Somatosensory index determination device, method and equipment for intelligent driving system

Abstract

Disclosed are a somatosensory index determination device, method and equipment of an intelligent driving system, comprising at least one processor, which is configured to: determine a target feature vector corresponding to a current version of an intelligent driving system to be tested under multiple simulation scenarios; obtain at least one historical feature vector; determine the similarity between the current version of the intelligent driving system and at least one historical version of the intelligent driving system based on the target feature vector and the at least one historical feature vector; determine a candidate historical version of the intelligent driving system based on the similarity; and determine a predicted somatosensory index of the current version of the intelligent driving system based on the true value somatosensory index corresponding to the candidate historical version of the intelligent driving system and the corresponding similarity. The scheme obtains the somatosensory index prediction result of the current version of the intelligent driving system based on the true value somatosensory index corresponding to the historical version of the intelligent driving system similar to the somatosensory feature of the current version of the intelligent driving system and the corresponding similarity, thereby improving the somatosensory index evaluation efficiency.

Description

Technical Field

[0001] This disclosure relates to the field of autonomous driving simulation testing technology, specifically to a device, method, and equipment for determining the somatosensory indicators of an intelligent driving system. Background Technology

[0002] With the rapid development and widespread application of intelligent driving technology, the control precision and driving safety of vehicles have been significantly improved, providing solid technical support for intelligent mobility. The functions of autonomous driving systems are gradually evolving from basic functionality to a stage where the driving experience is acceptable. The occupant's sensory experience has become a core dimension for evaluating the overall performance of autonomous driving systems. The sensory experience directly determines occupant comfort, driving safety, and trust in the product, and is a key evaluation criterion for the design optimization, version verification, and large-scale mass production deployment of high-level autonomous driving systems.

[0003] Current autonomous driving evaluation systems have significant shortcomings. Related technologies rely on professional test drivers and passengers conducting real-world road tests to score driving and riding experiences. However, this approach is highly dependent on field testing or offline manual analysis, lacking automated performance indicators and algorithms that can be embedded in simulation evaluation processes. This makes it difficult to adapt to the needs of large-scale, automated simulation evaluations, resulting in low evaluation efficiency. Therefore, there is an urgent need for a method to determine performance indicators for intelligent driving systems to meet automation requirements. Summary of the Invention

[0004] To address the aforementioned technical issues, this disclosure provides a device, method, and equipment for determining the somatosensory indicators of an intelligent driving system, thereby achieving automation requirements and improving evaluation efficiency.

[0005] A first aspect of this disclosure provides a device for determining the somatosensory index of an intelligent driving system, comprising at least one processor configured to: determine a target feature vector corresponding to a current version of the intelligent driving system under test in multiple simulation scenarios; wherein the target feature vector is used to characterize the somatosensory quantitative features of the current version of the intelligent driving system in the multiple simulation scenarios; acquire at least one historical feature vector; the historical feature vector is a feature vector corresponding to at least one historical version of the intelligent driving system in the multiple simulation scenarios; determine the similarity between the current version of the intelligent driving system and the at least one historical version of the intelligent driving system based on the target feature vector and the at least one historical feature vector; determine candidate historical version intelligent driving systems in the at least one historical version of the intelligent driving system based on the similarity; and determine the predicted somatosensory index of the current version of the intelligent driving system based on the ground truth somatosensory index corresponding to the candidate historical version intelligent driving system and the corresponding similarity.

[0006] A second aspect of this disclosure provides a method for determining the somatosensory indicators of an intelligent driving system, comprising: Determine the target feature vectors of the current version of the intelligent driving system to be tested in multiple simulation scenarios; wherein, the target feature vectors are used to characterize the haptic quantification features of the current version of the intelligent driving system in the multiple simulation scenarios; Obtain at least one historical feature vector; the historical feature vector is the feature vector corresponding to at least one historical version of the intelligent driving system under the multiple simulation scenarios. Based on the target feature vector and the at least one historical feature vector, the similarity between the current version of the intelligent driving system and the at least one historical version of the intelligent driving system is determined. Based on the similarity, candidate historical version intelligent driving systems are determined in the at least one historical version intelligent driving system. Based on the true haptic index corresponding to the candidate historical version of the intelligent driving system and the corresponding similarity, the predicted haptic index of the current version of the intelligent driving system is determined.

[0007] A third aspect of this disclosure provides a device for determining the somatosensory indicators of an intelligent driving system, comprising: The first determining module is used to determine the target feature vector of the current version of the intelligent driving system under test in multiple simulation scenarios; wherein, the target feature vector is used to characterize the haptic quantification features of the current version of the intelligent driving system in the multiple simulation scenarios; The first acquisition module is used to acquire at least one historical feature vector; the historical feature vector is the feature vector corresponding to at least one historical version of the intelligent driving system under the multiple simulation scenarios. The second determining module is used to determine the similarity between the current version of the intelligent driving system and the at least one historical version of the intelligent driving system based on the target feature vector and the at least one historical feature vector. The third determining module is used to determine candidate historical version intelligent driving systems in the at least one historical version intelligent driving system based on the similarity. The fourth determining module is used to determine the predicted haptic index of the current version of the intelligent driving system based on the true haptic index corresponding to the candidate historical version of the intelligent driving system and the corresponding similarity.

[0008] A fourth aspect of this disclosure provides a computer program product that, when executed by an instruction processor, performs a method for determining the somatosensory indicators of an intelligent driving system as proposed in a second aspect of this disclosure.

[0009] A fifth aspect of this disclosure provides an electronic device comprising: a processor; a memory for storing processor-executable instructions; and a processor for reading executable instructions from the memory and executing the instructions to implement the method for determining the somatosensory indicators of the intelligent driving system provided in the second aspect of the disclosure.

[0010] A sixth aspect of this disclosure provides a computer-readable storage medium storing a computer program for executing the somatosensory index determination method for the intelligent driving system provided in the second aspect of the present disclosure.

[0011] The somatosensory index determination device for an intelligent driving system provided in this embodiment can determine the target feature vectors corresponding to the current version of the intelligent driving system under test in multiple simulation scenarios. Based on the target feature vectors and the historical feature vectors corresponding to at least one historical version of the intelligent driving system in the multiple simulation scenarios, the similarity between the current version of the intelligent driving system and at least one historical version of the intelligent driving system is determined. Therefore, by comparing the similarity between the current version of the intelligent driving system and at least one historical version of the intelligent driving system, candidate historical version intelligent driving systems with similar somatosensory features to the current version of the intelligent driving system are selected from at least one historical version of the intelligent driving system. Then, the predicted somatosensory index of the current version of the intelligent driving system is obtained by combining the collected ground truth somatosensory indexes corresponding to the candidate historical version intelligent driving systems and the corresponding similarity weighted average. This solution eliminates the need for large-scale manual real-road testing of the current version of the intelligent driving system. Instead, it uses simulated target feature vectors to automatically filter out historical versions of intelligent driving systems that have similar haptic characteristics to the current version. Based on the ground truth haptic indicators and corresponding similarities of the historical versions, it can quickly obtain the haptic indicator prediction results for the current version of the intelligent driving system. This improves the efficiency of haptic indicator evaluation for intelligent driving systems and effectively reduces the cost of haptic indicator evaluation. Furthermore, it enables the rapid completion of haptic indicator evaluation requirements for intelligent driving systems during the version development phase. Attached Figure Description

[0012] Figure 1 This is a schematic diagram of the structure of a somatosensory index determination device for an intelligent driving system provided in an exemplary embodiment of this disclosure.

[0013] Figure 2 This is a structural diagram of an electronic device provided in an exemplary embodiment of this disclosure.

[0014] Figure 3 This is a flowchart illustrating a method for determining the somatosensory indicators of an intelligent driving system provided in an exemplary embodiment of this disclosure.

[0015] Figure 4This is a flowchart illustrating a method for determining the somatosensory indicators of an intelligent driving system provided in another exemplary embodiment of this disclosure.

[0016] Figure 5 This is a schematic diagram of the structure of a somatosensory index determination device for an intelligent driving system provided in another exemplary embodiment of this disclosure. Detailed Implementation

[0017] To explain this disclosure, exemplary embodiments of the disclosure will now be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the disclosure, and not all of them. It should be understood that the disclosure is not limited to exemplary embodiments.

