weighted database into kinematic states to obtain kinematic states, and partitioning speed ranges to obtain vehicle speed states; and a construction module for constructing a double-chain Markov model based on the kinematic states and the vehicle speed states.

[0013] According to a third aspect of this application, a computer-readable storage medium is provided, the storage medium storing a computer program for performing the method as described in the first aspect or any implementation thereof.

[0014] The method, apparatus, and storage medium for constructing typical vehicle operating conditions provided in this application classify driving behavior data through the road operating environment. By fully considering multi-dimensional features during the sampling phase, more targeted typical operating conditions can be constructed. A weighting mechanism reflecting actual traffic flow is introduced to improve the environmental adaptability and engineering versatility of subsequent models. The generated weighted database provides more accurate data for the subsequent construction of a double-chain Markov model. Data processing using the weighted database generates kinematic and vehicle speed states that reflect the operating conditions. The road operating environment reflects macroscopic clustering states, while the kinematic and vehicle speed states reflect microscopic continuous states, thus overcoming the limitations of single-chain Markov models. A double-chain Markov model is constructed based on the joint state of kinematic and vehicle speed states. This double-chain Markov model can then construct more representative, adaptable, and scalable typical vehicle operating conditions, providing more valuable input for related engineering research and product development. Attached Figure Description

[0015] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0016] Figure 1 This is a flowchart illustrating a typical vehicle operating condition construction method provided in an exemplary embodiment of this application.

[0017] Figure 2 This is a schematic diagram of the data preprocessing flow provided in an exemplary embodiment of this application.

[0018] Figure 3 This is a schematic diagram of a typical vehicle operating condition construction device provided in an exemplary embodiment of this application.

[0019] Figure 4 This is a structural diagram of an electronic device provided in an exemplary embodiment of this application. Detailed Implementation

[0020] Hereinafter, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0021] Currently, the methods for constructing vehicle driving conditions can be mainly divided into two categories: experimental driving conditions based on standards and specifications, and data-driven driving conditions based on actual driving data. In the standard and specification approach, standard test cycles uniformly established by international or national organizations are typically used, such as the New European Driving Cycle (NEDC) and the Worldwide Harmonized Light Vehicles Test Cycle (WLTC). These standard driving conditions have the advantages of uniformity and strong comparability of results, and are widely used in areas such as vehicle energy consumption testing and emissions certification. However, their test scenarios are mostly based on specific road types and driving behaviors, making it difficult to fully cover the diversity of road structures, the dynamic complexity of traffic conditions, and the differences in driver behavior in the current traffic environment. With the increasing severity of urban traffic congestion and the growing diversification of driving modes, the representativeness of these standard driving conditions in real-world traffic scenarios is gradually becoming limited.

[0022] In recent years, with the development of vehicle-mounted terminals and data acquisition technologies, methods for constructing driving conditions based on real vehicle driving data have gradually become an important research direction. This method processes and analyzes massive amounts of trajectory data to extract driving segments that reflect actual driving characteristics, thereby constructing representative driving conditions. Common techniques include cluster analysis of velocity sequences, extraction of typical kinematic segments, and synthetic reconstruction based on velocity-acceleration distributions. However, existing methods, in the process of data processing and modeling, typically treat sample data from different sources or regions as statistically uniform sets, failing to fully consider the spatial heterogeneity in traffic operations, such as differences in road network structure, traffic function, and flow characteristics across different regions. Due to the lack of effective modeling of the complexity of road structure and traffic environment, the constructed driving condition models may suffer from insufficient representativeness and weak adaptability when applied across regions and scenarios.

[0023] Therefore, to address the aforementioned issues and effectively integrate road structure and spatial feature information when modeling real driving data, this application proposes a method for constructing typical vehicle operating conditions that is more representative and has better transferability. Figure 1 For example, Figure 1 This is a flowchart illustrating a typical vehicle operating condition construction method provided in an exemplary embodiment of this application. First, a pre-defined double-chain Markov model is constructed through the following steps: The collected driving behavior data is classified based on the road operating environment to obtain multi-dimensional driving behavior data (see...). Figure 1 (S110), and secondly, based on traffic flow, multi-dimensional driving behavior data is weighted to generate a weighted database (see S110). Figure 1 (S120). Then, the weighted database is divided into kinematic states to obtain the kinematic states, and into speed ranges to obtain the vehicle speed states (see S120). Figure 1(S130), then, based on the kinematic state and vehicle speed state, a double-chain Markov model is constructed (see S130). Figure 1 (S140). Finally, based on the preset double-chain Markov model, typical vehicle operating conditions are generated (see S140). Figure 1 (S150). In massive data, the spatial heterogeneity of traffic operation is fully considered for classification and weighting, effectively integrating road structure and spatial feature information, and combining Markov chain probabilistic modeling to overcome the shortcomings of traditional methods and existing data-driven methods, and construct a more accurate reflection of typical vehicle operating conditions in complex traffic environments.

