A neural network-based whole vehicle ride comfort multi-dimensional comprehensive evaluation method

Abstract

The present application relates to the field of vehicle evaluation technology, and relates to a vehicle ride comfort multi-dimensional comprehensive evaluation method based on a neural network. Test data of a test vehicle and user characteristics of a driving evaluator are obtained. Based on the test data, six-dimensional evaluation indexes of the test vehicle are determined. The six-dimensional evaluation indexes and the user characteristics are input into a prediction model to obtain prediction scores of six first-level dimensions respectively. According to the test data, a test scene is determined, and weights of the six first-level dimensions are determined based on the test scene. According to the prediction scores of the six first-level dimensions and the weights, a ride comfort index is determined, and according to the ride comfort index, an evaluation result of the ride comfort of the test vehicle is determined. By establishing a six-dimensional hierarchical index system covering thermal comfort, steering stability, NVH performance, seat comfort, cabin comfort and intelligent driving comfort, a full-dimensional unified evaluation of the intelligent networked vehicle ride comfort is realized.

Description

Technical Field

[0001] This manual relates to the field of vehicle evaluation technology, and in particular to a multi-dimensional comprehensive evaluation method for overall vehicle ride comfort based on neural networks. Background Technology

[0002] Against the backdrop of the automotive industry's rapid development towards intelligence and electrification, consumers' demands for driving comfort are increasing, making it one of the core factors influencing users' car purchase decisions and brand satisfaction. Driving comfort involves multiple dimensions, including thermal comfort, handling stability, NVH (Noise, Vibration, and Harshness) performance, seat comfort, cabin performance, and intelligent driving performance. As a highly complex sensory system, the human body is sensitive to performance in all dimensions, and psychological states and usage scenarios also significantly influence subjective preferences, necessitating a comprehensive and integrated evaluation method.

[0003] However, current evaluation dimensions and mapping models for ride comfort are relatively singular, making it impossible to scientifically, accurately, and efficiently evaluate the overall ride comfort of a vehicle, and thus difficult to effectively answer the question of "how comfortable is this vehicle overall?"

[0004] Therefore, this specification provides a multi-dimensional comprehensive evaluation method for vehicle ride comfort based on neural networks. Summary of the Invention

[0005] This specification provides a multi-dimensional comprehensive evaluation method for vehicle ride comfort based on neural networks, in order to partially solve the aforementioned problems existing in the prior art.

[0006] The following technical solution is adopted in this specification:

[0007] This manual provides a multi-dimensional comprehensive evaluation method for vehicle ride comfort based on neural networks, including:

[0008] S1. Obtain test data measured by various measuring devices pre-installed on the test vehicle, and obtain user characteristics of the driving evaluator on the test vehicle;

[0009] S2. Based on the test data, and according to the preset evaluation method, determine the current six-dimensional evaluation index of the test vehicle. The six-dimensional evaluation index includes six primary dimensions, and each primary dimension includes six secondary indicators. The six primary dimensions are thermal comfort, handling stability, NVH performance, seat comfort, cabin comfort, and intelligent driving performance.

[0010] S3. Input the six-dimensional evaluation index and the user features into the pre-trained prediction model to obtain the predicted scores of the six primary dimensions output by the prediction model.

[0011] S4. Based on the test data, determine the test scenario for the test vehicle, and determine the weights of the six primary dimensions based on the test scenario;

[0012] S5. Based on the predicted scores of the six primary dimensions and the weights of the six primary dimensions, determine the ride comfort index of the test vehicle, and based on the ride comfort index, determine the ride comfort evaluation result of the test vehicle.

[0013] Based on the aforementioned technical means, this solution establishes a six-dimensional hierarchical index system covering thermal comfort, handling stability, NVH performance, seat comfort, cabin comfort, and intelligent driving comfort (i.e., intelligent driving performance), achieving a unified evaluation of the "driving comfort" of intelligent connected vehicles across all dimensions. By inputting the user characteristics of driving evaluators along with the six-dimensional evaluation indicators into a neural network prediction model, the evaluation results can reflect the differences in subjective feelings among different users, achieving personalized comfort prediction that integrates "human-vehicle-scenario" considerations.

[0014] Furthermore, S4 specifically includes:

[0015] Based on the prior experience of experts, the subjective weights of the six primary dimensions are determined respectively; and based on the preset test dataset, the objective weights of the six primary dimensions are determined respectively.

[0016] Based on the subjective and objective weights of the six primary dimensions, determine the combined weights of the six primary dimensions respectively;

[0017] Determine the scenario weight adjustment matrix corresponding to five typical driving scenarios, and determine the test scenario of the test vehicle based on the test data. The test scenario is one of the five typical driving scenarios, namely, urban congestion scenario, highway cruising scenario, mountain road driving scenario, intelligent driving cruise scenario, and long-distance travel scenario.

[0018] Based on the test scenario, the weight adjustment coefficients for the six primary dimensions are determined from the scenario weight adjustment matrix.

[0019] The weights of the six primary dimensions are determined based on their respective combined weights and weight adjustment coefficients.

[0020] Based on the aforementioned technical means, this solution proposes a combined subjective and objective weighting mechanism, and introduces scenario-adaptive weight adjustment, enabling the evaluation results to dynamically change with typical scenarios such as urban congestion, highway cruising, mountain driving, intelligent driving cruise, and long-distance travel. This mechanism solves the problem of "averaging out scenario differences" caused by fixed weights in existing solutions, making the comfort conclusions of the same vehicle in different scenarios more consistent with real user perception, and enhancing the practicality and credibility of the evaluation results.

[0021] Furthermore, based on the prior experience of experts, the subjective weights of the six primary dimensions are determined, specifically including:

[0022] Obtain the evaluation results of experts on the pairwise importance comparison of the six primary dimensions, and the evaluation results are expressed as ternary fuzzy numbers;

[0023] Each ternary fuzzy number is aggregated and defuzzified sequentially to determine a sixth-order judgment matrix;

[0024] Based on the sixth-order judgment matrix, the subjective weights of the six first-level dimensions are determined respectively.

[0025] Furthermore, the pre-set test dataset includes six-dimensional evaluation indicators of test data collected from various types of vehicles in five typical driving scenarios, as well as subjective scores of six primary dimensions given by driving evaluators, and uses one type of vehicle matched with one typical driving scenario as a sample.

[0026] Based on a pre-set test dataset, determine the objective weights for each of the six primary dimensions, specifically including:

[0027] Based on the preset test dataset, the subjective rating matrix of each sample in six primary dimensions is determined, and each subjective rating in the subjective rating matrix is ​​normalized to determine the normalized subjective rating matrix.

[0028] For each primary dimension of each sample, calculate the proportion of the sample’s subjective rating on that primary dimension to the total subjective rating of that primary dimension.

[0029] Based on the calculated proportions, calculate the information entropy of each first-level dimension;

[0030] The information entropy of each first-level dimension is converted into a difference coefficient, and based on the difference coefficient of each first-level dimension, the objective weights of the six first-level dimensions are determined respectively.

[0031] Furthermore, the method further includes step S6:

[0032] Based on the driving comfort index of the test vehicle and the preset target driving comfort index of the test vehicle, a loss value is determined, and the gradient of the input of the prediction model is determined based on the loss value.

[0033] Based on the gradient of the input to the prediction model, the six-dimensional evaluation index of the test vehicle is updated, and the updated six-dimensional evaluation index of the test vehicle is continuously input into the prediction model to iteratively update the six-dimensional evaluation index of the test vehicle until a preset iteration termination condition is reached.

[0034] Furthermore, the method further includes step S7:

[0035] For each secondary indicator in the six-dimensional evaluation index at the end of the iteration, the improvement range of the secondary indicator is determined based on its value at the end of the iteration and its current value.

[0036] Based on the improvement magnitude of the secondary indicator, determine the improvement ratio of the secondary indicator.