[0018] It should be noted that, unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps set forth in these embodiments do not limit the scope of this disclosure.

[0019] Application Overview

[0020] Traditional methods for evaluating the haptic feedback of intelligent driving systems rely heavily on real-vehicle field testing or offline manual analysis. They lack automated haptic feedback indicators and algorithms that can be embedded into the simulation evaluation process, making it difficult to adapt to the needs of large-scale, automated simulation evaluation.

[0021] In related technologies, the motion data collected by vehicle sensors is used as the basis for comfort evaluation based on simple statistical calculations using only a single kinematic signal such as acceleration or jerk. This approach, which uses a single kinematic characteristic parameter, is too one-sided and cannot fully cover the changes in driving sensation under complex driving conditions. It is also difficult to accurately reflect the real sensory experience brought about by typical control behaviors such as rapid acceleration, heavy braking, and intermittent braking.

[0022] To address the aforementioned technical issues, this disclosure provides a device for determining the somatosensory index of an intelligent driving system. The device includes at least one processor configured to: determine target feature vectors corresponding to the current version of the intelligent driving system under multiple simulation scenarios; acquire at least one historical feature vector; determine the similarity between the current version of the intelligent driving system and at least one historical version of the intelligent driving system based on the target feature vectors and the at least one historical feature vector; determine candidate historical version intelligent driving systems among the at least one historical version of the intelligent driving system based on the similarity; and determine the predicted somatosensory index of the current version of the intelligent driving system based on the ground truth somatosensory index and the corresponding similarity of the candidate historical version intelligent driving systems. This solution eliminates the need for large-scale manual real-road testing of the current version of the intelligent driving system. It automatically filters out historical version intelligent driving systems with similar somatosensory features using simulated target feature vectors, and then quickly obtains the predicted somatosensory index of the current version of the intelligent driving system based on the ground truth somatosensory index and the corresponding similarity of the historical version of the intelligent driving system. This improves the efficiency of somatosensory index evaluation for intelligent driving systems and effectively reduces the cost of somatosensory index evaluation. Furthermore, it enables rapid completion of somatosensory index evaluation requirements for intelligent driving systems during the version development phase.

[0023] Exemplary device

[0024] Figure 1 This is a schematic diagram of the structure of a motion perception index determination device for an intelligent driving system provided in an exemplary embodiment of this disclosure. Figure 1 As shown, the somatosensory index determination device 100 includes at least one processor 101, which is configured to: Determine the target feature vectors of the current version of the intelligent driving system under test in multiple simulation scenarios.

[0025] Among them, the target feature vector is used to characterize the haptic quantification features of the current version of the intelligent driving system in multiple simulation scenarios.

[0026] In some embodiments, the aforementioned simulation scenarios refer to preset test scenarios used for simulation testing of intelligent driving systems. Multiple simulation scenarios can cover different road types, different traffic flow densities, or different driving conditions, etc., to reproduce various longitudinal or lateral acceleration and deceleration scenarios that intelligent driving systems may encounter in daily driving. For example, multiple simulation scenarios may include lane changing scenarios, cruise scenarios, and following vehicle acceleration and deceleration scenarios.

[0027] In some embodiments, ergonomics refers to the subjective feeling of occupant comfort caused by the vehicle's dynamic behavior during the execution of longitudinal and / or lateral control by the intelligent driving system. Ergonomic metrics are used to quantify the subjective feeling of occupant comfort.

[0028] In some embodiments, since the current version of the intelligent driving system may exhibit two different types of dynamic behaviors during control: longitudinal events (e.g., heavy braking, intermittent braking, rapid acceleration) and lateral events (e.g., sharp turns, frequent lane changes, lateral swaying), the target feature vector can be used to characterize the longitudinal haptic quantification features of the current version of the intelligent driving system in multiple simulation scenarios, or the target feature vector can be used to characterize the lateral haptic quantification features of the current version of the intelligent driving system in multiple simulation scenarios. It should be noted that the target feature vectors characterizing the longitudinal or lateral haptic quantification features can be calculated using different types of parameters.

[0029] Obtain at least one historical feature vector; the historical feature vector is the feature vector corresponding to at least one historical version of the intelligent driving system in multiple simulation scenarios.

[0030] In some embodiments, at least one historical feature vector is obtained by conducting simulation tests on at least one historical version of the intelligent driving system using the same or different simulation evaluation platform, simulation scenario library, and test parameters (such as sampling frequency and simulation step size) as the current version of the intelligent driving system. This at least one historical feature vector can be stored in the database of the simulation evaluation platform. When obtaining historical feature vectors, one can directly retrieve the historical feature vectors corresponding to all historical versions of the intelligent driving system from the database, or, based on information such as the compatible vehicle models of the current version of the intelligent driving system, filter out a portion of the historical versions of the intelligent driving system within the same series, and then retrieve the historical feature vectors corresponding to those historical versions from the database.

[0031] It should be noted that when the target feature vector is used to characterize the longitudinal haptic quantification features of the current version of the intelligent driving system in multiple simulation scenarios, historical feature vectors are obtained to characterize the longitudinal haptic quantification features of the previous version of the intelligent driving system in multiple simulation scenarios; when the target feature vector is used to characterize the lateral haptic quantification features of the current version of the intelligent driving system in multiple simulation scenarios, historical feature vectors are obtained to characterize the lateral haptic quantification features of the previous version of the intelligent driving system in multiple simulation scenarios.

[0032] Based on the target feature vector and at least one historical feature vector, determine the similarity between the current version of the intelligent driving system and at least one historical version of the intelligent driving system.

[0033] In some embodiments, similarity is used to characterize the geometric similarity between the current version of the intelligent driving system and the historical version of the intelligent driving system in a somatosensory feature space constructed from multiple simulation scenarios.

[0034] In some embodiments, a preset similarity algorithm can be used to calculate the matching value between the target feature vector and each historical feature vector, and this matching value is determined as the similarity between the current version of the intelligent driving system and the historical version of the intelligent driving system. The similarity algorithm may include cosine similarity algorithm, Euclidean distance algorithm, Manhattan distance algorithm, etc. When the similarity algorithm is an Euclidean distance algorithm, Manhattan distance algorithm, etc., the smaller the calculated distance, the higher the similarity of the somatosensory features between the current version of the intelligent driving system and the corresponding historical version of the intelligent driving system.

[0035] For example, let's take the cosine similarity algorithm as an example. Using the cosine similarity algorithm, the cosine value between the target feature vector and each historical feature vector is calculated. This cosine value is then used to determine the similarity between the current version of the intelligent driving system and the corresponding historical version. The closer the cosine value is to 1, the more consistent the directions of the two vectors are, and the higher the similarity of the sonic features between the two versions of the intelligent driving system.

[0036] Based on similarity, candidate historical version intelligent driving systems are identified from at least one historical version of intelligent driving systems.

[0037] In some embodiments, a similarity threshold can be preset, and historical versions of intelligent driving systems with similarity greater than the threshold can be identified as candidate historical versions of intelligent driving systems. Alternatively, all historical versions of intelligent driving systems can be sorted from largest to smallest similarity, and the top N historical versions of intelligent driving systems in the sorted list can be selected as candidate historical versions of intelligent driving systems. Or, based on a preset similarity threshold, a subset of historical versions of intelligent driving systems with similarity greater than the threshold can be selected from at least one historical version of intelligent driving systems, and then sorted from largest to smallest similarity, and the top N historical versions of intelligent driving systems in the sorted list can be selected as candidate historical versions of intelligent driving systems. Here, N is a preset positive integer, for example, N equals 3.

[0038] Based on the true haptic metrics corresponding to candidate historical versions of intelligent driving systems and their corresponding similarities, the predicted haptic metrics for the current version of intelligent driving systems are determined.

[0039] In some embodiments, the true haptic index corresponding to the aforementioned candidate historical version of the intelligent driving system is a ground truth value measured manually. This true haptic index is obtained by real occupants through comprehensive scoring based on the haptic comfort across all scenarios during real road tests. It should be noted that the test scenarios covered by the true haptic index are consistent with multiple simulation scenarios of the current version of the intelligent driving system.

