State space model-based autonomous driving system trustworthiness estimation method and device

Abstract

The application provides a state space model-based automatic driving system trust degree estimation method and device, and relates to the technical field of automatic driving, wherein the method comprises the following steps: acquiring event indexes, experience indexes and behavior indexes in a current time period; acquiring a current trust degree of a driver to an automatic driving system in the current time period; inputting the current trust degree, the event indexes, the experience indexes and the behavior indexes into a state space model to obtain a trust degree of the driver to the automatic driving system in a next time period.The application can automatically capture the implicit relationship between each index and the change of the trust degree of the driver based on the event indexes, the experience indexes and the behavior indexes of the driver in the current time period and by using a pre-trained state space model, so that the trust degree of the driver to the automatic driving system is automatically inferred by the state space model, and the trust degree estimation efficiency of the automatic driving system is improved.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving technology, and in particular to a method and apparatus for estimating the trust level of an autonomous driving system based on a state-space model. Background Technology

[0002] In autonomous driving scenarios, the driver's trust in the AD (Automated Driving) system is fundamental to the driver's use of the AD system and is also an important standard for measuring the quality of the AD system. The driver's trust in the AD system is a highly abstract concept, and the degree of trust will change dynamically with the human-computer interaction experience.

[0003] In related technologies, questionnaires are typically used to obtain drivers' trust in AD systems at different time periods.

[0004] However, among the aforementioned technologies, the use of questionnaires to obtain drivers' trust in the AD system is difficult to scale up, which reduces the efficiency of trust estimation for autonomous driving systems. Summary of the Invention

[0005] This invention provides a method and apparatus for estimating the trust level of an autonomous driving system based on a state-space model, in order to address the shortcomings of existing technologies that reduce the efficiency of trust level estimation in autonomous driving systems.

[0006] This invention provides a method for estimating the trust level of an autonomous driving system based on a state-space model, comprising the following steps.

[0007] The system acquires event metrics, experience metrics, and behavioral metrics within the current time period. The event metrics include metrics that affect the driver's trust in the autonomous driving system. The experience metrics include metrics related to the interaction between the driver and the autonomous driving system. The behavioral metrics include metrics related to the driver's use of the autonomous driving system. Obtain the driver's current level of trust in the autonomous driving system during the current time period; The current trust level, the event index, the experience index, and the behavior index are input into the state space model to obtain the driver's trust level in the autonomous driving system in the next time period, which is output by the state space model. The state space model is trained based on at least one driver sample in different first time periods, including first event index samples, first experience index samples, first behavior index samples, and trust labels.

[0008] According to the present invention, a trust estimation method for an autonomous driving system based on a state-space model is provided, wherein the event indicators include at least one of the following: the number of lane departures under the autonomous driving system, the number of times the autonomous driving system exits, the number of times the driver brakes suddenly under the autonomous driving system, and the number of times the driver makes a lane change while the vehicle is in the autonomous driving system.

[0009] According to the present invention, a trust estimation method for an autonomous driving system based on a state-space model is provided, wherein the experience indicators include the success rate of intelligent lane changing and / or the success rate of adaptive cruise under the autonomous driving system.

[0010] According to the present invention, a trust estimation method for an autonomous driving system based on a state-space model is provided, wherein the behavioral indicators include the percentage of mileage the driver uses the autonomous driving system during the current time period and / or the number of times the driver actively takes over control of the vehicle.

[0011] According to the present invention, a trust estimation method for an autonomous driving system based on a state-space model is provided, wherein the state-space model is trained in the following manner: Acquire first event indicator samples, first experience indicator samples, and first behavior indicator samples from multiple driver samples in different first time periods; The average state space parameters are obtained by estimating the state space parameters of the initial state space model based on the trust labels of each driver sample to the autonomous driving system at the end of the current first time period sample, the trust labels of each driver sample to the autonomous driving system at the end of the next first time period sample, each first event index sample, each first experience index sample, and each first behavior index sample. The state-space model is determined based on the average state-space parameters.

[0012] According to the present invention, a trust estimation method for an autonomous driving system based on a state-space model is provided, wherein the state-space model is trained in the following manner: Obtain the second event index sample, the second experience index sample, and the second behavior index sample of the driver sample in different second time periods; Personalized state space parameters are obtained by estimating the state space parameters of the initial state space model using a linear mixed-effects model based on the driver's trust label for the autonomous driving system at the end of the current second time period sample, the driver's trust label for the autonomous driving system at the end of the next second time period sample, each second event indicator sample, each second experience indicator sample, and each second behavior indicator sample. Based on the personalized state space parameters, the state space model corresponding to the driver sample is determined.