[0037] Furthermore, the secondary indicators under thermal comfort include temperature uniformity, thermal response speed, airflow comfort, radiant heat asymmetry, relative humidity suitability, and seat surface thermal comfort; the secondary indicators under handling stability include lateral acceleration comfort, longitudinal acceleration smoothness, roll control, pitch control, steering self-centering, and dynamic response; the secondary indicators under NVH performance include idling cabin noise, constant speed driving noise, acceleration noise quality, road noise, wind noise performance, and vibration comfort; the secondary indicators under seat comfort include body pressure distribution uniformity, hip pressure concentration, lumbar support, seat vibration isolation, material breathability, and long-distance travel fatigue; the secondary indicators under cabin comfort include spaciousness, field of vision, human-machine interaction convenience, cabin odor quality, ambient lighting comfort, and storage space convenience; and the secondary indicators under intelligent driving performance include adaptive cruise smoothness, lane keeping stability, lane change smoothness, takeover transition naturalness, system status transparency, trust, and sense of security.

[0038] This manual provides a multi-dimensional comprehensive evaluation device for vehicle ride comfort based on neural networks, including:

[0039] The acquisition module is used to acquire test data measured by various measuring devices pre-installed on the test vehicle, and to acquire user characteristics of the driving evaluator on the test vehicle;

[0040] The first determining module is used to determine the current six-dimensional evaluation index of the test vehicle based on the test data and according to a preset evaluation method. The six-dimensional evaluation index includes six primary dimensions, and each primary dimension includes six secondary indicators. The six primary dimensions are thermal comfort, handling stability, NVH performance, seat comfort, cabin comfort, and intelligent driving performance.

[0041] The prediction module is used to input the six-dimensional evaluation indicators and the user features into a pre-trained prediction model to obtain the predicted scores of the six primary dimensions output by the prediction model.

[0042] The second determining module is used to determine the test scenario of the test vehicle based on the test data, and to determine the weights of the six primary dimensions based on the test scenario.

[0043] The third determining module is used to determine the ride comfort index of the test vehicle based on the predicted scores of the six primary dimensions and the weights of the six primary dimensions, and to determine the ride comfort evaluation result of the test vehicle based on the ride comfort index.

[0044] This specification provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned method for multi-dimensional comprehensive evaluation of vehicle ride comfort based on neural networks.

[0045] This specification provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements a multi-dimensional comprehensive evaluation method for vehicle ride comfort based on a neural network.

[0046] The above-mentioned technical solutions adopted in this specification can achieve the following beneficial effects:

[0047] This solution establishes a hierarchical six-dimensional indicator system covering thermal comfort, handling stability, NVH performance, seat comfort, cabin comfort, and intelligent driving comfort (i.e., intelligent driving capability), achieving a unified evaluation of the "driving comfort" of intelligent connected vehicles across all dimensions. By inputting the user characteristics of driving evaluators along with the six-dimensional evaluation indicators into a neural network prediction model, the evaluation results can reflect the differences in subjective feelings among different users, achieving personalized comfort prediction that integrates "human-vehicle-scenario" considerations. Attached Figure Description

[0048] The accompanying drawings, which are included to provide a further understanding of this specification and form part of this specification, illustrate exemplary embodiments and are used to explain this specification, but do not constitute an undue limitation thereof. In the drawings:

[0049] Figure 1 A flowchart illustrating a multi-dimensional comprehensive evaluation method for vehicle ride comfort based on neural networks, provided in the embodiments of this specification;

[0050] Figure 2 This is a schematic diagram of a multi-dimensional comprehensive evaluation device for vehicle ride comfort based on neural networks, as provided in this specification.

[0051] Figure 3 This specification provides a corresponding Figure 1 A schematic diagram of the structure of an electronic device. Detailed Implementation

[0052] To make the objectives, technical solutions, and advantages of this specification clearer, the technical solutions of this specification will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this specification, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments in this specification without creative effort are within the scope of protection of this application.

[0053] In embodiments of this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0054] The technical solutions provided in the various embodiments of this specification are described in detail below with reference to the accompanying drawings.

[0055] Figure 1 A flowchart illustrating a multi-dimensional comprehensive evaluation method for vehicle ride comfort based on neural networks, provided in this specification, includes the following steps:

[0056] S1: Obtain test data measured by various measuring devices pre-installed on the test vehicle, and obtain user characteristics of the driving evaluator on the test vehicle.

[0057] This specification describes the process of performing a multi-dimensional comprehensive evaluation of vehicle ride comfort based on neural networks. In the embodiments described herein, this process can be executed by a server. However, this specification does not limit the type of device or platform used to perform this multi-dimensional comprehensive evaluation of vehicle ride comfort based on neural networks; for example, a personal computer, mobile terminal, or other such device or platform can also be used. For ease of description, the following description uses a server as the executing entity.

[0058] In one or more embodiments of this specification, the server may acquire test data measured by various measuring devices pre-installed on the test vehicle, as well as acquire user characteristics of the driving evaluator on the test vehicle.

[0059] The selection of test vehicles can cover different market positioning, such as economy, mid-range, comfort, luxury, and sporty models, and different power forms, including gasoline vehicles, pure electric vehicles, and hybrid vehicles. Therefore, there are various types of test vehicles available for selection. Each test vehicle is equipped with measuring equipment, including temperature sensors, wind speed sensors, radiation thermometers, humidity sensors, inertial measurement units, acoustic acquisition systems, vibration accelerometers, laser rangefinders, body pressure distribution measuring pads, VOC detectors, illuminometers, electrocardiographs, force-measuring steering wheels, and data synchronization acquisition devices.

[0060] Of course, driving evaluators can drive test vehicles in various test scenarios and collect test data through various measuring devices on the test vehicles. Test scenarios can include urban congestion scenarios, characterized by an average vehicle speed of less than 30 km / h and a stopping frequency of more than 0.5 times per minute, with a data collection time of no less than 30 minutes; high-speed cruising scenarios, characterized by an average vehicle speed of more than 100 km / h and a speed fluctuation of less than 5 km / h, with a data collection time of no less than 20 minutes; and mountain road driving scenarios, driving on continuous curved road sections with an average curvature radius of less than 100m, with a data collection time of no less than 15 minutes. The five typical driving scenarios include: intelligent driving cruise scenario, intelligent driving assistance system is activated, L2 or higher level assisted driving function is activated on highway, and the data collection time is not less than 20 minutes; long-distance travel scenario, continuous driving time exceeds 120 minutes, etc. Of course, it may also include test scenarios such as driving at a speed of 40km / h to 60km / h on a standard test road surface and collecting data for not less than 10 minutes on a rough road surface. This manual does not limit the test scenarios.

[0061] The user characteristics of driving evaluators include age, gender, driving experience, height, weight, preference type (such as preference for comfort), and scenario type (i.e., the type of test scenario).

[0062] It is worth noting that, in this manual, test data can be collected by driving various types of test vehicles in various test scenarios (there are 5 test scenarios). If a total of n test vehicles collect test data, n×5 types of 36 secondary indicators can be extracted. Each type of test vehicle can be equipped with no fewer than 10 evaluators, whose age range is between 25 and 55 years old, with a male-to-female ratio of approximately 1:1. After experiencing each test scenario, the evaluators will score each of the six primary dimensions using an eleven-point scale from 0 to 10, and will also score the overall comfort using a scale from 0 to 100.

[0063] Of course, to align the subjective evaluations of different evaluators, eliminate unreliable and outlier data, and obtain an objective "subjective score," a consistency test can be performed on the evaluators' subjective scores. The test method uses Kendall's coefficient of harmony to measure consistency. When the calculated Kendall's coefficient of harmony is greater than 0.5, the consistency is considered acceptable; otherwise, the scores need to be re-evaluated or a new batch of evaluators needs to be selected.