[0040] In some embodiments, when the target feature vector is used to characterize the longitudinal haptic quantification features of the current version of the intelligent driving system in multiple simulation scenarios, the true haptic index is actually the ground true value corresponding to the longitudinal haptic comfort, and correspondingly, the predicted haptic index is the predicted value corresponding to the longitudinal haptic comfort; when the target feature vector is used to characterize the lateral haptic quantification features of the current version of the intelligent driving system in multiple simulation scenarios, the true haptic index is actually the ground true value corresponding to the lateral haptic comfort, and correspondingly, the predicted haptic index is the predicted value corresponding to the lateral haptic comfort.

[0041] For example, the true-value ergonomic metric refers to the average number of kilometers driven by the intelligent driving system before a control event causing discomfort to the occupants is actually triggered; the predicted ergonomic metric refers to the average number of kilometers driven by the intelligent driving system before a control event causing discomfort to the occupants is expected to be triggered. A higher true-value or predicted ergonomic metric indicates better ergonomic performance of the intelligent driving system. For instance, taking the control event as a longitudinal event, if the true-value ergonomic metric is 240 km / event, it means that the corresponding intelligent driving system experiences one heavy braking event or intermittent braking event causing discomfort on average every 240 kilometers.

[0042] In some embodiments, the predicted haptic index of the current version of the intelligent driving system can be obtained by weighting the ground truth haptic index corresponding to the candidate historical version of the intelligent driving system and the corresponding similarity.

[0043] For example, if K candidate historical versions of intelligent driving systems are identified, the predicted haptic index of the current version of the intelligent driving system can be calculated according to the following formula (1), based on the true haptic index and the corresponding similarity of the K candidate historical versions of intelligent driving systems: = (1) in, This indicates the predicted haptic feedback metrics of the current version of the intelligent driving system. This represents the true-value haptic index corresponding to the j-th historical version of the intelligent driving system. This represents the similarity between the j-th historical version of the intelligent driving system and the current version of the intelligent driving system.

[0044] For example, selecting the top-3 candidate historical versions of intelligent driving systems, the similarity S between historical version 1 and the current version is... i1 =0.9, true mean squared motion index MPJ1 = 200km / time; similarity S between historical version 2 and the current version i2 =0.8, true mean squared motion index MPJ2 = 250km / time; similarity S between historical version 3 and the current version. i3=0.7, true value of the driving experience index MPJ3 = 300km / time, substituting into the above formula (1), the predicted driving experience index of the current version of the intelligent driving system is calculated as follows: =

[0045] The somatosensory index determination device for an intelligent driving system provided in this embodiment can determine the target feature vectors corresponding to the current version of the intelligent driving system under test in multiple simulation scenarios. Based on the target feature vectors and the historical feature vectors corresponding to at least one historical version of the intelligent driving system in the multiple simulation scenarios, the similarity between the current version of the intelligent driving system and at least one historical version of the intelligent driving system is determined. Therefore, by comparing the similarity between the current version of the intelligent driving system and at least one historical version of the intelligent driving system, candidate historical version intelligent driving systems with similar somatosensory features to the current version of the intelligent driving system are selected from at least one historical version of the intelligent driving system. Then, the predicted somatosensory index of the current version of the intelligent driving system is obtained by combining the collected ground truth somatosensory indexes corresponding to the candidate historical version intelligent driving systems and the corresponding similarity weighted average. This solution can automatically and quickly obtain the prediction results of the current version of the intelligent driving system's motion index without conducting large-scale manual real-road tests. This not only effectively reduces the time, manpower, and vehicle costs required for real-road testing in the process of determining motion indexes, but also improves the efficiency of motion index evaluation for intelligent driving systems. Furthermore, it can quickly complete the evaluation requirements of the motion indexes of intelligent driving systems during the version development stage, which also effectively reduces the cost of motion index evaluation for intelligent driving systems.

[0046] In some embodiments of this disclosure, such as Figure 1 As shown, processor 101 executes the process of determining the target feature vectors of the current version of the intelligent driving system under test in multiple simulation scenarios, specifically configured as follows: Obtain the correction failure rate of the current version of the intelligent driving system in the simulation scenario.

[0047] In some embodiments, the aforementioned correction failure rate characterizes the probability that the current version of the intelligent driving system will actually cause sensory discomfort in various simulation scenarios. For a detailed explanation of the specific implementation method for determining the correction failure rate, please refer to the following embodiments; the specific implementation method will not be repeated here.

[0048] In some embodiments, the aforementioned correction failure rate can be stored in the database of the simulation evaluation platform, and can be directly retrieved from the database when needed. Moreover, each simulation scenario corresponds to its own correction failure rate, so a correspondence between simulation scenario identifiers and correction failure rates can be established in advance; when it is necessary to obtain the correction failure rate corresponding to a certain simulation scenario, the correction failure rate corresponding to the simulation scenario indicated by the simulation scenario identifier can be determined based on the simulation scenario identifier of that simulation scenario and the correspondence.

[0049] The target feature vector is determined based on multiple correction failure rates.

[0050] In some embodiments, the correction failure rates corresponding to multiple simulation scenarios are used as vector elements, and the multiple correction failure rates are arranged according to a preset scenario sorting rule to generate a target feature vector that represents the overall tactile quantitative characteristics of the current version of the intelligent driving system.

[0051] For example, there may be K simulation scenarios, and the correction failure rate corresponding to each of the K simulation scenarios is denoted as . , , ..., The target feature vector is denoted as F1, and F1 can be represented as follows: F1 = [ , , ..., ] For example, with K=4, four simulation scenarios including highway cruising, urban congestion, ramp merging, and emergency braking, and the preset scenario ordering rule being [highway cruising, urban congestion, ramp merging, emergency braking], let's assume highway cruising: =0.12, Urban congestion: =0.27, merging from the ramp: =0.05, emergency braking: =0.31. According to the scene sorting rules, after arranging and combining the above four correction failure rates, the target feature vector F1 is obtained as follows: F1= [0.12, 0.27, 0.05, 0.31] Based on the above embodiments, since the target feature vector can be determined based on the correction failure rate of the current version of the intelligent driving system in each simulation scenario, it is possible to quantify the comprehensive tactile performance of the current version of the intelligent driving system in multiple simulation scenarios and accurately characterize the comprehensive tactile features of the current version of the intelligent driving system. This can meet the automated operation requirements of large-scale, multi-round intelligent driving system version simulation evaluation and further improve the evaluation efficiency.

[0052] In some embodiments of this disclosure, such as Figure 1As shown, processor 101 executes the process of obtaining the correction failure rate of the current version of the intelligent driving system in the simulation scenario, specifically configured as follows: Obtain the total number of simulation samples, the number of failed motion simulation samples, and the number of failed simulation samples for the current version of the intelligent driving system in the simulation scenario.

[0053] In some embodiments, the total number of simulation samples mentioned above refers to the total number of samples for which the current version of the intelligent driving system has completed all simulation tests in the corresponding simulation scenario; the number of failed haptic simulation samples refers to the number of simulation samples in which the current version of the intelligent driving system actually causes occupant discomfort in the simulation scenario, provided that the simulation runs normally; the number of simulation failure samples refers to the number of samples in the simulation scenario where the simulation test terminates abnormally due to algorithm logic errors or abnormal simulation operation. It should be noted that the number of failed haptic simulation samples and the number of simulation failure samples do not overlap.

[0054] For example, after each round of simulation test is completed, the simulation evaluation platform will automatically count and store the number of the above three types of samples for each tested version of the intelligent driving system in each simulation scenario. Therefore, the total number of simulation samples, the number of haptic simulation failure samples, and the number of simulation failure samples corresponding to each simulation scenario can be obtained directly from the database of the simulation evaluation platform.

[0055] The correction failure rate is determined based on the total number of simulation samples, the number of failed motion simulation samples, and the number of failed simulation samples.

[0056] In some embodiments, the motion simulation failure rate can be determined first based on the total number of simulation samples, the number of failed motion simulation samples, and the number of simulation failure samples. Then, the motion simulation failure rate is corrected to obtain the corrected failure rate. For details, please refer to the detailed description in the following embodiments; the specific embodiments disclosed herein will not be repeated here.

[0057] It should be noted that the above steps are illustrated by way of the correction failure rate for a single simulation scenario. For calculating the correction failure rate for other simulation scenarios, the steps in the above embodiments can be referred to.