[0024] The following text combines Figure 1 This application provides a more detailed description of the typical vehicle operating condition construction method provided in the embodiments.

[0025] In S110, the collected driving behavior data is classified based on the road operating environment to obtain multi-dimensional driving behavior data.

[0026] In some embodiments, firstly, to ensure the scientific validity and representativeness of the typical vehicle operating conditions, a multi-dimensional hierarchical data collection and processing model based on traffic flow ratios, which balances coverage and balance, can be adopted in the data collection stage. This model fully considers multi-dimensional characteristics such as road type, traffic time period, and spatial area during the sampling phase, and introduces a flow weighting mechanism to enable the sampling distribution to more realistically reflect the actual traffic operation. For example, the target city's road network can be hierarchically divided, hierarchical collection requirements can be clarified, data representativeness judgment indicators can be defined, and sample distribution bias can be corrected through sample weighting or supplementary collection when necessary.

[0027] Current research generally fails to adequately consider the variability of road operating conditions during work condition modeling. For example, different road types (e.g., highways, expressways, arterial roads, secondary arterial roads, and local roads) exhibit significantly different driving behavior patterns, and traffic flow distribution in urban road networks is significantly non-uniform. Some road segments may appear infrequently in the data but have high actual traffic frequency; conversely, some frequently occurring trajectory segments may have a limited contribution to overall traffic, causing unweighted models to structurally deviate from the characteristics of mainstream work conditions. From an engineering perspective, typical work conditions not only need to be mathematically representative but also ensure coverage of typical operating environments from a traffic engineering perspective. Therefore, incorporating macro-statistical information such as traffic flow into work condition models to reasonably reflect the importance of roads in actual traffic operation will help enhance the model's practical guiding value. Furthermore, road type, as an important dimension in traffic planning and vehicle control strategy design, significantly influences speed change trends, acceleration / deceleration behavior, and start-stop states, and should not be ignored.

[0028] Therefore, to address the issue of data usage not taking into account the differences between various dimensions, one possible approach is to divide the road operating environment into three dimensions: road type, time period, and region. Based on the road operating environment, the collected driving behavior data can be categorized into road type, time period, and region dimensions. The combination of these dimensions forms a sample space combination unit.

[0029] For the road type dimension, the data collection can cover different types of roads, including expressways (H), rapid transit (E), arterial roads (M), secondary arterial roads (S), and local roads (L), to reflect the differences in road levels.

[0030] In terms of time period, the data collection can distinguish typical traffic operation periods, including peak period (P), off-peak period (O), and low-peak period (Q), to reflect the time-varying characteristics of traffic flow.

[0031] For the regional dimension, the data collection scope can cover different areas such as the central urban area (C), the near suburbs (N), and the suburbs (R) to reflect the spatial differences in commuting behavior and traffic organization.

[0032] By combining the above three dimensions, 5×3×3=45 sample space combination units can be formed, denoted as: In some embodiments, to ensure the completeness and balance of sampling, a minimum sampling quantity constraint can be explicitly specified for each sample space combination. In other words, the number of effective trajectory samples in each combination unit should not be less than a preset lower limit, thereby avoiding model distortion caused by the scarcity of samples of a certain category. Therefore, after sampling is completed, the sample coverage can be statistically verified to ensure the number of effective trajectory samples.

[0033] For example, the actual sample volume of a sample space combination unit is greater than or equal to the preset target sample volume; or the traffic flow deviation rate of the sample space combination unit is less than the preset deviation rate. Here, the traffic flow deviation rate refers to the deviation between the actual sample proportion of a sample space combination unit and the corresponding target traffic flow proportion. In other words, two types of coverage judgment indicators are introduced: sampling coverage rate and traffic flow deviation rate. For example, the sampling coverage rate refers to the ratio of the actual sample volume to the target sample volume for each sample space, and the coverage rate must not be less than 80%, meaning that each combination must complete at least 80% of the planned sampling volume. The traffic flow deviation rate refers to the deviation between the proportion of each category of samples in the total sample size and the proportion of that category in the target traffic flow. The traffic flow deviation rate can be calculated using Formula 1: Formula 1; In Formula 1, sample proportion refers to the proportion of each category of samples in the total sample size, and traffic flow proportion refers to the proportion of each category of samples in the target traffic flow. ijk This refers to the flow deviation rate. To ensure that the sampling distribution is as close as possible to the actual traffic conditions, the flow deviation rate should ideally not exceed 15%. If the flow deviation rate of any sample spatial combination unit exceeds 15%, it is considered insufficient sampling, and supplementary sampling is required.