[0064] Let X (36 dimensions) represent the objective feature vector composed of 36 secondary indicators, and let Y represent the subjective vector score composed of the subjective scores of the six primary dimensions given by the evaluators of the test vehicle and test scenario (in the case of multiple evaluators, the average of each primary dimension and the overall comfort is taken). Thus, the test dataset D can be constructed:

[0065]

[0066] Where N is the total number of test samples, equal to the product of the number of vehicles collecting test data and the number of test scenarios. Each test sample corresponds to an X and a Y. The dataset is randomly divided into training, validation, and test sets in an 8:1:1 ratio. Alternatively, it can be subjective ratings from six primary dimensions provided by commentators. ) and overall comfort rating ( )composition.

[0067] S2: Based on the test data, and according to the preset evaluation method, determine the current six-dimensional evaluation index of the test vehicle. The six-dimensional evaluation index includes six primary dimensions, and each primary dimension includes six secondary indicators. The six primary dimensions are thermal comfort, handling stability, NVH performance, seat comfort, cabin comfort, and intelligent driving performance.

[0068] In one or more embodiments of this specification, the server can determine the current six-dimensional evaluation index of the test vehicle based on the acquired test data and according to a preset evaluation method. This six-dimensional evaluation index includes six primary dimensions, each of which includes six secondary indicators. The six primary dimensions are thermal comfort, handling stability, NVH performance, seat comfort, cabin comfort, and intelligent driving performance. That is, a total of 36 secondary indicators. In this six-dimensional evaluation index, each primary dimension is characterized by its six secondary indicators.

[0069] The six primary dimensions of the six-dimensional evaluation index can be denoted as A, B, C, D, E, and F, which correspond to thermal comfort, handling stability, NVH performance, seat comfort, cabin comfort, and intelligent driving performance, respectively.

[0070] In this specification, the evaluation methods for determining the six secondary indicators that characterize each primary dimension can be determined as described below.

[0071] The secondary indicators under dimension A thermal comfort are A1 temperature uniformity, A2 thermal response speed, A3 airflow comfort, A4 radiant heat asymmetry, A5 relative humidity suitability, and A6 seat surface thermal comfort.

[0072] A1 temperature uniformity is the standard deviation of temperature values ​​at preset measuring points within the test vehicle's cabin. Nine measuring points are preset, located on the left and right sides of the driver's head, in front of the chest, in front of the abdomen, at the left and right ankles, at the front passenger's head, and on the left and right sides of the rear passenger's head. The calculation formula is:

[0073]

[0074] in For the first Temperature values ​​at each measuring point A1 is the arithmetic mean of the temperatures at nine measuring points, and its unit is degrees Celsius.

[0075] The A2 thermal response speed, taking cabin temperature rise as an example, is the time required for the average cabin temperature to reach 90% of the difference between the set temperature and the initial temperature after the vehicle's air conditioning system is started, measured in seconds.

[0076] A3 airflow comfort is defined as the average airflow velocity in the area around the driver's head, measured in meters per second.

[0077] A4 radiative thermal asymmetry is defined as the absolute value of the difference in average radiative temperature between the left and right sides of the cabin, expressed in degrees Celsius.

[0078] A5 relative humidity suitability is defined as the relative humidity of the air inside the cabin, expressed as a percentage.

[0079] A6 seat surface thermal comfort is defined as the surface temperature of the contact area of ​​the center area of ​​the seat cushion, in degrees Celsius.

[0080] The secondary indicators under the B-dimensional handling stability are B1 lateral acceleration comfort, B2 longitudinal acceleration smoothness, B3 roll control, B4 pitch control, B5 steering self-centering, and B6 dynamic responsiveness.

[0081] B1 Lateral acceleration comfort is defined as the root mean square value of the lateral acceleration of the test vehicle when cornering, measured in meters per second squared.

[0082] B2 longitudinal acceleration smoothness is defined as the rate of change of longitudinal acceleration with respect to time during the acceleration or deceleration of the test vehicle. The calculation formula is as follows:

[0083]

[0084] in For longitudinal acceleration, For measuring time, B2 is measured in meters per cubic second.

[0085] B3 roll control is defined as the ratio of the vehicle's roll angle to its lateral acceleration during steady-state cornering, measured in degrees per second of gravitational acceleration.

[0086] B4 pitch control is defined as the maximum change in the vehicle's pitch angle during braking or acceleration, measured in degrees.

[0087] B5 steering self-centering is defined as the time required for the steering wheel to return to the center position after being released from the deflection position, measured in seconds.

[0088] B6 dynamic responsiveness is defined as the time required for the yaw rate to reach 90% of its steady-state value after a step steering input, in milliseconds.

[0089] The secondary indicators under the C-dimensional NVH performance are C1 Idle in-vehicle noise, C2 Constant speed driving noise, C3 Acceleration noise quality, C4 Road noise, C5 Wind noise performance, and C6 Vibration comfort.

[0090] C1 is the in-vehicle noise at idle speed, which is defined as the A-weighted equivalent continuous sound pressure level at the driver's right ear position when the engine is idling.

[0091] C2 is the noise level at a constant speed, defined as the A-weighted equivalent continuous sound pressure level at the driver's right ear position when the vehicle is traveling at a constant speed of 120 km / h.

[0092] C3 acceleration noise quality is defined as the A-weighted maximum sound pressure level at the driver's right ear position under full-throttle acceleration conditions.

[0093] C4 road noise is defined as the A-weighted equivalent continuous sound pressure level at the driver's right ear position when traveling at a specified speed on a standard rough road surface.

[0094] The C5 wind noise performance is defined as the A-weighted sound pressure level at the driver's right ear position caused by airflow excitation when the vehicle is traveling at a constant speed of 120 km / h.

[0095] C6 vibration comfort is defined as the root mean square value of frequency-weighted acceleration at the surface of the seat cushion, calculated according to ISO 2631 standard. The calculation formula is as follows:

[0096]

[0097] in The acceleration signal after frequency weighting processing. For measuring time, C6 is measured in meters per second squared.

[0098] The secondary indicators under the D-dimensional seat comfort are D1 uniformity of body pressure distribution, D2 concentration of hip pressure, D3 lumbar support, D4 seat vibration isolation, D5 material breathability, and D6 fatigue during long-distance travel.

[0099] D1 is the uniformity of body pressure distribution, which is defined as the standard deviation of pressure values ​​in the contact area of ​​the seat cushion, in kilopascals.

[0100] D2 hip pressure concentration is defined as the maximum pressure value in the area directly below the ischial tuberosity of the human body, measured in kilopascals.

[0101] D3 lumbar support is defined as the average support force in the contact area between the lumbar support and the human waist, measured in Newtons.

[0102] D4 Seat Vibration Isolation is defined as the vibration transmission rate between the seat cushion surface and the seat mounting point, and is calculated using the following formula:

[0103]

[0104] in The root mean square value of frequency-weighted acceleration at the surface of the seat cushion. The root mean square value of the frequency-weighted acceleration at the seat mounting point.

[0105] D5 material breathability is defined as the water vapor permeability of the seat surface covering material, measured in grams per square meter per hour.

[0106] D6 Long-Distance Travel Fatigue Index is defined as the cumulative discomfort index obtained by assessing the change in body pressure distribution after two hours of continuous travel, with a value ranging from 1 to 10.

[0107] The secondary indicators under the E-dimensional cabin comfort are E1 spaciousness, E2 field of vision, E3 human-machine interaction convenience, E4 cabin odor quality, E5 ambient lighting comfort, and E6 storage space convenience.

[0108] E1 space comfort is defined as the vertical distance from the driver evaluator's point H to the inner surface of the test vehicle's roof, in millimeters.

[0109] E2 field of vision is defined as the angle of obstruction caused by the A-pillar to the driver's left frontal field of vision, measured in degrees.

[0110] E3 Human-Computer Interaction Convenience is defined as a comprehensive score of the accessibility of the main operating controls, derived through human factors engineering analysis, with a value range of 3 to 10.

[0111] E4 cabin odor quality is defined as the total concentration of volatile organic compounds in the cabin, expressed in micrograms per cubic meter.

[0112] The E5 ambient lighting comfort rating is defined as a comprehensive score of the cabin lighting system's illuminance uniformity and glare control, with a value ranging from 3 to 10 points.