[0058] Based on the above embodiments, since the correction failure rate can be determined based on the total number of simulation samples, the number of failed haptic simulation samples, and the number of simulation failure samples, test interference caused by non-haptic reasons is eliminated. This allows the correction failure rate to more accurately reflect the probability that the current version of the intelligent driving system will actually cause haptic discomfort in the corresponding simulation scenario, improving the accuracy of feature quantification and thus ensuring the accuracy of subsequent predictions of haptic indicators. Moreover, all statistical data can be directly obtained from the simulation evaluation platform without manual statistics and data entry, further improving the automation level of the entire process and saving labor costs.

[0059] In some embodiments of this disclosure, a longitudinal event is used as an example for illustration. When the number of failed motion simulation samples is equal to the number of simulation samples that actually cause longitudinal motion discomfort to occupants in the current version of the intelligent driving system under that simulation scenario, such as... Figure 1 As shown, processor 101 executes the function of obtaining the number of failed motion simulation samples in the current version of the intelligent driving system in the simulation scenario, specifically configured as follows: Determine the target haptic score corresponding to a longitudinal event occurring in the simulation scenario for the current version of the intelligent driving system.

[0060] In some embodiments, the aforementioned target ergonomic score refers to a quantitative score of the degree of ergonomic comfort brought to the occupant by the corresponding intelligent driving operation during the occurrence of a longitudinal event; wherein, a higher score corresponds to a stronger degree of ergonomic comfort, and vice versa. For the specific implementation method of determining the target ergonomic score, please refer to the detailed description in the following embodiments, which will not be repeated here.

[0061] In some embodiments, the target haptic score can be stored in the database of the simulation evaluation platform, and can be directly retrieved from the database when needed. Furthermore, the same simulation scene may contain multiple types of vertical events, each with its own target haptic score. Therefore, a correspondence between vertical event identifiers and target haptic scores can be pre-established. When it is necessary to obtain the target haptic score corresponding to a certain type of vertical event, the target haptic score corresponding to the vertical event indicated by that vertical event identifier can be determined based on the vertical event identifier of that type of vertical event and the correspondence. For example, if both a light braking event and a heavy braking event occur simultaneously, the target haptic score corresponding to the light braking event and the target haptic score corresponding to the heavy braking event are calculated separately.

[0062] Based on the target haptic score and the preset haptic threshold, the number of haptic failure scenarios in the current version of the intelligent driving system is determined.

[0063] In some embodiments, after obtaining the target haptic scores of all longitudinal events in the same simulation scenario, the target haptic scores of all longitudinal events can be compared with a preset haptic threshold one by one. If the target haptic score is lower than or equal to the preset haptic threshold, it indicates that the haptic comfort corresponding to the longitudinal event does not meet the requirements, and the simulation scenario sample corresponding to the longitudinal event is a haptic failure scenario sample. By counting the haptic failure scenario samples, the number of haptic failure scenarios of the current version of the intelligent driving system in the simulation scenario can be obtained. If the target haptic score is greater than the preset haptic threshold, it indicates that the haptic comfort corresponding to the longitudinal event meets the requirements, and the simulation scenario sample corresponding to the longitudinal event does not belong to the haptic failure scenario sample and is not included in the statistical range of the number of haptic failure scenarios.

[0064] It should be noted that the implementation method for obtaining the number of failed motion simulation samples in each simulation scenario can refer to the above embodiments. For different longitudinal events, the corresponding preset motion thresholds may be the same or different; for different simulation scenarios, the corresponding preset motion thresholds may be the same or different, and the embodiments of this disclosure do not limit this.

[0065] In other embodiments, if the target haptic score is lower than or equal to a preset haptic threshold, it indicates that the haptic comfort level corresponding to the longitudinal event does not meet the requirements, i.e., the haptic comfort level corresponding to the longitudinal event is low comfort. When the target haptic score is greater than the preset haptic threshold, the haptic comfort level corresponding to the longitudinal event can be further divided. Specifically, if the target haptic score falls within a first threshold range, the haptic comfort level corresponding to the longitudinal event is medium comfort; if the target haptic score falls within a second threshold range, the haptic comfort level corresponding to the longitudinal event is high comfort.

[0066] For example, assuming a preset somatosensory threshold of 0.5, the first threshold range is (0.5, 0.8], and the second threshold range is (0.8, ...). If the target comfort score is lower than or equal to 0.5, it indicates that the comfort level corresponding to the longitudinal event does not meet the requirements, meaning the comfort level for that longitudinal event is low. If the target comfort score is higher than 0.5, the comfort level corresponding to the longitudinal event can be further determined by comparing the target comfort score with a first threshold range and a second threshold range. Thus, the comfort level (denoted as Comfort Level) corresponding to the target comfort score S can be represented as follows:

[0067] It should be noted that when the number of failed motion simulation samples is equal to the number of simulation samples in which the current version of the intelligent driving system actually causes longitudinal motion discomfort to the occupants in that simulation scenario, the target motion score corresponding to the longitudinal event occurring in the simulation scenario can be calculated first, as described in the above embodiment. Then, based on the target motion score and a preset motion threshold, the number of failed motion simulation samples in that simulation scenario can be determined. Similarly, when the number of failed motion simulation samples is equal to the number of simulation samples in which the current version of the intelligent driving system actually causes lateral motion discomfort to the occupants in that simulation scenario, the statistical logic of the longitudinal motion failure scenario in the aforementioned embodiment can be referred to. The determination can be made based on the motion score corresponding to the identified lateral event and a preset lateral motion threshold. Alternatively, the implementation method in related technologies can be referred to, which will not be elaborated upon in this disclosure.

[0068] Based on the above embodiments, the target haptic score corresponding to a longitudinal event occurring in the simulation scenario for the current version of the intelligent driving system can be determined. Based on the target haptic score and a preset haptic threshold, the number of haptic failure scenarios for the current version of the intelligent driving system can be determined. This solution can automatically identify haptic simulation failure samples that do not meet the haptic comfort requirements in the simulation scenario through the matching relationship between the target haptic score and the preset haptic threshold, without the need for manual intervention. Furthermore, this solution can adapt to the automated identification needs of batch simulation testing, thereby improving the accuracy and efficiency of haptic simulation failure sample statistics and providing a reliable data foundation for subsequent calculation of the failure rate.

[0069] In some embodiments of this disclosure, such as Figure 1 As shown, processor 101 executes the process of determining the target haptic score corresponding to a longitudinal event occurring in the simulation scenario within the current version of the intelligent driving system, specifically configured as follows: Obtain the target simulation data corresponding to the current version of the intelligent driving system in the simulation scenario.

[0070] In some embodiments, the target simulation data mentioned above refers to time-series smoothed data obtained by filtering the initial simulation data collected during the operation of the current version of the intelligent driving system in various simulation scenarios; wherein, the target simulation data includes key motion parameters such as acceleration and vehicle speed during vehicle operation. The implementation method for filtering the initial simulation data to obtain the target simulation data can be referred to the detailed description in the following embodiments, which will not be repeated here.

[0071] In some embodiments, the target simulation data can be stored in the database of the simulation evaluation platform, and can be directly retrieved from the database when needed. Furthermore, each simulation scenario corresponds to its own target simulation data, so a correspondence between simulation scenario identifiers and target simulation data can be pre-established; when it is necessary to obtain the target simulation data corresponding to a certain simulation scenario, the target simulation data corresponding to the simulation scenario indicated by the simulation scenario identifier can be determined based on the simulation scenario identifier of that simulation scenario and the correspondence.

[0072] Based on the acceleration time series data in the target simulation data, determine the jerk time series data.

[0073] In some embodiments, jerk is an indicator characterizing the rate of change of acceleration, and jerk is typically expressed as Jerk, with units of m / s². 3 The difference Δt = t between adjacent time stamps can be calculated based on the timestamps of acceleration time series data. i+1 -t iThen, the central difference method is used to calculate the jerk based on the acceleration time series data and the difference calculation step size. Of course, the difference calculation step size can be Δt, 2Δt, or 3Δt, and this embodiment of the disclosure does not limit it.

[0074] For example, using 2Δt as the differential calculation step size, as shown in the following formula (2), the central difference method is used to calculate the jerk time series data based on the acceleration time series data and the differential calculation step size.