[0034] In some embodiments, to ensure the accuracy and rationality of the collected data, a systematic preprocessing method can be implemented during the data processing stage to preprocess the collected data, thereby ensuring that the data subsequently used in the model is of high quality. Preprocessing methods may include outlier identification, redundant data cleaning, and kinematic segment rationality determination.

[0035] Figure 2 This is a schematic diagram of the data preprocessing flow provided in an exemplary embodiment of this application, as shown below. Figure 2 As shown, the first step is to perform anomaly detection on the raw data to ensure data health. Anomalies include: inconsistent states, data redundancy at the same time step, speed anomalies, and acceleration anomalies. For handling inconsistent states, if logical contradictions exist between parameters that prevent consistency, the relevant data will be considered invalid and deleted. For example, if a pure electric vehicle has zero current but non-zero acceleration, this logical contradiction should be handled as described above. For data redundancy elimination, if multiple records exist at the same timestamp, only one representative record is retained, and the remaining redundant data is deleted to ensure the uniqueness of the time series. For handling acceleration anomalies, if the acceleration value exceeds the positive threshold of 4 m / s², or the deceleration value exceeds the negative threshold of 8 m / s², the relevant data will be identified as anomalies and removed. For handling speed anomalies, if the vehicle speed exceeds 130 km / h (the theoretical upper limit), it is considered out-of-limit data and will also be deleted.

[0036] See also Figure 2Next, the data after anomaly detection is segmented, for example, by duration. Anomaly segments are then removed based on motion segments to ensure data rationality and effectiveness. Anomaly segments can include long periods of parking, intermittent low-speed driving, and long idling times; these segments do not provide effective assistance to model recognition and can be directly deleted. For example, a comprehensive judgment is made on the collected kinematic segments: segments with a maximum speed not exceeding 10 km / h and a duration of less than 50 seconds are considered low-speed, short-duration segments and will be directly deleted. Segments with an idling time exceeding 180 seconds are also considered anomalies. For cases with excessively long idling times, the idling time of the segment is truncated to 180 seconds, and the remaining portion is retained. After the anomaly detection and anomaly segment removal preprocessing, the data can be output for subsequent data classification. The preprocessing effectively removes invalid or unreasonable data, improving the accuracy and representativeness of subsequent modeling and analysis.

[0037] See also Figure 1 In S120, multi-dimensional driving behavior data is weighted based on traffic flow to generate a weighted database.

[0038] To further enhance the representativeness and engineering applicability of the constructed typical vehicle operating conditions, after data preprocessing and classification, a traffic flow-based weighting mechanism is introduced to quantitatively assign weights to each layer of sample space. That is, based on traffic flow, the sample space combination units that meet the preset requirements are weighted to generate a weighted database.

[0039] One possible implementation involves statistically analyzing traffic flow data for each sample space combination unit. This traffic flow data represents the total number of vehicles passing through the sample space combination unit per unit time. Based on the traffic flow data for each sample space combination unit, a weight value is assigned to the corresponding sample space combination unit; the weight value is proportional to the traffic flow data. In other words, within a defined three-dimensional hierarchical space S, traffic flow data is statistically analyzed for each road type, time period, and area combination unit. Traffic flow is measured based on the total number of vehicles passing through the sample space per unit time.

[0040] For example, to reflect the importance of each sample space in the overall traffic situation, weights are determined based on the number of vehicles passing through. The weights are defined as follows: Formula 2; In Formula 2, Represents the sample space combination unit The number of vehicles passing through the internal statistics The weights represent the relative importance of each sample space in the overall traffic distribution.

[0041] After the weights are calculated, based on The original samples in the database are weighted. Each sample spatial combination unit is assigned a corresponding weight to ensure that the contribution of that unit is consistent with its actual traffic flow proportion in subsequent work condition construction. When some units have insufficient sample numbers, corrections can be made through sample duplication, supplementary collection, or interpolation, while maintaining consistency with the target weight ratio. Through this weighting mechanism, the final weighted database can simultaneously reflect traffic flow distribution characteristics and sampling coverage, effectively avoiding distortion problems caused by sample imbalance, and providing high-fidelity input data for subsequent work condition synthesis based on Markov processes.