[0113] E6 storage space convenience is defined as the total volume of available storage space in the front row, in liters.

[0114] The secondary indicators under the F-dimensional intelligent driving performance are F1 adaptive cruise smoothness, F2 lane keeping stability, F3 lane change smoothness, F4 takeover transition naturalness, F5 system status transparency, and F6 trust and sense of security.

[0115] F1 adaptive cruise comfort is defined as the root mean square value of the rate of change of longitudinal acceleration in adaptive cruise control mode, in meters per cubic second.

[0116] F2 lane keeping stability is defined as the root mean square value of the vehicle's lateral position deviation in lane keeping assist mode, in meters.

[0117] F3 lane change smoothness is defined as the peak value of lateral acceleration during automatic lane change, measured in meters per second squared.

[0118] F4 takeover transition naturalness is defined as the time required for control to be transferred when switching from autonomous driving mode to manual driving mode, in seconds.

[0119] F5 system status transparency is defined as a score given by the human-machine interface on the completeness of the presentation of the intelligent driving system's operating status information, with a value ranging from 3 to 10.

[0120] F6 Trust and Safety is defined as the rate of change of the driver's heart rate variability relative to the baseline value when the driver is in intelligent driving mode, expressed as a percentage.

[0121] It is worth noting that the methods for determining the above 36 secondary indicators can all adopt existing solutions, and are not restricted in this specification, nor are they the protection points that this solution intends to protect.

[0122] S3: Input the six-dimensional evaluation index and the user features into the pre-trained prediction model to obtain the predicted scores of the six primary dimensions output by the prediction model.

[0123] In one or more embodiments of this specification, the server can input the determined six-dimensional evaluation index and user features into a pre-trained prediction model to obtain the prediction scores for the six primary dimensions output by the trained model.

[0124] S4: Based on the test data, determine the test scenario for the test vehicle, and determine the weights of the six primary dimensions based on the test scenario.

[0125] In one or more embodiments of this specification, the server can determine the test scenario for collecting test data from the acquired test data. Then, based on the determined test scenario, it determines the weights for each of the six primary dimensions. These weights for the six primary dimensions can be a set of preset weights.

[0126] S5: Determine the ride comfort index of the test vehicle based on the predicted scores of the six primary dimensions and the weights of the six primary dimensions, and determine the ride comfort evaluation result of the test vehicle based on the ride comfort index.

[0127] In one or more embodiments of this specification, the server can determine the driving comfort index of the test vehicle based on the predicted scores of the six primary dimensions and the weights of the six primary dimensions, and determine the driving comfort evaluation result of the test vehicle based on the driving comfort index. The higher the driving comfort index, the better the driving comfort evaluation result.

[0128] based on Figure 1This paper presents a multi-dimensional comprehensive evaluation method for vehicle ride comfort based on neural networks. By establishing a six-dimensional hierarchical index system covering thermal comfort, handling stability, NVH performance, seat comfort, cabin comfort, and intelligent driving comfort (i.e., intelligent driving capability), it achieves a unified evaluation of the "ride comfort" of intelligent connected vehicles across all dimensions. By inputting the user characteristics of driving evaluators along with the six-dimensional evaluation indicators into the neural network prediction model, the evaluation results can reflect the differences in subjective feelings among different users, achieving personalized comfort prediction that integrates "human-vehicle-scenario" considerations.

[0129] Furthermore, in one or more embodiments of this specification, S4 may specifically involve the server determining the subjective weights of the six primary dimensions based on preset expert prior experience, and determining the objective weights of the six primary dimensions based on a preset test dataset D (i.e., the test dataset D constructed above).

[0130] The method for determining the subjective weights of the six primary dimensions is as follows: the server obtains the evaluation results of experts on the pairwise importance comparisons of the six primary dimensions. The evaluation results are expressed as ternary fuzzy numbers, such as the evaluation result of the importance between the i-th primary dimension and the j-th (i,j∈[1,6]) primary dimension. , For the first The minimum value of the pre-determined assessment results by the experts. For the first The reference value, or most likely value, of the assessment results predetermined by the experts. For the first The maximum value of the pre-determined evaluation results by the experts. The correspondence between the linguistic description and the triangular fuzzy number is as follows: equally important (1,1,1), slightly important (2,3,4), significantly important (4,5,6), strongly important (6,7,8), and extremely important (8,9,9).

[0131] Then, the ternary fuzzy numbers are aggregated and defuzzified in sequence to determine the sixth-order judgment matrix.

[0132] Given N experts, the expression for aggregating and defuzzifying the ternary fuzzy numbers is as follows:

[0133]

[0134]

[0135]

[0136] In the formula, To The result after aggregation. To The result after aggregation processing , Indicates to The result after aggregation processing , Indicates to The result after aggregation processing . for The result after deblurring. It is a sixth-order judgment matrix.

[0137] Finally, the server can determine the subjective weights of the six primary dimensions based on the sixth-order judgment matrix.

[0138]

[0139]

[0140]

[0141] In the formula, This represents the geometric mean of the elements in each row of a sixth-order judgment matrix. This represents the subjective weight of the first-level dimension i. This represents the subjective weight of each first-level dimension.

[0142] Before determining the subjective weights, the consistency ratio of the sixth-order judgment matrix can be calculated. If the consistency ratio is less than 0.1, the consistency test is passed; otherwise, the sixth-order judgment matrix needs to be corrected. Each element of the corrected sixth-order judgment matrix is ​​then substituted into the calculation. In the formula.

[0143] In addition, after determining the subjective weights, the method for determining the objective weights of the six primary dimensions is to determine a pre-set test dataset, which includes the six-dimensional evaluation indicators of test data collected from various types of vehicles in five typical driving scenarios and the subjective scores of the six primary dimensions given by driving evaluators, and to match one type of vehicle with one typical driving scenario as a sample.

[0144] The server determines an N×6 subjective rating matrix for each sample across six primary dimensions based on a pre-set test dataset, and normalizes each subjective rating in the subjective rating matrix to obtain the normalized subjective rating matrix.

[0145] Next, for each sample and each first-level dimension, calculate the proportion of that sample's subjective rating on that first-level dimension to the total subjective ratings for that first-level dimension. Then, based on the calculated proportions, calculate the information entropy for each first-level dimension. The calculation formula is:

[0146]

[0147]

[0148] In the formula, Let i be the subjective rating of the i-th sample in the j-th first-level dimension after normalization, where i,j∈[1,6]. The percentage of the score of the i-th sample on the j-th dimension in the total score of the j-th first-level dimension. Let be the information entropy of the j-th first-level dimension. N is the total number of samples. When Definition when equal to zero It equals zero.

[0149] Finally, the server converts the information entropy of each first-level dimension into a dissimilarity coefficient, and based on the dissimilarity coefficient of each first-level dimension, determines the objective weights of the six first-level dimensions respectively. The formula is:

[0150]

[0151]

[0152]

[0153] In the formula, Let be the variance coefficient of the j-th first-level dimension. Let be the objective weight of the j-th first-level dimension. This represents the objective weight of each first-level dimension.

[0154] Therefore, after determining the subjective and objective weights of the six primary dimensions, the server can determine the combined weights of the six primary dimensions based on their respective subjective and objective weights.

[0155] Specifically, combined weights It can be set as:

[0156]

[0157] These are the combination coefficients, with values ​​ranging from zero to one.

[0158] The optimization objective is to minimize the total deviation between the combined weights and the subjective and objective weights.

[0159]

[0160] set up ,Will Substitute into the objective function and expand:

[0161]

[0162] right Take the derivative and set it to zero:

[0163]

[0164] The Nash equilibrium combination coefficients are obtained by solving:

[0165]

[0166] Will Substitute into the calculation to determine the final combined weights:

[0167]

[0168] The combined weights are normalized to ensure that the sum of the weights equals 1. Let j be the combined weight of the j-th first-level dimension, which after normalization is: :

[0169]

[0170] Next, the server determines the scenario weight adjustment matrix corresponding to five typical driving scenarios, and determines the test scenario for the test vehicle based on the test data. The test scenario is one of the five typical driving scenarios. Urban congestion scenarios High-speed cruising scenario, Mountain road driving scenarios Intelligent driving cruise scenario, Long-distance travel scenario.