[0013] According to the present invention, a method for estimating the trust level of an autonomous driving system based on a state-space model is provided, the method further includes: The estimation model is constructed based on the following formula (1): The updated model is constructed based on the following formula (2): in, This represents the end time of the driver's sample for the autonomous driving system in the current first time period. Trust rating labels This indicates the end time of the driver's sample for the autonomous driving system in the next first time period. Trust rating labels This represents the first event metric sample. This represents the first experience indicator sample. This represents the first behavioral indicator sample. All of these are state-space parameters to be estimated. Linear transition matrix, Noise term; The combination of the estimated model and the updated model is determined as the initial state space model.

[0014] According to the present invention, a method for estimating the trust level of an autonomous driving system based on a state-space model is provided. The method inputs the current trust level, the event metric, the experience metric, and the behavior metric into a state-space model to obtain the driver's trust level in the autonomous driving system in the next time period, as output by the state-space model. This includes: The current trust level, the event index, and the experience index are input into the estimation model of the state space model corresponding to the driver to obtain the driver's initial trust level of the autonomous driving system in the next time period output by the estimation model. The initial trust level and the behavioral indicators are input into the update model of the state space model corresponding to the driver. The initial trust level is corrected by a Kalman filter based on the behavioral indicators to obtain the driver's trust level in the autonomous driving system in the next time period, which is output by the update model.

[0015] The present invention also provides a trust estimation device for an autonomous driving system based on a state-space model, comprising: The first acquisition unit is used to acquire event indicators, experience indicators and behavior indicators within the current time period. The event indicators include indicators that affect the driver's trust in the autonomous driving system. The experience indicators include indicators related to the interaction between the driver and the autonomous driving system. The behavior indicators include indicators related to the driver's use of the autonomous driving system. The second acquisition unit is used to acquire the driver's current level of trust in the autonomous driving system during the current time period; An estimation unit is used to input the current trust level, the event index, the experience index, and the behavior index into a state space model to obtain the driver's trust level in the autonomous driving system in the next time period, as output by the state space model. The state space model is trained based on at least one driver sample in different first time periods, including first event index samples, first experience index samples, first behavior index samples, and trust labels.

[0016] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the confidence estimation method for an autonomous driving system based on a state-space model as described above.

[0017] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the confidence estimation method for an autonomous driving system based on a state-space model as described above.

[0018] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the trust estimation method for an autonomous driving system based on a state-space model as described above.

[0019] This invention provides a method and apparatus for estimating trust in autonomous driving systems based on a state-space model. It acquires event metrics, experience metrics, and behavioral metrics within the current time period, and obtains the driver's current trust level in the autonomous driving system during this time period. The current trust level, event metrics, experience metrics, and behavioral metrics are then input into a pre-trained state-space model to obtain the driver's trust level in the autonomous driving system for the next time period, output by the state-space model. Therefore, this invention can automatically capture the implicit relationship between the driver's event metrics, experience metrics, and behavioral metrics within the current time period using a pre-trained state-space model. This enables automatic inference of the driver's trust level in the autonomous driving system through the state-space model, improving the efficiency of trust estimation for autonomous driving systems. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0021] Figure 1 This is a flowchart illustrating the trust estimation method for autonomous driving systems based on state-space models provided in this embodiment of the invention.

[0022] Figure 2 This is a schematic diagram of the training process of the state-space model provided in an embodiment of the present invention.

[0023] Figure 3 This is a schematic diagram illustrating the evolution of various indicators and trust levels over time during the use of the autonomous driving system by a driver sample provided in this embodiment of the invention.

[0024] Figure 4 This is a schematic diagram of the structure of the trust estimation device for an autonomous driving system based on a state-space model provided in an embodiment of the present invention.

[0025] Figure 5 This is a schematic diagram of the physical structure of the electronic device provided in an embodiment of the present invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0027] The following is combined with Figures 1-3 This invention describes a trust estimation method for autonomous driving systems based on a state-space model. The execution entity of this state-space model-based trust estimation method can be an electronic device such as an in-vehicle device, computer, tablet computer, terminal, or server, or it can be a state-space model-based trust estimation device installed in such an electronic device. This state-space model-based trust estimation device can be implemented through software, hardware, or a combination of both.