[0042] In S130, the weighted database is divided into kinematic states to obtain the kinematic states, and the speed range is divided to obtain the vehicle speed states.

[0043] When dividing the weighted database into kinematic states and velocity ranges, the Self-Organizing Map (SOM) algorithm can be used to perform unsupervised clustering of the fragment data, identifying representative micro-travels as kinematic states (main states). Then, discrete data points are used as sub-states, which, together with the main states, generate a joint transition probability matrix. Guided by the joint TPM (Transition Probability Matrix), Markov Chain Monte Carlo (MCMC) is used to generate multiple sets of candidate driving condition sequences.

[0044] As one possible implementation, the efficient unsupervised clustering method—the SOM algorithm—is first used to classify and reduce the kinematic states of large-scale driving data. Multiple driving segments are extracted from a weighted database, and sub-segments of preset durations are extracted from these segments to form the input dataset. These sub-segments include vehicle speed data and kinematic features. The input dataset is then fed into a preset self-organizing map neural network to perform unsupervised clustering of the kinematic features of the sub-segments, thereby dividing driving behavior into multiple kinematic states. The vehicle speed data of the sub-segments is further divided according to preset speed ranges to obtain multiple vehicle speed states. When constructing the double-chain Markov model, kinematic states and vehicle speed states at the same time stamp are associated and paired to form joint state pairs. Based on the joint state pairs in the time series, the state transition patterns are statistically analyzed to generate a joint transition probability matrix.

[0045] For example, a 4-second segment is extracted from each driving segment in the weighted database and used as the input set. Subsequently, during SOM training, the neuron weights in the competitive layer are iteratively updated according to the following formula: Formula 3; In Formula 3, Indicates the update rate; This represents the best-matching neuron that wins in the current iteration; It is the neighboring neurons, and the influence of the neighboring neurons depends on their distance from the winning neuron; For neighborhood functions, This represents the input set (a 4-second segment). represents the neuron weight, and t represents the number of iterations.

[0046] in, It can be calculated using Formula 4: Formula 4; In Formula 4, The Manhattan distance between neurons is represented by t, which represents the number of iterations; the effective width of the neighborhood is t. This can be set using Formula 5: Formula 5; In Formula 5, and It is a constant variable, and t represents the number of iterations.

[0047] To control the update rate, It can be defined using Formula Six: Formula Six; In Formula Six, It is a constant variable, and t represents the number of iterations.

[0048] As learning progresses, the update rate gradually decreases. This gradual decay mechanism ensures that the neuron weights can be coarsely adjusted to finely tuned, eventually converging stably to different cluster centers. Through this process, driving segments can be clustered into six typical kinematic states based on speed and acceleration values ​​using a SOM neural network: idling, rapid acceleration, gradual acceleration, cruising, gradual deceleration, and rapid deceleration.

[0049] To more accurately reflect real-world dynamic driving behavior, please continue reading. Figure 1 In S140, a double-chain Markov model is constructed based on the kinematic state and vehicle speed state. That is, the two variables of kinematic state and vehicle speed state are combined to construct a double-chain Markov model, and the double-chain Markov model is used to realize the construction process of driving conditions.

[0050] In some embodiments, the evolution of the vehicle's operating state is modeled as a discrete-time, finite-state-space Markov stochastic process. A Markov process is a statistically significant sequence of stochastic states that satisfies the property of no aftereffect; that is, the state transitions of the system at any given time depend only on the current state and are independent of historical states. Mathematically, its core definition can be expressed as: Formula 7; In Formula 7, P(... | ...) represents conditional probability. The left side of the vertical bar | represents the event to be determined, and the right side represents the known conditions. X t and X t+1 There are two random variables, X t X represents the system state at time step t. t+1 This represents the system state at the next time step t+1. In the discrete-time case, this stochastic process is usually called a Markov chain, and it is characterized by a transition probability matrix (TPM). The ordinary state probability transition matrix is ​​denoted by Equation 8: Formula 8; P nk It is a probability value. The indices n and k correspond to the state x. i and x j , where P nk This indicates that the system is currently in state S. n Transition to state S k The probability of the matrix. This matrix has the following basic properties: non-negativity, each element is non-negative and less than 1, that is: The probability summation property states that for any current state i, the sum of the probabilities of transitioning to all possible subsequent states equals 1, ensuring the completeness and consistency of state evolution. In vehicle operating condition modeling applications, the vehicle speed-acceleration combined state can be viewed as a finite set of system states. During vehicle operation on the road, its state evolution follows an approximate Markov property: when the time granularity is sufficiently small, the vehicle's state at the next moment is mainly determined by the current moment, with minimal influence from earlier state histories.