[0171] Scene weight adjustment matrix The expression is:

[0172]

[0173] Scene weight adjustment matrix Line 1 Column elements Indicates in the scene Next The adjustment coefficients for the weights of the primary dimensions. The scene weight adjustment matrix corresponds to five scenarios in order: urban congestion, highway cruising, mountain driving, intelligent driving cruise, and long-distance travel. The columns correspond to six primary dimensions in order: thermal comfort, handling stability, NVH performance, seat comfort, cabin comfort, and intelligent driving performance. An adjustment coefficient greater than 1 indicates that this primary dimension is more important in that scenario, and its weight is increased; an adjustment coefficient equal to 1 indicates that its importance remains unchanged; an adjustment coefficient less than 1 indicates that this primary dimension is relatively less important, and its weight is reduced.

[0174] Test data may include vehicle condition parameters, such as average vehicle speed. lateral acceleration peak Continuous driving time Parking frequency Intelligent driving system status The test scenario can be determined based on the following logic:

[0175] like If it is in an active state, then the current scene is .

[0176] Otherwise, if km / h and times / minute, then the current scenario is .

[0177] Otherwise, if If km / h, then further judgment is needed: The current scenario is in minutes. Otherwise, the current scenario is .

[0178] Otherwise, if If the m / s² value occurs frequently, then the current scenario is... .

[0179] If none of the above conditions are met, the system will default to a high-speed cruise scenario, and the current scenario will be [missing information]. .

[0180] Finally, based on the test scenario, the server determines the weight adjustment coefficients for each of the six primary dimensions from the scenario weight adjustment matrix. Then, based on the combined weights and weight adjustment coefficients for each of the six primary dimensions, the weights for each of the six primary dimensions are determined.

[0181] Specifically, after determining the test scenario, the server adjusts the scenario weight matrix. Extract the first Rows as adjustment coefficient vectors .

[0182] Calculate the adjusted weights:

[0183]

[0184] Normalize the adjusted weights:

[0185]

[0186] Get the scene The final weight vector Each weight in the final weight vector is then used as a weight for one of the six first-level dimensions.

[0187] In one or more embodiments of this specification, the test dataset D described above can be used to train a prediction model. The prediction model is selected as a Physical Information Neural Network (PINN) model, abbreviated as PINN. The input of the PINN is designed to have two parts, one part being an objective feature vector. The dimension is 36, corresponding to the measurement values ​​of 36 secondary indicators. The second part is the user feature vector. The dimensions are seven, including age. ,gender Driving experience ,height ,weight Preference type Scene type .

[0188] The input is selected from the above test dataset. In Concatenate the user feature vectors of the commentators (in the case of multiple commentators, select the one that gave the closest subjective rating). The commentator's user characteristics are used as the user feature vector. The output of the prediction model, PINN, is used as a label. Based on the difference between the prediction model's output and the label, a loss value is determined, and the prediction model is trained with the minimum loss value as the optimization objective.

[0189] Of course, in this specification, the input features can also be standardized before inputting PINN. The standardization formula is as follows:

[0190]

[0191] in For the first The mean of each feature on the test dataset The standard deviation is denoted as .

[0192] The PINN architecture employs a fully connected structure, comprising an input layer, hidden layers, and an output layer. The input layer has a dimension of 36, corresponding to the concatenation of 36 objective features and 7 user features. The hidden layer consists of four fully connected layers, each sequentially performing linear transformation, batch normalization, activation function operation, and random deactivation. The first hidden layer has an input dimension of 43 and an output dimension of 128. The second hidden layer has an input dimension of 128 and an output dimension of 64. The third hidden layer has an input dimension of 64 and an output dimension of 32. The fourth hidden layer has an input dimension of 32 and an output dimension of 16. The activation function used is the SiLU function. The output layer adopts a multi-task structure, including two output branches: the first output branch predicts the subjective scores for the six primary dimensions. It maps the output of the fourth hidden layer to six dimensions through a linear transformation, then maps it to the 0-1 interval using the Sigmoid function, and multiplies it by 10 to obtain the dimension scores ranging from 0 to 10. The second output branch is used to predict the Comprehensive Comfort Index (CCI). It maps the output of the fourth hidden layer to one dimension through a linear transformation, then maps it to the 0-1 range using the Sigmoid function, and multiplies it by 100 to obtain a comprehensive score ranging from 0 to 100. .

[0193] The training objective of a physical information neural network is to minimize the total loss function. It includes four parts:

[0194]

[0195] in , , These are the preset constraint weight coefficients.

[0196] Data fitting loss The calculation method is as follows:

[0197]

[0198] in For the sample size, For the first Dimensional predicted score, For the true score (i.e., the subjective rating in the test dataset) ), The overall comfort index is predicted. The true values ​​(i.e., subjective ratings in the test dataset) ).

[0199] Stevens power-law constrained loss The calculation method is as follows:

[0200] Constructing Constraint Loss:

[0201]

[0202] in The number of secondary indicators participating in the constraints. For the first The values ​​of the secondary indicators, For the subjective scores corresponding to the first-level dimension, This represents the partial derivative of the subjective score with respect to the secondary indicators. For the first The perceptual power index of each secondary indicator.

[0203] The power exponent values ​​are as follows: 0.67 for sound pressure indicators C1 to C5, 0.95 for vibration indicators C6 and B1, 1.0 for temperature indicators A1 to A6, 1.1 for pressure indicators D1 and D2, and 1.0 for other indicators.

[0204] Monotonicity constraint loss The calculation method is as follows:

[0205] For cost-related metrics, the subjective comfort score should monotonically decrease as the metric value increases, meaning the partial derivative should be negative. The ReLU function is used to penalize positive gradients.

[0206]

[0207] in This is a set of indexes for cost-related indicators. The number of indicators in the set. .

[0208] Bounded constraint loss The calculation method is as follows:

[0209]

[0210] The constraint weight coefficients adopt a gradual adjustment strategy, and the calculation formula is as follows:

[0211]

[0212] in For the current training round, It is a time constant. This is the maximum weight value.

[0213] In one or more embodiments of this specification, the 36 secondary indicators are divided into three types according to their relationship with comfort:

[0214] Benefit-type indicators are those whose higher values ​​indicate better comfort, including D5, E1, E3, E5, E6, and F5.

[0215] Cost-related indicators are those where a smaller value indicates better comfort. These include A1, A2, A4, B1, B2, B3, B4, B5, B6, C1, C2, C3, C4, C5, C6, D1, D2, D4, D6, E2, E4, F1, F2, F3, F4, and F6.

[0216] The optimal index is an index with an optimal value, including A3, A5, A6, and D3. The optimal value of A3 is 0.3 m / s, the optimal value of A5 is 50%, the optimal value of A6 is 28℃, and the optimal value of D3 is 60 N.

[0217] In one or more embodiments of this specification, the expression for determining the ride comfort index of the test vehicle based on the predicted scores of the six primary dimensions and the weights of the six primary dimensions is as follows:

[0218]

[0219] in These correspond to the six primary dimensions A, B, C, D, E, and F, respectively. Since the sum of the weights equals one and the scores for each dimension range from 0 to 10, therefore... The value range is from 0 to 100.

[0220] Furthermore, it can also be based on The evaluation level is determined by the value of the rating:

[0221] when A score between 90 and 100 is rated A+, indicating an excellent level.

[0222] when A score between 80 and 90 is graded A, indicating an excellent level.

[0223] when A score between 70 and 80 is graded as B, indicating a good level.

[0224] when A score between 60 and 70 is graded as C, indicating a passing level.

[0225] when A score between 50 and 60 is graded as D, indicating a poor level.

[0226] when A score below 50 is grade E, indicating failure.