[0028] Figure 1 This is a flowchart illustrating the trust estimation method for autonomous driving systems based on state-space models provided in this embodiment of the invention. Figure 1As shown, the trust estimation method for autonomous driving systems based on the state-space model includes the following steps: Step 101: Obtain event metrics, experience metrics, and behavior metrics within the current time period. The event metrics include metrics that affect the driver's trust in the autonomous driving system. The experience metrics include metrics related to the interaction between the driver and the autonomous driving system. The behavior metrics include metrics related to the driver's use of the autonomous driving system.

[0029] The event metrics include at least one of the following: the number of lane departures under the autonomous driving system, the number of times the autonomous driving system disengages, the number of times the driver brakes suddenly under the autonomous driving system, and the number of times the vehicle sways left and right or deviates from the predetermined trajectory under the autonomous driving system. The experience metrics include the success rate of intelligent lane changing and / or the success rate of adaptive cruise control under the autonomous driving system. The behavioral metrics include the percentage of mileage the driver uses under the autonomous driving system during the current time period and / or the number of times the driver actively takes over vehicle control.

[0030] For example, the length of the current time period can be set based on needs. For instance, if the current time period is three days, the system can collect data on the number of lane departures, fallbacks, emergency braking, and lane departures when the driver uses the autonomous driving system during the current time period. These events are negative indicators that affect the driver's trust in the autonomous driving system. The system can also collect the intelligent lane change success rate (ILC success rate) and adaptive cruise control success rate (ALC success rate) under the autonomous driving system. These are positive indicators. In addition, the system can collect the percentage of mileage the driver uses in the current time period (AD percentage) and the number of times the driver actively takes over vehicle control in the next time period, which can also be called the number of times the driver actively takes over. The AD percentage and the number of times the driver actively takes over can reflect the driver's usage of the autonomous driving system.

[0031] It should be noted that the experience indicators may also include the number of times the driver presses the accelerator and adjusts the speed under the autonomous driving system, but this invention does not limit this.

[0032] Step 102: Obtain the driver's current level of trust in the autonomous driving system during the current time period.

[0033] For example, when using it for the first time, if the current time period is the first time period, then the driver's current level of trust in the autonomous driving system during the current time period can be the level of trust obtained through a questionnaire survey. The level of trust can be represented by a trust score. For example, the driver's trust score for the autonomous driving system obtained through a questionnaire survey can be a value between 0 and 100, normalized to a value in the range [0,1]. If the driver does not have an initial level of trust, the average of the trust scores of all other drivers for the autonomous driving system obtained through the questionnaire survey is used as the initial level of trust. If the current time period is not the first time period, then the driver's current level of trust in the autonomous driving system during the current time period is the level of trust output by the state space model below.

[0034] Step 103: Input the current trust level, the event index, the experience index, and the behavior index into the state space model to obtain the driver's trust level in the autonomous driving system in the next time period output by the state space model; wherein, the state space model is trained based on the first event index sample, the first experience index sample, the first behavior index sample, and the trust level label of at least one driver sample in different first time periods.

[0035] Trust ratings can be obtained through questionnaires.

[0036] For example, the current trust level, event indicators, experience indicators, and behavior indicators are input into a pre-trained state space model. Through the state space model, the implicit temporal state of trust level is learned using Kalman filtering, capturing the implicit relationship between each indicator and the driver's trust level changes. Finally, the driver's trust level in the autonomous driving system in the next time period is obtained from the state space model output.

[0037] When obtaining the driver's trust level in the autonomous driving system in the next time period, the driver's trust level in the next time period, the event indicators, experience indicators, and behavioral indicators in the next time period can be input into the state space model. The output of the state space model is the driver's trust level in the next time period after that, with the trust level being a value in the range [0,1]. By repeating this process, the driver's trust level in the autonomous driving system in different time periods can be obtained. In this process, it is only necessary to obtain the driver's trust level in the autonomous driving system in the first time period through a questionnaire survey. Subsequently, the driver's trust level in the autonomous driving system in other time periods can be obtained through the pre-trained state space model, without the need for frequent questionnaire surveys.

[0038] It should be noted that the driver's trust level in the autonomous driving system at different times can be used to evaluate the version of the autonomous driving system, as well as to assess what specific scenarios can cause changes in trust level.