[0051] Based on the Markov model, in some embodiments, the joint state can be defined by Equation 9: Formula Nine; In Formula Nine, Represents the set of six micro-travel states This refers to six micro-travel states: {idle, rapid acceleration, gradual acceleration, cruise, gradual deceleration, rapid deceleration}. Represents the set of vehicle speed sub-states Each element Indicates the time interval The average speed within the area.

[0052] In some embodiments, transition counts can be performed on all adjacent joint state pairs in the historical sequence; based on the transition counts, the joint probability of transitioning from the current joint state to the next joint state is estimated; wherein the joint probability is mathematically represented as the product of a first probability component and a second probability component; the first probability component represents the transition probability between kinematic states, and the second probability component represents the transition probability between vehicle speed states given the next kinematic state.

[0053] As one possible implementation, in a doubly-linked Markov model, starting from the current state... Transition to the next state The joint probability can be expressed as Equation 10: Formula 10; In Formula 10, This represents the transition probability between kinematic states; This represents the transition probability between vehicle speed states given the next kinematic state. This probability expresses the state at which the vehicle speed transitions from the current kinematic state to the next kinematic state. and current vehicle speed status Based on this, it simultaneously transitions to a new kinematic state. and new vehicle speed status The possibilities are used to form a comprehensive joint transition probability matrix (TPM).

[0054] The combined TPM can be represented as shown in Formula 11: Formula 11; In Formula 11, T represents ij The cumulative sum of each element in T ij Let p be a row vector representing the probability distribution of transitions from state (i, j) (e.g., a joint state at times t and t+1, or a pair of states of the system at a certain moment) to all possible next joint states. p represents the transition probability, and n and m represent the number of states in two different dimensions of the system, such as the number of kinematic states and the number of vehicle speed states. To apply the MCMC method, we first determine... The cumulative sum of each element in the formula is shown in Formula Twelve: Formula 12; In Formula Twelve, T ij This represents the probability distribution of transitions from state (i, j) to all possible next joint states. T represents ij The cumulative sum of each element in the system, where p represents the transition probability, and n and m represent the number of states in two different dimensions of the system, respectively.

[0055] Each element in a combined TPM can be accessed through For example, formula thirteen: Formula Thirteen; In Formula Thirteen, T ij Let p represent the probability distribution of transitioning from state (i, j) to all possible next joint states, where p represents the transition probability, and n and m represent the number of states in two different dimensions of the system, respectively.

[0056] In Monte Carlo sampling, a random value is selected. When the conditions are met At that time, set and The next state is The operators " / " and "%" represent the rounding down and rounding up functions, respectively.

[0057] After the construction of the double-chain Markov model is completed, it is defined as the preset double-chain Markov model. In S150, typical vehicle operating conditions are generated based on the preset double-chain Markov model.

[0058] In some embodiments, based on a preset double-chain Markov model, starting from the initial kinematic state and vehicle speed state, a kinematic state sequence and a vehicle speed state sequence are generated synchronously based on a joint transition probability matrix; based on the kinematic state sequence and the vehicle speed state sequence, typical vehicle operating conditions are generated.

[0059] As one possible implementation, the target driving distance is determined based on preset road labels, serving as the termination condition for generating typical vehicle operating conditions. The initial kinematic state is set to idle speed, and the initial vehicle speed state is set to 0, serving as the initial operating condition state. Based on a preset double-chain Markov model, starting from the initial kinematic and vehicle speed states, and based on a joint transition probability matrix, a kinematic state sequence and a vehicle speed state sequence are generated simultaneously. This includes: starting from the initial operating condition state, randomly selecting numbers from a uniformly distributed preset dataset; searching for an index position that satisfies preset conditions in the joint transition probability matrix based on the random numbers; determining the next operating condition state based on the index position; and adding the next operating condition state to the kinematic state sequence and the vehicle speed state sequence.