[0227] Furthermore, the evaluation results for driving and riding comfort include: the overall comfort index. And the corresponding levels, and the scores for each of the six primary dimensions. to It may also include radar chart representations of scores for six primary dimensions, weak point dimension identification, defined as dimensions with scores below six, etc.

[0228] In one or more embodiments of this specification, steps S1 to S5 address the problem of "how to evaluate the comfort of a vehicle." S1-S5 can only tell us "how many points a vehicle scored," but cannot determine "how the specific indicators should be modified to obtain a certain score." Therefore, this specification also includes step S6:

[0229] The server determines the loss value based on the test vehicle's ride comfort index and the preset target ride comfort index, and then determines the gradient of the prediction model's input based on the loss value.

[0230] Then, based on the gradient of the input to the prediction model, the six-dimensional evaluation index of the test vehicle is updated, and the updated six-dimensional evaluation index of the test vehicle is continuously input into the prediction model to iteratively update the six-dimensional evaluation index of the test vehicle until the preset iteration termination condition is reached.

[0231] It is worth noting that for each iteration update, the loss value... Let t represent the t-th iteration. The gradient of the prediction model's input is determined based on the loss value, i.e., the gradient of the loss value with respect to the input parameters is calculated. .

[0232]

[0233] Based on gradient Update the 36 secondary metrics of the input prediction model:

[0234]

[0235] In the formula, and These are the 36 secondary indicators input into the prediction model at iteration t+1 and t, respectively. The learning rate controls the step size for each update.

[0236] Then Project onto the feasible region:

[0237]

[0238] and The first The updated lower and upper limits for each secondary indicator. It may be beyond the scope of engineering feasibility (e.g., temperature uniformity becomes negative), in which case it is "pulled back" to the nearest boundary value.

[0239] Finally, the iteration termination condition:

[0240] like Then the iteration terminates, where This is the preset convergence threshold.

[0241] If the number of iterations Exceeding the maximum number of iterations Then the iteration terminates.

[0242] If the gradient norm The iteration terminates if the value is less than the gradient threshold.

[0243] If none of the above three conditions are met, the next iteration continues.

[0244] The optimized vector is obtained after the iteration. This refers to a set of secondary indicator values ​​required to achieve the target driving comfort index.

[0245] Among them, the target driving comfort index is set according to the market positioning of the vehicle model. :

[0246] The target of economy cars No less than 60 points;

[0247] The goal of comfort models No less than 75 points;

[0248] The target of luxury models No less than 85 points;

[0249] The target of flagship models No less than 90 points.

[0250] Or set based on competitor benchmarking results. Equal to the main competitors Add three points to the value.

[0251] Furthermore, this specification may also provide details for each secondary indicator. Calculate its pair Sensitivity:

[0252]

[0253] Sensitivity measures how much the Comprehensive Comfort Index (CCI) changes when a secondary indicator changes by a small amount. Sensitivity is calculated at the current design point. This indicates that the current six-dimensional evaluation index for the test vehicle includes 36 secondary indicators. Furthermore, these 36 secondary indicators can be sorted from highest to lowest sensitivity.

[0254] Furthermore, the server can also determine the improvement magnitude of each secondary indicator in the six-dimensional evaluation metrics at the end of the iteration, based on the value of that secondary indicator at the end of the iteration and its current value. Then, based on the improvement magnitude of that secondary indicator, the improvement ratio of that secondary indicator is determined.

[0255] The improvement margin for each secondary indicator is calculated as follows:

[0256]

[0257] Calculate the percentage improvement:

[0258]

[0259] In the formula, This represents the k-th secondary index in the six-dimensional evaluation index at the end of the iteration. This indicates the kth secondary indicator included in the current six-dimensional evaluation index of the test vehicle. This indicates the improvement margin of the k-th secondary indicator, that is, how much the k-th secondary indicator needs to change from its current value to the target value. This represents the improvement ratio of the k-th secondary indicator, that is, the ratio of the magnitude of change required for the k-th secondary indicator to the current value.

[0260] Therefore, prioritization can be determined by combining the sensitivity ranking and the extent of improvement:

[0261] Secondary indicators that rank in the top third in terms of sensitivity and show an improvement percentage greater than 10% are given high priority.

[0262] Secondary indicators with sensitivity ranking in the middle third or with an improvement percentage between 5% and 10% are given a medium priority.

[0263] Other secondary indicators have low priority.

[0264] In summary, the server can output a list of development goals, listing the following for each secondary metric: metric number, metric name, and current value. Target value Scope of improvement Sensitivity Priority level.

[0265] In the development target list, prioritize them from highest to lowest priority, and within the same priority level, prioritize them from highest to lowest sensitivity.

[0266] This solution establishes a hierarchical six-dimensional indicator system covering thermal comfort, handling stability, NVH performance, seat comfort, cabin comfort, and intelligent driving capabilities, achieving a unified evaluation of the "driving comfort" of intelligent connected vehicles across all dimensions. Compared to existing technologies that use different standards for each dimension, making it difficult to reach a unified conclusion for the entire vehicle, this solution outputs a comprehensive comfort index (CCI) (0-100 points) using a unified quantitative caliber. This directly answers the question "How comfortable is the overall vehicle?" and simultaneously identifies shortcomings in each of the six dimensions, providing a consistent benchmark for product definition and benchmarking.

[0267] This approach incorporates psychophysical principles into the subjective-objective mapping modeling process. Constraints such as Stevens power law, monotonicity, and boundedness (and optionally, "optimization constraints" for optimality indicators) are embedded into the neural network training loss function. This ensures that the model satisfies fundamental perceptual laws while fitting subjective ratings. This reduces the risk of "contrary-to-common sense mappings" that might occur with purely black-box models (e.g., increased noise / vibration predicting greater comfort). Furthermore, it significantly improves the interpretability of the model output. Engineers can use gradient and sensitivity analysis to determine "which objective indicators have the greatest impact on comfort and in what direction," thus transforming the model results into actionable improvement strategies.

[0268] By incorporating physical constraints and prior structures, this scheme achieves high predictive stability and generalization ability even in engineering scenarios with limited sample size. Unlike traditional machine learning methods that typically rely on a large number of labeled samples, this scheme can suppress overfitting and maintain reasonable monotonic / bounded output even with small to medium sample sizes. It is suitable for scenarios where data is insufficient in the early development stages of new vehicle models, but rapid comfort predictions and trend judgments are required, thereby improving R&D efficiency and reducing the cost of repeated subjective experiments.

[0269] This solution proposes a weighting mechanism combining subjective and objective weights, and introduces a scenario-adaptive weight adjustment matrix. This allows the evaluation results to dynamically change according to typical scenarios such as urban congestion, highway cruising, mountain driving, intelligent driving cruise, and long-distance travel. This mechanism solves the problem of "averaging out scenario differences" caused by fixed weights in existing technologies, making the comfort conclusions of the same vehicle in different scenarios more consistent with real user perception, and enhancing the practicality and credibility of the evaluation results.

[0270] This solution further provides a reverse solution and target back-calculation method based on the PINN model. It can deduce the target range of key objective indicators from the target CCI, and provide optimization priorities by combining sensitivity and improvement magnitude. This expands comfort evaluation from "post-event scoring" to "pre-event target setting and process guidance," forming an engineering closed loop of "evaluation—target setting—optimization—re-evaluation," thereby shortening the development iteration cycle and improving resource investment efficiency.

[0271] Based on the neural network-based multi-dimensional comprehensive evaluation method for vehicle ride comfort provided in one or more embodiments of this specification, following the same approach, this specification also provides a corresponding neural network-based multi-dimensional comprehensive evaluation device for vehicle ride comfort, such as... Figure 2 As shown.

[0272] Figure 2 This specification provides a schematic diagram of a multi-dimensional comprehensive evaluation device for vehicle ride comfort based on neural networks, specifically including:

[0273] The acquisition module 200 is used to acquire test data measured by various measuring devices pre-installed on the test vehicle, and to acquire user characteristics of the driving evaluator on the test vehicle.