[0075] (2) Where j(t) i ) represents t i The corresponding accelerometer at each moment, (t i+1 ) represents t i+1 acceleration at any moment (t i-1 ) represents t i-1 The acceleration at time t, where Δt is the time interval between adjacent timestamps.

[0076] Based on target simulation data, jerk timing data, and preset event judgment thresholds, longitudinal events within the simulation scene are identified.

[0077] In some embodiments, a longitudinal event refers to a control behavior event that triggers occupant discomfort due to dynamic performance during longitudinal control by the intelligent driving system, and is identifiable. Longitudinal events can include different types of control behavior events; for example, longitudinal events can include heavy braking events, intermittent braking events, and sudden acceleration from a standstill, which can cause changes in occupant comfort.

[0078] In some embodiments, the preset event determination threshold may include multiple sets of event determination thresholds, each set corresponding to different types of longitudinal events. For example, when the longitudinal event is a heavy braking event, a first event determination threshold corresponding to the heavy braking event can be determined from the preset event determination thresholds; when the longitudinal event is a point braking event, a second event determination threshold corresponding to the point braking event can be determined from the preset event determination thresholds. Each set of event determination thresholds is predetermined based on occupant perception calibration data from real-vehicle road tests, and can also be adaptively adjusted according to vehicle characteristics and differences in occupant groups.

[0079] For example, for a single simulation sample including target simulation data and corresponding jerk timing data, the target simulation data and corresponding jerk timing data can be traversed along the timing dimension, and matched with preset event judgment thresholds. If the target simulation data and jerk timing data simultaneously satisfy a set of event judgment thresholds corresponding to a certain type of longitudinal event in the preset event judgment thresholds, then the longitudinal event of the type corresponding to that set of event judgment thresholds can be determined as a longitudinal event occurring in the simulation scenario. After completing the identification of all types of longitudinal events in the simulation scenario, all longitudinal events identified in the current simulation scenario can be obtained. For example, if the target simulation data and corresponding jerk timing data satisfy the first event judgment threshold corresponding to a heavy braking event, then a heavy braking event can be determined to have occurred; if the target simulation data and jerk timing data satisfy the second event judgment threshold corresponding to a point braking event, then a point braking event can be determined to have occurred.

[0080] For example, consider the detection of a hard braking event. The first event determination threshold may include a first acceleration threshold a. th1 Second acceleration threshold a th2 First jerk threshold j th1 Second accelerometer threshold j th2 Threshold T for the shortest duration of heavy braking min and minimum vehicle speed threshold v min When the target simulation data and the corresponding jerk timing data satisfy the following combination of conditions (1) or (2), it is determined to be a re-braking event: Conditional combination (1): a(t) th1 And |j(t)|>j th1 And t e -t s >T min And v(t)>v min .

[0081] Conditional combination (2): |j(t)|>j th2 And a(t+Δt) th2 And t e -t s >T min And v(t)>v min .

[0082] Where a(t) represents the acceleration at any time t, |j(t)| represents the absolute value of the acceleration at any time t, a(t+Δt) represents the acceleration at time t+Δt, v(t) represents the velocity at any time t, and the duration refers to the continuous duration of the combination of the above conditions. s t represents the start time of the heavy braking event. e ​​t represents the end time of the heavy braking event. e -t s Indicates the duration of the hard braking event.

[0083] For example, consider the detection of intermittent braking events. The second event determination threshold may include the lower limit of intermittent braking acceleration, a. min_brake Upper limit of braking acceleration a max_brake Threshold T for the shortest duration of intermittent braking min The longest duration of intermittent braking, T max and the minimum acceleration threshold j of point braking min When the target simulation data and the corresponding jerk timing data meet the following combination of conditions (3), it is determined to be a point braking event: a min_brake < a(t) < a max_brake And T min <t e -t s < T max And |j(t)| > j min (3) Among them, t s t represents the start time of the braking event. e t represents the end time of the braking event. e -t s The duration of the braking event is represented by . a(t) represents the acceleration at any time t, and |j(t)| represents the absolute value of the acceleration at any time t.

[0084] Obtain the multidimensional kinematic feature vector of the longitudinal event.

[0085] In some embodiments, for each identified longitudinal event, a multidimensional kinematic feature vector of that longitudinal event is obtained. The aforementioned multidimensional kinematic feature vector can be a two-dimensional kinematic feature vector, a three-dimensional kinematic feature vector, or a four-dimensional kinematic feature vector; this disclosure does not limit the specific type of kinematic feature vector.

[0086] For example, taking a four-dimensional kinematic feature vector as an example, the multi-dimensional kinematic feature vector can include the absolute value of peak jerk, the absolute value of average jerk, the duration of the event, and the peak acceleration within the duration of the longitudinal event. First, based on the target simulation data and jerk time-series data, the absolute value of peak jerk, the absolute value of average jerk, the duration of the event, and the peak acceleration can be obtained. Then, by arranging these features in a preset order, the multi-dimensional kinematic feature vector corresponding to the longitudinal event can be obtained. For all longitudinal events identified in the current simulation scene, the multi-dimensional kinematic feature vector corresponding to each longitudinal event is obtained.

[0087] Based on preset feature weights and multidimensional kinematic feature vectors, the target somatosensory score corresponding to the longitudinal event is determined by a weighted somatosensory function.

[0088] In some embodiments, the preset feature weights include multiple weight coefficients, each corresponding to a different dimension of a feature in a multidimensional kinematic feature vector. The preset feature weights are obtained through regression training on a large number of haptic scoring samples collected during real-vehicle road tests. The weighted haptic function is a linear weighted function composed of the preset feature weights and the multidimensional kinematic feature vector, used to calculate the target haptic score corresponding to a longitudinal event.

[0089] In some embodiments, the comfort sub-score corresponding to each dimension feature in the multi-dimensional kinematic feature vector can be calculated first, and then the comfort sub-scores can be weighted and summed based on preset feature weights to obtain the target somatosensory score corresponding to the longitudinal event. Specifically, the comfort sub-score corresponding to each dimension feature can be calculated based on the value of the dimension feature and the corresponding preset sub-score mapping rule.

[0090] For example, let's take a four-dimensional kinematic feature vector as an example, where the multi-dimensional kinematic feature vector is used instead of a four-dimensional kinematic feature vector. The four-dimensional kinematic feature vector is denoted as F2 = [a min , T, j avg , j max1 The weighted somatosensory function is shown in the following formula (3). Based on the four-dimensional kinematic feature vector, the target somatosensory score corresponding to the longitudinal event is calculated using the following formula (3).

[0091] (3)

[0092]

[0093]

[0094]

[0095] Among them, f a f represents the peak acceleration comfort sub-score. T The duration comfort score is represented by f. p f represents the peak acceleration comfort sub-score. j w represents the average jerk comfort sub-score. a w T w P w j These are the weight coefficients corresponding to the four features, and S is the final target haptic score for this longitudinal event.

[0096] Where T represents the duration of the longitudinal event, which is equal to the end time t of the longitudinal event. e With the start time t of the longitudinal event s Time difference; It refers to the absolute value of the peak accelerometer during the duration of a longitudinal event. It refers to the minimum absolute value of acceleration within the duration of a longitudinal event. This refers to the absolute value of the average jerk over the duration of a longitudinal event; a max T represents the preset acceleration threshold. max j represents the preset duration threshold. max2 This indicates the preset jerk threshold.

[0097] It should be noted that for all vertical events identified in the current simulation scene, the steps in the above embodiments can be directly used to obtain the target somatosensory score corresponding to each vertical event.

[0098] Based on the above embodiments, since longitudinal events occurring in the vehicle under the corresponding simulation scenario are accurately identified based on target simulation data, accelerometer timing data, and preset event judgment thresholds, and the target ergonomic score is calculated by weighting the multi-dimensional kinematic feature vector of the longitudinal event with preset feature weights, the impact of longitudinal events on occupant ergonomics can be accurately quantified. This avoids the error of relying solely on single acceleration data to determine ergonomics in traditional methods, thereby improving the accuracy of subsequent ergonomic simulation failure sample judgment and ensuring the reliability of the target feature vector. This lays a good foundation for subsequent accurate calculation of version similarity and determination of predicted ergonomic indicators.