[0060] In other words, we can first determine the total distance required to reach the target vehicle speed profile based on preset road labels, using this as the termination condition for state iteration. Secondly, we perform initialization, setting the initial state of the vehicle speed profile to idle and a vehicle speed of 0. Then, from the current state Starting with a pre-defined dataset that is evenly distributed Random numbers are randomly drawn from the middle. Next, find the matching criteria in the joint TPM. index position In this way, the next state can be effectively sampled and determined. In other words, the cumulative sum vector transforms the discrete probability distribution into a partition on the interval [0, 1]. The probability that a random number u falls into a particular interval ([0, P1], (P1, P1+P2], (P1+P2, 1]) is exactly equal to the original probability (P1, P2, P3) of that interval. This ensures that the sampling process strictly follows the transition probability T. ij The defined distribution.

[0061] Next, the newly generated state is appended to the sequence, and the travel distance is calculated based on the current speed and time step. If the cumulative distance has not yet reached the target, the next state transition continues. Finally, when the cumulative distance meets the termination condition, the complete vehicle speed sequence is extracted. This is used as a valid driving condition sample generated in one step. Through this complete step, highly realistic driving conditions, i.e., speed-time curves, can be generated systematically and efficiently as typical vehicle operating conditions. Typical vehicle operating conditions depend on the original dataset; driving conditions are essentially simplified microcosms of the dataset and are sufficiently representative. Typical driving conditions for each dataset can be formed by classifying the dataset, such as urban, suburban, and highway driving, which facilitates subsequent simulation training.

[0062] Figure 3 This is a schematic diagram of the structure of a typical vehicle operating condition construction device provided in an exemplary embodiment of this application, as shown below. Figure 3 As shown, the typical vehicle operating condition construction device 3 includes: a generation module 31, used to generate typical vehicle operating conditions based on a preset double-chain Markov model; a classification module 32, used to classify the collected driving behavior data based on the road operating environment to obtain multi-dimensional driving behavior data; a weighting module 33, used to weight the multi-dimensional driving behavior data based on traffic flow to generate a weighted database; a partitioning module 34, used to partition the weighted database into kinematic states to obtain kinematic states, and to partition the speed ranges to obtain vehicle speed states; and a construction module 35, used to construct a double-chain Markov model based on the kinematic states and vehicle speed states.

[0063] As one possible implementation, the segmentation module 34 can be configured to: extract multiple driving segments from a weighted database, and extract sub-segments of a preset duration from the driving segments to form an input dataset; wherein, the sub-segments include vehicle speed data and kinematic features; input the input dataset into a preset self-organizing map neural network, and perform unsupervised clustering on the kinematic features of the sub-segments, thereby dividing the driving behavior into multiple kinematic states; divide the vehicle speed data of the sub-segments according to a preset speed range to obtain multiple vehicle speed states; wherein, the construction module 35 can be configured to: associate and pair the kinematic states and vehicle speed states at the same timestamp to form joint state pairs; based on the joint state pairs in the time series, statistically analyze the state transition patterns and generate a joint transition probability matrix.

[0064] As one possible implementation, the building module 35 can be configured to: count the transitions of all adjacent joint state pairs in the historical sequence; and estimate the joint probability of transitioning from the current joint state to the next joint state based on the transition counts; wherein the joint probability is mathematically represented as the product of a first probability component and a second probability component; the first probability component represents the transition probability between kinematic states, and the second probability component represents the transition probability between vehicle speed states given the next kinematic state.

[0065] As one possible implementation, the generation module 31 can be configured to: based on a preset double-chain Markov model, starting from the initial kinematic state and vehicle speed state, and based on the joint transition probability matrix, simultaneously generate a kinematic state sequence and a vehicle speed state sequence; and based on the kinematic state sequence and the vehicle speed state sequence, generate typical vehicle operating conditions.

[0066] As one possible implementation, the typical vehicle operating condition construction device 3 may include: determining the target driving distance based on preset road labels as the termination condition for generating typical vehicle operating conditions; setting the initial kinematic state to idle speed and the initial vehicle speed state to 0 as the initial operating condition state; wherein, the generation module 31 may be configured to: starting from the initial operating condition state, extracting random numbers from a uniformly distributed preset dataset; searching for the index position that satisfies the preset condition from the joint transition probability matrix based on the random numbers; determining the next operating condition state based on the index position; and adding the next operating condition state to the kinematic state sequence and the vehicle speed state sequence.

[0067] As one possible implementation, the classification module 32 can be configured to: divide the collected driving behavior data into road type dimension, time period dimension and region dimension based on the road operating environment; and form a sample space combination unit based on the combination of road type dimension, time period dimension and region dimension.