[0274] The first determining module 202 is used to determine the current six-dimensional evaluation index of the test vehicle based on the test data and according to a preset evaluation method. The six-dimensional evaluation index includes six primary dimensions, and each primary dimension includes six secondary indicators. The six primary dimensions are thermal comfort, handling stability, NVH performance, seat comfort, cabin comfort, and intelligent driving performance.

[0275] Prediction module 204 is used to input the six-dimensional evaluation index and the user features into a pre-trained prediction model to obtain the predicted scores of the six primary dimensions output by the prediction model.

[0276] The second determining module 206 is used to determine the test scenario of the test vehicle based on the test data, and to determine the weights of the six primary dimensions based on the test scenario.

[0277] The third determining module 208 is used to determine the driving comfort index of the test vehicle based on the predicted scores of the six primary dimensions and the weights of the six primary dimensions, and to determine the driving comfort evaluation result of the test vehicle based on the driving comfort index.

[0278] Optionally, the second determining module 206 is further configured to: determine the subjective weights of the six primary dimensions based on the prior experience of experts; determine the objective weights of the six primary dimensions based on a preset test dataset; determine the combined weights of the six primary dimensions based on the subjective and objective weights of the six primary dimensions; determine the scenario weight adjustment matrix corresponding to the five typical driving scenarios; and determine the test scenario of the test vehicle based on the test data, wherein the test scenario is one of the five typical driving scenarios, namely, urban congestion scenario, highway cruising scenario, mountain road driving scenario, intelligent driving cruise scenario, and long-distance travel scenario; determine the weight adjustment coefficients of the six primary dimensions from the scenario weight adjustment matrix based on the test scenario; and determine the weights of the six primary dimensions based on the combined weights and weight adjustment coefficients of the six primary dimensions.

[0279] Optionally, the second determining module 206 is further configured to obtain the evaluation results of the experts on the pairwise importance comparison of the six primary dimensions, wherein the evaluation results are expressed as ternary fuzzy numbers, and each ternary fuzzy number is aggregated and defuzzified in sequence to determine a sixth-order judgment matrix, and the subjective weights of the six primary dimensions are determined according to the sixth-order judgment matrix.

[0280] Optionally, the preset test dataset includes six-dimensional evaluation indicators of test data collected from various types of vehicles in five typical driving scenarios and subjective scores of six primary dimensions given by driving evaluators, and uses one type of vehicle matched with one typical driving scenario as a sample.

[0281] The second determining module 206 is further configured to determine the subjective rating matrix of each sample in six primary dimensions according to the preset test dataset, and normalize each subjective rating in the subjective rating matrix to determine the normalized subjective rating matrix. For each primary dimension of each sample, the proportion of the subjective rating of the sample in that primary dimension to the total subjective rating of that primary dimension is calculated. Based on the calculated proportions, the information entropy of each primary dimension is calculated, the information entropy of each primary dimension is converted into a difference coefficient, and the objective weights of the six primary dimensions are determined based on the difference coefficients of each primary dimension.

[0282] Optionally, the device further includes an update module 210;

[0283] The update module 210 is used to determine a loss value based on the driving comfort index of the test vehicle and the preset target driving comfort index of the test vehicle, determine the gradient of the input of the prediction model based on the loss value, update the six-dimensional evaluation index of the test vehicle based on the gradient of the input of the prediction model, and continue to input the updated six-dimensional evaluation index of the test vehicle into the prediction model to iteratively update the six-dimensional evaluation index of the test vehicle until a preset iteration termination condition is reached.

[0284] Optionally, the device further includes a fourth determining module 212;

[0285] The fourth determining module 212 is used to determine the improvement range of each secondary indicator in the six-dimensional evaluation index at the end of the iteration, based on the value of the secondary indicator at the end of the iteration and the current value, and to determine the improvement ratio of the secondary indicator based on the improvement range of the secondary indicator.

[0286] Optionally, the secondary indicators under thermal comfort in the first determining module 202 are temperature uniformity, thermal response speed, airflow comfort, radiant heat asymmetry, relative humidity suitability, and seat surface thermal comfort; the secondary indicators under handling stability are lateral acceleration comfort, longitudinal acceleration smoothness, roll control, pitch control, steering self-centering, and dynamic response; the secondary indicators under NVH performance are idling in-vehicle noise, constant speed driving noise, acceleration noise quality, road noise, wind noise performance, and vibration comfort; the secondary indicators under seat comfort are body pressure distribution uniformity, hip pressure concentration, lumbar support, seat vibration isolation, material breathability, and long-distance riding fatigue; the secondary indicators under cabin comfort are spaciousness, field of vision, human-machine interaction convenience, cabin odor quality, ambient lighting comfort, and storage space convenience; and the secondary indicators under intelligent driving performance are adaptive cruise smoothness, lane keeping stability, lane change smoothness, takeover transition naturalness, system status transparency, trust, and sense of security.

[0287] This specification also provides a computer-readable storage medium storing a computer program that can be used to execute the above-described... Figure 1 This paper presents a multi-dimensional comprehensive evaluation method for vehicle ride comfort based on neural networks.

[0288] This instruction manual also provides Figure 3 The diagram shows a schematic structural representation of the electronic device. Figure 3As shown, at the hardware level, this electronic device includes a processor, internal bus, network interface, memory, and non-volatile memory, and may also include other hardware required for business operations. The processor reads the corresponding computer program from the non-volatile memory into memory and then runs it to achieve the above. Figure 1 This paper presents a multi-dimensional comprehensive evaluation method for vehicle ride comfort based on neural networks.

[0289] Of course, in addition to software implementation, this specification does not exclude other implementation methods, such as logic devices or a combination of hardware and software. In other words, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0290] In the 1990s, improvements to a technology could be clearly distinguished as either hardware improvements (e.g., improvements to the circuit structure of diodes, transistors, switches, etc.) or software improvements (improvements to the methodology). However, with technological advancements, many methodological improvements today can be considered direct improvements to the hardware circuit structure. Designers almost always obtain the corresponding hardware circuit structure by programming the improved methodology into the hardware circuit. Therefore, it cannot be said that a methodological improvement cannot be implemented using hardware physical modules. For example, a Programmable Logic Device (PLD) (such as a Field Programmable Gate Array (FPGA)) is such an integrated circuit whose logic function is determined by the user programming the device. Designers can program and "integrate" a digital system onto a PLD themselves, without needing chip manufacturers to design and manufacture dedicated integrated circuit chips. Furthermore, nowadays, instead of manually manufacturing integrated circuit chips, this programming is mostly implemented using "logic compiler" software. Similar to the software compiler used in program development, the original code before compilation must also be written in a specific programming language, called a Hardware Description Language (HDL). There are many HDLs, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, and RHDL (Ruby Hardware Description Language). Currently, the most commonly used are VHDL (Very-High-Speed ​​Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should also understand that by simply performing some logic programming on the method flow using one of these hardware description languages ​​and programming it into an integrated circuit, the hardware circuit implementing the logical method flow can be easily obtained.

[0291] The controller can be implemented in any suitable manner. For example, it can take the form of a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro)processor, logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers, and embedded microcontrollers. Examples of controllers include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicon Labs C8051F320. A memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art will also recognize that, in addition to implementing the controller in purely computer-readable program code form, the same functionality can be achieved by logically programming the method steps to make the controller take the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, such a controller can be considered a hardware component, and the means included therein for implementing various functions can also be considered as structures within the hardware component. Alternatively, the means for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0292] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, a computer can be, for example, a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email device, game console, tablet computer, wearable device, or any combination of these devices.

[0293] For ease of description, the above devices are described in terms of function, divided into various units. Of course, in implementing this specification, the functions of each unit can be implemented in one or more software and / or hardware components.