[0099] In some embodiments of this disclosure, such as Figure 1 As shown, processor 101 executes the process of acquiring the target simulation data corresponding to the current version of the intelligent driving system in the simulation scenario, specifically configured as follows: Obtain the initial simulation data of the current version of the intelligent driving system in the simulation scenario.

[0100] For example, the aforementioned initial simulation data refers to the original time-series dataset of vehicle control, natively output by the simulation platform and without filtering or noise reduction processing, within the complete operating cycle of a single simulation scenario for the current version of the intelligent driving system. This initial simulation data may include acceleration time-series data, speed time-series data, and timestamps. After selecting the current version of the intelligent driving system to be tested, the set of simulation scenarios required for this evaluation (such as highway cruising, urban congestion, ramp merging, emergency braking, etc.) can be selected from the platform scenario library of the intelligent driving simulation evaluation platform. For each simulation scenario, the current version of the intelligent driving system is run on the simulation evaluation platform to perform the simulation task, and the initial simulation data of each simulation scenario is collected in real time from the simulation evaluation platform.

[0101] The initial simulation data is filtered to obtain the target simulation data.

[0102] In some embodiments, since the initial simulation data output by the simulation platform typically contains high-frequency noise, numerical jitter, or abnormal data caused by solver instability, directly using the initial simulation data for processing would lead to severely distorted results. Therefore, the initial simulation data is filtered to obtain the target simulation data. This filters out high-frequency abnormal fluctuations introduced by system noise and data acquisition errors during the simulation process, resulting in stable target simulation data. The filtering process may include Gaussian filtering, median filtering, or other suitable filtering methods, which are not limited in this embodiment.

[0103] In other embodiments, when filtering the initial simulation data, all data included in the initial simulation data can be filtered, or only a portion of the initial simulation data (e.g., acceleration timing data) can be filtered. This disclosure does not limit this, and the specific method is determined according to the actual usage.

[0104] For example, median filtering is used as an example of filtering processing. Median filtering is performed on the acceleration time series data in the initial simulation data. Specifically, a fixed-length sliding window can be set and slides point by point on the acceleration time series data. Each time, all acceleration values ​​within the window are extracted, sorted, and the median is taken as the filtered acceleration value at the center point of the current window. After traversing the entire time series data, smooth acceleration data with high-frequency jitter removed can be obtained.

[0105] For example, at the current time t i Centered on a point, a sliding window is formed by taking k acceleration data points before and after the point (a total of 2k+1 acceleration data points). As shown in formula (4) below, all acceleration values ​​within the window are sorted, and the median of the sorted values ​​is taken as the filtered t. i By sliding the acceleration value corresponding to each moment point by point to complete the smoothing process of the entire acceleration time series data, the target acceleration time series data after removing high-frequency jitter can be obtained.

[0106] (4) It should be noted that the above steps can be performed for each simulation scenario to obtain the target simulation data for each scenario.

[0107] Based on the above embodiments, by filtering the initial simulation data to remove noise interference, the accuracy of the basic data for subsequent somatosensory score calculation can be guaranteed, and abnormal fluctuations in the initial simulation data can be avoided from interfering with the accuracy of longitudinal event recognition and somatosensory score, thereby improving the reliability of the somatosensory score results.

[0108] In some embodiments of this disclosure, such as Figure 1 As shown, processor 101 executes a process to determine the failure rate based on the total number of simulation samples, the number of failed motion simulation samples, and the number of simulation failure samples. Specifically, it is configured as follows: The motion simulation failure rate is determined based on the total number of simulation samples, the number of failed motion simulation samples, and the number of simulation failure samples.

[0109] For example, the number of valid simulation samples can be obtained based on the total number of simulation samples and the number of simulation failure samples, so as to eliminate invalid simulation failure samples; then, the haptic failure rate in the valid sample space can be obtained based on the ratio of the number of haptic simulation failure samples to the number of valid simulation samples. As shown in the following formula (5): (5) in, Indicates the failure rate of sensory perception. This indicates the number of failed motion simulation samples. This indicates the number of simulation failure samples. This indicates the total number of simulation samples.

[0110] For example, suppose the current version of the intelligent driving system performs a simulation test task in a certain simulation scenario on the simulation platform. The simulation test task contains a total of 1,000 simulation samples. Among them, 100 simulation samples fail due to simulation anomalies such as algorithm bugs. Among the remaining 900 valid simulation samples, there are 50 simulation samples with failed body perception. After substituting into the above formula, we can obtain the body perception failure rate corresponding to this simulation scenario as 50 / (1000-100)≈5.56%.

[0111] The correction failure rate is determined based on the total number of simulation samples, the number of failed motion simulation samples, the number of simulation failure samples, and the motion failure rate.

[0112] To avoid the influence of simulated failure samples on the test results, the number of simulated failure samples implied in the estimated number of simulated failure samples can be obtained based on the number of simulated failure samples and the perceived failure rate. This allows for the correction of the perceived failure rate, resulting in a corrected failure rate. Compared to the original perceived failure rate, the corrected failure rate better reflects the actual perceived failure situation across the entire simulated sample space. This is illustrated in the following formula (6): (6) in, Indicates the failure rate of the correction. This indicates the number of failed motion simulation samples. This indicates the number of simulation failure samples. This represents the total number of simulation samples. This indicates the failure rate of the sensory experience.

[0113] Based on the above embodiments, the corrected failure rate can be determined based on the total number of simulation samples, the number of failed motion simulation samples, the number of simulation failure samples, and the motion failure rate. Therefore, by correcting the original motion failure rate, the interference of failed simulation samples on the results can be eliminated, improving the accuracy of the corrected failure rate in representing the real motion performance, making the final target feature vector more reliable, and thus improving the accuracy of predicting motion indicators.

[0114] In some embodiments of this disclosure, such as Figure 1 As shown, processor 101 performs a process based on the target feature vector and at least one historical feature vector to determine the similarity between the current version of the intelligent driving system and at least one historical version of the intelligent driving system, specifically configured as follows: The target feature vector and historical feature vector are standardized respectively to obtain the standardized target feature vector and the standardized historical feature vector.

[0115] In some embodiments, Z-score standardization can be used to standardize the target feature vector and each historical feature vector. Specifically, the statistical mean and statistical standard deviation of all historical feature vectors can be determined first; then, based on the statistical mean and statistical standard deviation, the component of each dimension in the target feature vector and each historical feature vector is standardized in turn, that is, each component is subtracted from the statistical mean and then divided by the statistical standard deviation, finally obtaining the standardized target feature vector and standardized historical feature vectors in which each component conforms to a normal distribution.

[0116] It should be noted that standardization can eliminate the interference of differences in units or numerical ranges between different feature dimensions on similarity calculation, thereby improving the stability and accuracy of subsequent similarity calculations.

[0117] Similarity is determined based on the standardized target feature vector and the standardized historical feature vector.

[0118] In some embodiments, the cosine similarity between the standardized target feature vector and the standardized historical feature vector can be calculated. Typically, the cosine similarity value ranges from [-1, 1]. By linearly shifting and scaling the cosine similarity to map it to the interval [0, 1], the similarity between the current version of the intelligent driving system and each historical version of the intelligent driving system can be determined.

[0119] For example, suppose the standardized target feature vector after standardization is =[ (1) , (2) ,…, (k)The standardized historical feature vector is =[ , ,…, The cosine similarity can be calculated using the following formula (7); and the similarity mapped to the [0,1] interval can be calculated using the following formula (8).

[0120] (7) (8) in, The cosine similarity between the standardized target feature vector and the standardized historical feature vector is represented. This indicates the similarity between the current version of the intelligent driving system and previous versions. This represents the magnitude of the standardized target feature vector. This represents the magnitude of the standardized historical feature vector.

[0121] The similarity obtained using the above method can accurately reflect the degree of similarity between the current version of the intelligent driving system and previous versions of the intelligent driving system in the same multiple simulation scenarios at the level of somatosensory features, providing a reliable basis for the selection of subsequent candidate versions of the intelligent driving system.