[0068] As one possible implementation, the weighting module 33 can be configured to: perform weighting processing on sample space combination units that meet preset requirements based on traffic flow, and generate a weighted database; wherein, the sample space combination units that meet the preset requirements include: the actual collection volume of the sample space combination unit is greater than or equal to the preset target collection volume; or the traffic flow deviation rate of the sample space combination unit is less than the preset deviation rate; wherein, the traffic flow deviation rate refers to the deviation between the actual sample ratio of the sample space combination unit and the corresponding target traffic flow ratio.

[0069] As one possible implementation, the typical vehicle operating condition construction device 3 can be configured to: statistically analyze the traffic flow data of each sample space combination unit; wherein the traffic flow data represents the total number of vehicles passing through the sample space combination unit per unit time; wherein, the multi-dimensional driving behavior data is weighted based on the traffic flow to generate a weighted database, including: assigning a weight value to the corresponding sample space combination unit according to the traffic flow data of each sample space combination unit; wherein the weight value is proportional to the traffic flow data.

[0070] An electronic device includes: a processor; a memory for storing processor-executable instructions; and a processor for executing a typical vehicle operating condition construction method according to embodiments provided in this application.

[0071] Below, for reference Figure 4 This application describes an electronic device according to embodiments thereof. The electronic device may be either or both of a first device and a second device, or a standalone device independent of them, which may communicate with the first device and the second device to receive acquired input signals from them.

[0072] Figure 4 A block diagram of an electronic device according to an embodiment of this application is illustrated.

[0073] like Figure 4 As shown, the electronic device 40 includes one or more processors 41 and a memory 42.

[0074] The processor 41 may be a central processing unit (CPU) or other form of processing unit with data processing and / or instruction execution capabilities, and may control other components in the electronic device 40 to perform desired functions.

[0075] The memory 42 may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and / or cache memory. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 41 may execute the program instructions to implement the typical vehicle operating condition construction method of the various embodiments of this application described above, and / or other desired functions. Various contents such as input signals, signal components, and noise components may also be stored in the computer-readable storage medium.

[0076] In one example, the electronic device 40 may also include an input device 43 and an output device 44, which are interconnected via a bus system and / or other forms of connection mechanism (not shown).

[0077] When the electronic device is a standalone device, the input device 43 can be a communication network connector for receiving the collected input signals from the first device and the second device.

[0078] In addition, the input device 43 may also include, for example, a keyboard, a mouse, etc.

[0079] The output device 44 can output various information to the outside, including determined distance information, direction information, etc. The output device 44 may include, for example, a display, a speaker, a printer, and a communication network and its connected remote output devices, etc.

[0080] Of course, for the sake of simplicity, Figure 4 Only some of the components of the electronic device 40 relevant to this application are shown in this illustration; components such as buses, input / output interfaces, etc., are omitted. In addition, the electronic device 40 may include any other suitable components depending on the specific application.

[0081] Computer program products can be written in any combination of one or more programming languages ​​to perform the operations of the embodiments of this application. The programming languages ​​include object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computing device, partially on the user's computing device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0082] A computer-readable storage medium stores a computer program for executing a method for constructing typical vehicle operating conditions according to embodiments provided in this application.

[0083] Computer-readable storage media may take the form of any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may, for example, include, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: electrical connections having one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0084] The above description has been given for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of this application to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations thereof.

Claims

1. A method for constructing typical vehicle operating conditions, characterized in that, include: Based on a pre-defined double-chain Markov model, typical vehicle operating conditions are generated. The preset double-chain Markov model is constructed through the following steps: Based on the road operating environment, the collected driving behavior data is classified to obtain multi-dimensional driving behavior data; Based on traffic flow, multi-dimensional driving behavior data is weighted to generate a weighted database; The weighted database is divided into kinematic states to obtain kinematic states, and into speed ranges to obtain vehicle speed states. Based on the kinematic state and the vehicle speed state, a double-chain Markov model is constructed.

2. The method for constructing typical vehicle operating conditions according to claim 1, characterized in that, The weighted database is divided into kinematic states to obtain kinematic states, and into velocity ranges to obtain vehicle speed states, including: Multiple driving segments are extracted from the weighted database, and sub-segments of a preset duration are extracted from the driving segments to form an input dataset; wherein, the sub-segments include vehicle speed data and kinematic features; The input dataset is fed into a preset self-organizing map neural network, and the kinematic features of the sub-segments are clustered in an unsupervised manner to divide driving behavior into multiple kinematic states. The vehicle speed data of the sub-segment is divided according to a preset speed range to obtain multiple vehicle speed states; Based on the kinematic state and the vehicle speed state, a double-chain Markov model is constructed, including: The kinematic states and vehicle speed states at the same timestamp are associated and paired to form joint state pairs; Based on the joint state pairs in the time series, statistical state transition patterns are analyzed to generate a joint transition probability matrix.