[0294] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0295] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0296] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0297] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0298] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0299] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0300] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic or disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0301] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0302] Those skilled in the art will understand that the embodiments of this specification can be provided as methods, systems, or computer program products. Therefore, this specification may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this specification may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0303] This specification can be described in the general context of computer-executable instructions that are executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a specific task or implement a specific abstract data type. This specification can also be practiced in distributed computing environments, where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0304] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0305] The above description is merely an embodiment of this specification and is not intended to limit this specification. Various modifications and variations can be made to this specification by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this specification should be included within the scope of the claims of this specification.

Claims

1. A multi-dimensional comprehensive evaluation method for vehicle ride comfort based on neural networks, characterized in that, include: S1. Obtain test data measured by various measuring devices pre-installed on the test vehicle, and obtain user characteristics of the driving evaluator on the test vehicle; S2. Based on the test data, and according to the preset evaluation method, determine the current six-dimensional evaluation index of the test vehicle. The six-dimensional evaluation index includes six primary dimensions, and each primary dimension includes six secondary indicators. The six primary dimensions are thermal comfort, handling stability, NVH performance, seat comfort, cabin comfort, and intelligent driving performance. S3. Input the six-dimensional evaluation index and the user features into the pre-trained prediction model to obtain the predicted scores of the six primary dimensions output by the prediction model. S4. Based on the test data, determine the test scenario for the test vehicle, and determine the weights of the six primary dimensions based on the test scenario; S5. Based on the predicted scores of the six primary dimensions and the weights of the six primary dimensions, determine the ride comfort index of the test vehicle, and based on the ride comfort index, determine the ride comfort evaluation result of the test vehicle. The method further includes steps S6 and S7; S6: Based on the driving comfort index of the test vehicle and the preset target driving comfort index of the test vehicle, the six-dimensional evaluation index of the test vehicle is iteratively updated. S7: For each secondary indicator in the six-dimensional evaluation index at the end of the iteration, determine the improvement range of the secondary indicator based on its value at the end of the iteration and its current value; Based on the improvement magnitude of the secondary indicator, determine the improvement ratio of the secondary indicator.

2. The method for multi-dimensional comprehensive evaluation of vehicle ride comfort based on neural networks as described in claim 1, characterized in that, S4 specifically includes: Based on the prior experience of experts, the subjective weights of the six primary dimensions are determined respectively; and based on the preset test dataset, the objective weights of the six primary dimensions are determined respectively. Based on the subjective and objective weights of the six primary dimensions, determine the combined weights of the six primary dimensions respectively; Determine the scenario weight adjustment matrix corresponding to five typical driving scenarios, and determine the test scenario of the test vehicle based on the test data. The test scenario is one of the five typical driving scenarios, namely, urban congestion scenario, highway cruising scenario, mountain road driving scenario, intelligent driving cruise scenario, and long-distance travel scenario. Based on the test scenario, the weight adjustment coefficients for the six primary dimensions are determined from the scenario weight adjustment matrix. The weights of the six primary dimensions are determined based on their respective combined weights and weight adjustment coefficients.

3. The method for multi-dimensional comprehensive evaluation of vehicle ride comfort based on neural networks as described in claim 2, characterized in that, Based on the prior experience of experts, the subjective weights of the six primary dimensions are determined, specifically including: Obtain the evaluation results of experts on the pairwise importance comparison of the six primary dimensions, and the evaluation results are expressed as ternary fuzzy numbers; Each ternary fuzzy number is aggregated and defuzzified sequentially to determine a sixth-order judgment matrix; Based on the sixth-order judgment matrix, the subjective weights of the six first-level dimensions are determined respectively.

4. The method for multi-dimensional comprehensive evaluation of vehicle ride comfort based on neural networks as described in claim 2, characterized in that, The pre-set test dataset includes six-dimensional evaluation indicators of test data collected from various types of vehicles in five typical driving scenarios, and six primary-dimensional subjective scores given by driving evaluators. Each type of vehicle is matched with one typical driving scenario as a sample. Based on a pre-set test dataset, determine the objective weights for each of the six primary dimensions, specifically including: Based on the preset test dataset, the subjective rating matrix of each sample in six primary dimensions is determined, and each subjective rating in the subjective rating matrix is ​​normalized to determine the normalized subjective rating matrix. For each primary dimension of each sample, calculate the proportion of the sample’s subjective rating on that primary dimension to the total subjective rating of that primary dimension. Based on the calculated proportions, calculate the information entropy of each first-level dimension; The information entropy of each first-level dimension is converted into a difference coefficient, and based on the difference coefficient of each first-level dimension, the objective weights of the six first-level dimensions are determined respectively.

5. The multi-dimensional comprehensive evaluation method for vehicle ride comfort based on neural networks as described in claim 1, characterized in that, In S6, based on the test vehicle's ride comfort index and the test vehicle's preset target ride comfort index, the six-dimensional evaluation indicators of the test vehicle are iteratively updated, specifically including: Based on the driving comfort index of the test vehicle and the preset target driving comfort index of the test vehicle, a loss value is determined, and the gradient of the input of the prediction model is determined based on the loss value. Based on the gradient of the input to the prediction model, the six-dimensional evaluation index of the test vehicle is updated, and the updated six-dimensional evaluation index of the test vehicle is continuously input into the prediction model to iteratively update the six-dimensional evaluation index of the test vehicle until a preset iteration termination condition is reached.

6. The multi-dimensional comprehensive evaluation method for vehicle ride comfort based on neural networks as described in claim 1, characterized in that, The secondary indicators under thermal comfort are temperature uniformity, thermal response speed, airflow comfort, radiant heat asymmetry, relative humidity suitability, and seat surface thermal comfort; the secondary indicators under handling stability are lateral acceleration comfort, longitudinal acceleration smoothness, roll control, pitch control, steering self-centering, and dynamic response; the secondary indicators under NVH performance are idling cabin noise, constant speed driving noise, acceleration noise quality, road noise, wind noise performance, and vibration comfort; the secondary indicators under seat comfort are body pressure distribution uniformity, hip pressure concentration, lumbar support, seat vibration isolation, material breathability, and long-distance travel fatigue; the secondary indicators under cabin comfort are spaciousness, field of vision, human-machine interaction convenience, cabin odor quality, ambient lighting comfort, and storage space convenience; the secondary indicators under intelligent driving performance are adaptive cruise smoothness, lane keeping stability, lane change smoothness, takeover transition naturalness, system status transparency, trust, and sense of security.

7. A multi-dimensional comprehensive evaluation device for vehicle ride comfort based on neural networks, characterized in that, include: The acquisition module is used to acquire test data measured by various measuring devices pre-installed on the test vehicle, and to acquire user characteristics of the driving evaluator on the test vehicle; The first determining module is used to determine the current six-dimensional evaluation index of the test vehicle based on the test data and according to a preset evaluation method. The six-dimensional evaluation index includes six primary dimensions, and each primary dimension includes six secondary indicators. The six primary dimensions are thermal comfort, handling stability, NVH performance, seat comfort, cabin comfort, and intelligent driving performance. The prediction module is used to input the six-dimensional evaluation indicators and the user features into a pre-trained prediction model to obtain the predicted scores of the six primary dimensions output by the prediction model. The second determining module is used to determine the test scenario of the test vehicle based on the test data, and to determine the weights of the six primary dimensions based on the test scenario. The third determining module is used to determine the ride comfort index of the test vehicle based on the predicted scores of the six primary dimensions and the weights of the six primary dimensions, and to determine the ride comfort evaluation result of the test vehicle based on the ride comfort index. The device also includes an update module and an improvement module; The update module is used to iteratively update the six-dimensional evaluation indicators of the test vehicle based on the test vehicle's ride comfort index and the test vehicle's preset target ride comfort index. The improvement module is used to determine the improvement magnitude of each secondary indicator in the six-dimensional evaluation index at the end of the iteration, based on the value of the secondary indicator at the end of the iteration and the current value; and to determine the improvement ratio of the secondary indicator based on the improvement magnitude of the secondary indicator.

8. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the method described in any one of claims 1 to 6.

9. An electronic device, characterized in that, The method includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method described in any one of claims 1 to 6.

Images