[0122] Based on the above embodiments, the dimensional differences of different feature dimensions are eliminated through standardization processing, and the geometric similarity of the somatosensory features between different versions is accurately quantified by similarity. This can intuitively reflect the consistency of the tendency of different versions of intelligent driving systems to cause discomfort to passengers, providing an accurate and reliable basis for the selection of subsequent candidate versions of intelligent driving systems, thereby helping to improve the credibility and accuracy of the final predicted somatosensory indicators.

[0123] Exemplary electronic devices Figure 2 A structural diagram of an electronic device provided in an embodiment of this disclosure includes at least one processor 111 and a memory 112.

[0124] The processor 111 may be a central processing unit (CPU) or other form of processing unit with data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device 11 to perform desired functions.

[0125] The memory 112 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 111 may execute one or more computer program instructions to implement the somatosensory indicator determination method and / or other desired functions of the intelligent driving system of the various embodiments of this disclosure described above.

[0126] In one example, the electronic device 11 may also include an input device 113 and an output device 114, which are interconnected via a bus system and / or other forms of connection mechanism (not shown).

[0127] The input device 113 may also include, for example, a keyboard, a mouse, etc.

[0128] The output device 114 can output various information to the outside, including, for example, a display, a speaker, a printer, and a communication network and its connected remote output devices, etc.

[0129] Of course, for the sake of simplicity, Figure 2 Only some of the components of the electronic device 11 relevant to this disclosure are shown, omitting components such as buses, input / output interfaces, etc. In addition, the electronic device 11 may include any other suitable components depending on the specific application.

[0130] Exemplary methods Figure 3 This is a flowchart illustrating a method for determining motion perception indicators in an intelligent driving system according to an exemplary embodiment of this disclosure. The method can be derived from... Figure 1 At least one processor 101 in the electronic device 100 executes the commands. For example... Figure 3 As shown, the method may include the following steps 301 to 305.

[0131] Step 301: Determine the target feature vectors of the current version of the intelligent driving system to be tested in multiple simulation scenarios.

[0132] Among them, the target feature vector is used to characterize the haptic quantification features of the current version of the intelligent driving system in multiple simulation scenarios.

[0133] Step 302: Obtain at least one historical feature vector.

[0134] The historical feature vector is the feature vector corresponding to at least one historical version of the intelligent driving system in multiple simulation scenarios.

[0135] Step 303: Based on the target feature vector and at least one historical feature vector, determine the similarity between the current version of the intelligent driving system and at least one historical version of the intelligent driving system.

[0136] Step 304: Based on similarity, determine candidate historical version intelligent driving systems from at least one historical version intelligent driving system.

[0137] Step 305: Based on the true haptic indicators and the corresponding similarity of the candidate historical versions of the intelligent driving system, determine the predicted haptic indicators of the current version of the intelligent driving system.

[0138] like Figure 4 As shown above, in the above Figure 3 Based on the illustrated embodiment, step 301 may include steps 3011-3012.

[0139] Step 3011: Obtain the correction failure rate of the current version of the intelligent driving system in the simulation scenario.

[0140] Step 3012: Determine the target feature vector based on multiple correction failure rates.

[0141] In some embodiments, step 3011 may specifically include: obtaining the total number of simulation samples, the number of failed motion simulation samples, and the number of failed simulation samples of the current version of the intelligent driving system in the simulation scenario; and determining the correction failure rate based on the total number of simulation samples, the number of failed motion simulation samples, and the number of failed simulation samples.

[0142] In some embodiments, the step of obtaining the number of failed haptic simulation samples of the current version of the intelligent driving system in a simulation scenario specifically includes: determining the target haptic score corresponding to the longitudinal event occurring in the simulation scenario for the current version of the intelligent driving system; and determining the number of failed haptic simulation samples of the current version of the intelligent driving system based on the target haptic score and a preset haptic threshold.

[0143] In some embodiments, the step of determining the target somatosensory score corresponding to a longitudinal event occurring in the simulation scenario by the current version of the intelligent driving system specifically includes: acquiring target simulation data corresponding to the simulation scenario of the current version of the intelligent driving system; determining jerk timing data based on acceleration timing data in the target simulation data; identifying longitudinal events in the simulation scenario based on the target simulation data, jerk timing data, and a preset event judgment threshold; acquiring multidimensional kinematic feature vectors of the longitudinal events; and determining the target somatosensory score corresponding to the longitudinal events through a weighted somatosensory function based on preset feature weights and multidimensional kinematic feature vectors.

[0144] In some embodiments, the step of obtaining the target simulation data corresponding to the current version of the intelligent driving system in the simulation scenario specifically includes: obtaining the initial simulation data of the current version of the intelligent driving system in the simulation scenario; and filtering the initial simulation data to obtain the target simulation data.

[0145] In some embodiments, the step of determining the correction failure rate based on the total number of simulation samples, the number of failed motion simulation samples, and the number of failed simulation samples specifically includes: determining the motion simulation failure rate based on the total number of simulation samples, the number of failed motion simulation samples, and the number of failed simulation samples; and determining the correction failure rate based on the total number of simulation samples, the number of failed motion simulation samples, the number of failed simulation samples, and the motion simulation failure rate.

[0146] In some embodiments, step 303 may specifically include: standardizing the target feature vector and the historical feature vector respectively to obtain a standardized target feature vector and a standardized historical feature vector; and determining the similarity based on the standardized target feature vector and the standardized historical feature vector.

[0147] It is understood that the specific execution methods and corresponding beneficial effects of each step in the method for determining the somatosensory indicators of the intelligent driving system in the above embodiments have been described in detail in the corresponding embodiments of the exemplary device section above. Please refer to the corresponding execution methods and beneficial technical effects of the exemplary device section above, and they will not be repeated here.

[0148] Exemplary device Figure 5 This is a schematic diagram of the structure of a somatosensory index determination device for an intelligent driving system provided in an embodiment of this disclosure, as shown below. Figure 5As shown, the somatosensory index determination device 400 includes: a first determination module 401, used to determine the target feature vector corresponding to the current version of the intelligent driving system under multiple simulation scenarios; wherein the target feature vector is used to characterize the somatosensory quantitative features of the current version of the intelligent driving system under the multiple simulation scenarios; a first acquisition module 402, used to acquire at least one historical feature vector; the historical feature vector is the feature vector corresponding to at least one historical version of the intelligent driving system under the multiple simulation scenarios; a second determination module 403, used to determine the similarity between the current version of the intelligent driving system and the at least one historical version of the intelligent driving system based on the target feature vector and the at least one historical feature vector; a third determination module 404, used to determine candidate historical version of the intelligent driving system among the at least one historical version of the intelligent driving system based on the similarity; and a fourth determination module 405, used to determine the predicted somatosensory index of the current version of the intelligent driving system based on the ground truth somatosensory index corresponding to the candidate historical version of the intelligent driving system and the corresponding similarity.

[0149] In some embodiments, the first determining module 401 is specifically used to: obtain the correction failure rate of the current version of the intelligent driving system in the simulation scenario; and determine the target feature vector based on multiple correction failure rates.

[0150] In some embodiments, the first determining module 401 is specifically used to: obtain the total number of simulation samples, the number of failed motion simulation samples, and the number of failed simulation samples of the current version of the intelligent driving system in the simulation scenario; and determine the correction failure rate based on the total number of simulation samples, the number of failed motion simulation samples, and the number of failed simulation samples.

[0151] In some embodiments, the first determining module 401 is specifically used to: determine the target somatosensory score corresponding to the longitudinal event occurring in the simulation scenario by the current version of the intelligent driving system; and determine the number of somatosensory simulation failure samples of the current version of the intelligent driving system based on the target somatosensory score and a preset somatosensory threshold.

[0152] In some embodiments, the first determining module 401 is specifically configured to: acquire target simulation data corresponding to the current version of the intelligent driving system in the simulation scenario; determine jerk timing data based on acceleration timing data in the target simulation data; identify the longitudinal event in the simulation scenario based on the target simulation data, the jerk timing data, and a preset event determination threshold; acquire the multidimensional kinematic feature vector of the longitudinal event; and determine the target somatosensory score corresponding to the longitudinal event through a weighted somatosensory function based on preset feature weights and the multidimensional kinematic feature vector.

[0153] In some embodiments, the first determining module 401 is specifically used to: obtain the initial simulation data of the current version of the intelligent driving system in the simulation scenario; and perform filtering processing on the initial simulation data to obtain the target simulation data.