3. The method for constructing typical vehicle operating conditions according to claim 2, characterized in that, Based on joint state pairs in a time series, statistical state transition patterns are analyzed to generate a joint transition probability matrix, including: Count the transitions for all adjacent joint state pairs in the historical sequence; Based on the transition count, the joint probability of transitioning from the current joint state to the next joint state is estimated; The joint probability is mathematically represented as the product of a first probability component and a second probability component; the first probability component represents the transition probability between kinematic states, and the second probability component represents the transition probability between vehicle speed states given the next kinematic state.

4. The method for constructing typical vehicle operating conditions according to claim 3, characterized in that, Based on a pre-defined double-chain Markov model, typical vehicle operating conditions are generated, including: Based on the preset double-chain Markov model, starting from the initial kinematic state and vehicle speed state, and based on the joint transition probability matrix, the kinematic state sequence and vehicle speed state sequence are generated simultaneously. Based on the kinematic state sequence and the vehicle speed state sequence, typical vehicle operating conditions are generated.

5. The method for constructing typical vehicle operating conditions according to claim 4, characterized in that, The method for constructing typical vehicle operating conditions also includes: Based on the preset road labels, the target driving distance is determined as the termination condition for generating typical vehicle operating conditions; The initial kinematic state is set to idle speed, and the initial vehicle speed state is set to 0, which serves as the initial operating condition. Specifically, based on the preset double-chain Markov model, starting from the initial kinematic state and vehicle speed state, and based on the joint transition probability matrix, the kinematic state sequence and vehicle speed state sequence are generated simultaneously, including: Starting from the initial working condition, random numbers are drawn from a uniformly distributed preset dataset; Based on the random number, find the index position that satisfies the preset condition from the joint transition probability matrix; Based on the index position, determine the next operating condition. The next operating condition state is added to the kinematic state sequence and the vehicle speed state sequence.

6. The method for constructing typical vehicle operating conditions according to claim 1, characterized in that, The classification of collected driving behavior data based on the road operating environment yields multi-dimensional driving behavior data, including: Based on the road operating environment, the collected driving behavior data is divided into road type dimension, time period dimension, and region dimension; Based on the combination of road type dimension, time period dimension and region dimension, a sample space combination unit is formed.

7. The method for constructing typical vehicle operating conditions according to claim 6, characterized in that, Based on traffic flow, multi-dimensional driving behavior data is weighted to generate a weighted database, including: Based on traffic flow, the sample space combination units that meet the preset requirements are weighted to generate a weighted database; The sample space combination unit that meets the preset requirements includes: The actual collection volume of the sample space combination unit is greater than or equal to the preset target collection volume; or The flow deviation rate of the sample space combination unit is less than the preset deviation rate; wherein, the flow deviation rate refers to the deviation between the actual sample ratio of the sample space combination unit and the corresponding target traffic flow ratio.

8. The method for constructing typical vehicle operating conditions according to claim 6, characterized in that, The method for constructing typical vehicle operating conditions includes: Traffic flow data for each sample space combination unit is statistically analyzed; wherein, the traffic flow data represents the total number of vehicles passing through the sample space combination unit per unit time. This includes weighting multi-dimensional driving behavior data based on traffic flow to generate a weighted database, which includes: Based on the traffic flow data of each sample space combination unit, a weight value is assigned to the corresponding sample space combination unit; wherein the weight value is proportional to the traffic flow data.

9. A typical vehicle operating condition construction device, characterized in that, include: The generation module is used to generate typical vehicle operating conditions based on a preset double-chain Markov model. The classification module is used to classify the collected driving behavior data based on the road operating environment to obtain multi-dimensional driving behavior data; The weighting module is used to weight multi-dimensional driving behavior data based on traffic flow to generate a weighted database; The partitioning module is used to partition the weighted database into kinematic states to obtain kinematic states, and to partition the speed range to obtain vehicle speed states. A construction module is used to construct a double-chain Markov model based on the kinematic state and the vehicle speed state.

10. A computer-readable storage medium storing a computer program for performing a method for constructing typical vehicle operating conditions as described in any one of claims 1-8.

Images