[0154] In some embodiments, the first determining module 401 is specifically configured to: determine the motion simulation failure rate based on the total number of simulation samples, the number of motion simulation failure samples, and the number of simulation failure samples; and determine the correction failure rate based on the total number of simulation samples, the number of motion simulation failure samples, the number of simulation failure samples, and the motion simulation failure rate.

[0155] In some embodiments, the second determining module 403 is specifically used to: perform standardization processing on the target feature vector and the historical feature vector respectively to obtain a standardized target feature vector and a standardized historical feature vector; and determine the similarity based on the standardized target feature vector and the standardized historical feature vector.

[0156] The beneficial technical effects corresponding to the exemplary embodiments of this device can be found in the corresponding beneficial technical effects of the exemplary device section above, and will not be repeated here.

[0157] Exemplary computer program products and computer-readable storage media In addition to the methods and devices described above, embodiments of this disclosure may also provide a computer program product, including computer program instructions, which, when executed by a processor, cause the processor to perform the steps in the method for determining the somatosensory indicators of an intelligent driving system according to the various embodiments of this disclosure described in the "Exemplary Methods" section above.

[0158] Computer program products can be written in any combination of one or more programming languages ​​to perform operations of the embodiments of this disclosure. These 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 a user's computing device, partially on a user's computing device, as a standalone software package, partially on a user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0159] Furthermore, embodiments of this disclosure may also be computer-readable storage media storing computer program instructions thereon, which, when executed by a processor, cause the processor to perform the steps in the methods for determining the somatosensory indicators of an intelligent driving system according to the various embodiments of this disclosure described in the "Exemplary Methods" section above.

[0160] 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 include, but is not limited to, systems, apparatuses, or devices that are electrical, magnetic, optical, electromagnetic, infrared, or semiconductor, 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.

[0161] The basic principles of this disclosure have been described above with reference to specific embodiments. However, the advantages, benefits, and effects mentioned in this disclosure are merely examples and not limitations, and should not be considered as essential features of each embodiment of this disclosure. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the scope of this disclosure to the necessity of employing the aforementioned specific details for implementation.

[0162] Various modifications and variations can be made to this disclosure without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this disclosure and their equivalents, this disclosure is also intended to include such modifications and variations.

Claims

1. A device for determining the somatosensory indicators of an intelligent driving system, comprising at least one processor, the processor being configured to: Determine the target feature vectors of the current version of the intelligent driving system under test in multiple simulation scenarios; among which, The target feature vector is used to characterize the haptic quantification features of the current version of the intelligent driving system under the multiple simulation scenarios; Obtain at least one historical feature vector; the historical feature vector is the feature vector corresponding to at least one historical version of the intelligent driving system under the multiple simulation scenarios. Based on the target feature vector and the at least one historical feature vector, the similarity between the current version of the intelligent driving system and the at least one historical version of the intelligent driving system is determined. Based on the similarity, candidate historical version intelligent driving systems are determined in the at least one historical version intelligent driving system. Based on the true haptic index corresponding to the candidate historical version of the intelligent driving system and the corresponding similarity, the predicted haptic index of the current version of the intelligent driving system is determined.

2. The apparatus according to claim 1, wherein, The processor is configured to determine the target feature vectors of the current version of the intelligent driving system under test in multiple simulation scenarios, specifically as follows: Obtain the correction failure rate of the current version of the intelligent driving system in the simulation scenario; The target feature vector is determined based on multiple correction failure rates.

3. The apparatus according to claim 2, wherein, The processor is configured to obtain the correction failure rate of the current version of the intelligent driving system in the simulation scenario, specifically as follows: Obtain the total number of simulation samples, the number of failed motion simulation samples, and the number of failed simulation samples of the current version of the intelligent driving system in the simulation scenario; The correction failure rate is determined based on the total number of simulation samples, the number of failed motion simulation samples, and the number of failed simulation samples.

4. The apparatus according to claim 3, wherein, The processor is specifically configured to obtain the number of failed motion simulation samples of the current version of the intelligent driving system in the simulation scenario. Determine the target somatosensory score corresponding to a longitudinal event occurring in the simulation scenario in the current version of the intelligent driving system; Based on the target somatosensory score and the preset somatosensory threshold, the number of somatosensory simulation failure samples of the current version of the intelligent driving system is determined.

5. The apparatus according to claim 4, wherein, The processor is configured to determine the target haptic score corresponding to a longitudinal event occurring in the simulation scenario within the current version of the intelligent driving system. Obtain the target simulation data of the current version of the intelligent driving system in the simulation scenario; Based on the acceleration time-series data in the target simulation data, determine the jerk time-series data; Based on the target simulation data, the jerk timing data, and the preset event determination threshold, the longitudinal event in the simulation scene is identified; Obtain the multidimensional kinematic feature vector of the longitudinal event; Based on preset feature weights and the multidimensional kinematic feature vector, the target somatosensory score corresponding to the longitudinal event is determined by a weighted somatosensory function.

6. The apparatus according to claim 5, wherein, The processor is specifically configured to acquire the target simulation data corresponding to the current version of the intelligent driving system in the simulation scenario. Obtain the initial simulation data of the current version of the intelligent driving system in the simulation scenario; The initial simulation data is filtered to obtain the target simulation data.

7. The apparatus according to claim 3, wherein, The processor executes a process to determine the correction failure rate based on the total number of simulation samples, the number of failed motion simulation samples, and the number of failed simulation samples, specifically configured as follows: The motion simulation failure rate is determined based on the total number of simulation samples, the number of failed motion simulation samples, and the number of simulation failure samples. The correction failure rate is determined based on the total number of simulation samples, the number of failed motion simulation samples, the number of simulation failure samples, and the motion failure rate.

8. The apparatus according to claim 1, wherein, The processor performs a similarity determination between the current version of the intelligent driving system and the at least one historical version of the intelligent driving system based on the target feature vector and the at least one historical feature vector, specifically configured as follows: The target feature vector and the historical feature vector are standardized respectively to obtain the standardized target feature vector and the standardized historical feature vector. The similarity is determined based on the standardized target feature vector and the standardized historical feature vector.

9. A method for determining the somatosensory indicators of an intelligent driving system, comprising: Determine the target feature vectors of the current version of the intelligent driving system to be tested in multiple simulation scenarios; wherein, the target feature vectors are used to characterize the haptic quantification features of the current version of the intelligent driving system in the multiple simulation scenarios; Obtain at least one historical feature vector; the historical feature vector is the feature vector corresponding to at least one historical version of the intelligent driving system under the multiple simulation scenarios. Based on the target feature vector and the at least one historical feature vector, the similarity between the current version of the intelligent driving system and the at least one historical version of the intelligent driving system is determined. Based on the similarity, candidate historical version intelligent driving systems are determined in the at least one historical version intelligent driving system. Based on the true haptic index corresponding to the candidate historical version of the intelligent driving system and the corresponding similarity, the predicted haptic index of the current version of the intelligent driving system is determined.

10. A device for determining the somatosensory indicators of an intelligent driving system, comprising: The first determining module is used to determine the target feature vector of the current version of the intelligent driving system under test in multiple simulation scenarios; wherein, the target feature vector is used to characterize the haptic quantification features of the current version of the intelligent driving system in the multiple simulation scenarios; The first acquisition module is used to acquire at least one historical feature vector; the historical feature vector is the feature vector corresponding to at least one historical version of the intelligent driving system under the multiple simulation scenarios. The second determining module is used to determine the similarity between the current version of the intelligent driving system and the at least one historical version of the intelligent driving system based on the target feature vector and the at least one historical feature vector. The third determining module is used to determine candidate historical version intelligent driving systems in the at least one historical version intelligent driving system based on the similarity. The fourth determining module is used to determine the predicted haptic index of the current version of the intelligent driving system based on the true haptic index corresponding to the candidate historical version of the intelligent driving system and the corresponding similarity.

11. A computer-readable storage medium storing a computer program for executing the somatosensory index determination method of the intelligent driving system according to claim 9.

12. An electronic device, the electronic device comprising: processor; Memory used to store the processor's executable instructions; The processor is configured to read the executable instructions from the memory and execute the instructions to implement the method for determining the somatosensory indicators of the intelligent driving system as described in claim 9.

Images