[0039] This invention provides a method for estimating trust in autonomous driving systems based on a state-space model. It acquires event metrics, experience metrics, and behavioral metrics within the current time period, and obtains the driver's current trust level in the autonomous driving system during this time period. The current trust level, event metrics, experience metrics, and behavioral metrics are then input into a pre-trained state-space model to obtain the driver's trust level in the autonomous driving system for the next time period, output by the state-space model. This invention can automatically capture the implicit relationship between the driver's event metrics, experience metrics, and behavioral metrics within the current time period using a pre-trained state-space model. This allows for the automatic inference of the driver's trust level in the autonomous driving system through the state-space model, improving the efficiency of trust estimation and overcoming problems such as interference and inefficiency from questionnaire surveys.

[0040] In one embodiment, Figure 2 This is a schematic diagram of the training process of the state-space model provided in an embodiment of the present invention, as shown below. Figure 2 As shown, the state-space model is trained in the following manner: Step 201: Obtain the first event indicator sample, the first experience indicator sample, and the first behavior indicator sample of multiple driver samples in different first time periods.

[0041] For example, a sample of 20 to 40 drivers can be acquired and tracked for 4 to 6 weeks, with each 3-day period constituting a first time interval sample. For each first time interval sample, first event indicator samples, first experience indicator samples, and first behavior indicator samples are collected. The first event indicator samples include the first lane departure frequency under the autonomous driving system, the first time the autonomous driving system disengaged, the first emergency braking frequency of the driver sample under the autonomous driving system, and the first time the vehicle made a lane change maneuver. The first experience indicator samples include the first intelligent lane change success rate and the first adaptive cruise control success rate under the autonomous driving system. The first behavior indicator samples include the first mileage percentage of the driver sample using the autonomous driving system within the first time interval sample and the first number of times the driver actively took over the autonomous driving system. The first event indicator samples, first experience indicator samples, and first behavior indicator samples are scalable objective interactive data.

[0042] Step 202: Estimate the state space parameters of the initial state space model using a linear mixed-effects model based on the trust labels of each driver sample to the autonomous driving system at the end of the current first time period sample, the trust labels of each driver sample to the autonomous driving system at the end of the next first time period sample, each first event index sample, each first experience index sample, and each first behavior index sample.

[0043] For example, based on the time-series data of each driver sample, the state-space parameters of the initial state-space model are estimated using a linear mixed-effects model. Specifically, the trust label of each driver sample at the end of the current first time period sample and the trust label at the end of the next first time period sample are integrated. Combined with the first event indicator samples (first lane departure times, first number of times the autonomous driving system exited, first number of times of emergency braking, and first number of times the vehicle made a sudden stop), first experience indicator samples (first intelligent lane change success rate and first adaptive cruise success rate), and first behavior indicator samples (first mileage percentage and first number of times active takeover was performed) recorded within the current first time period sample, the state-space parameters of the initial state-space model are statistically fitted and estimated. Finally, average state-space parameters that can represent the average evolution law of the entire driver group are obtained. These average state-space parameters quantify the implicit relationship between the trust changes of each indicator sample and the driver sample.

[0044] Step 203: Determine the state space model based on the average state space parameters.

[0045] For example, when the average state space parameters are estimated, they can be substituted into the initial state space model to obtain the trained state space model. The trained state space model can then be deployed to the driver's in-vehicle device or cloud server, making it easier to estimate the driver's trust in the autonomous driving system at different time periods based on this state space model.

[0046] It should be noted that if there are missing or contradictory data in the process of obtaining the first event indicator sample, the first experience indicator sample, and the first behavior indicator sample, interpolation and logical correction can be used to ensure the quality of the training data. This invention will not elaborate on these points here.

[0047] In this embodiment, a linear mixed-effects model is used to estimate the state space parameters of the initial state space model based on the trust labels of each driver sample to the autonomous driving system at the end of the current first time period sample, the trust labels of each driver sample to the autonomous driving system at the end of the next first time period sample, each first event indicator sample, each first experience indicator sample, and each first behavior indicator sample. Finally, the state space model is obtained, which enables the trained state space model to automatically capture the implicit relationship between each indicator and the driver's trust change, thereby realizing the automatic inference of the driver's trust in the autonomous driving system through the state space model.

[0048] In one embodiment, the state-space model is trained in the following manner: The process involves acquiring second event index samples, second experience index samples, and second behavior index samples of the driver sample in different second time periods; using a linear mixed-effects model, based on the driver sample's trust label for the autonomous driving system at the end of the current second time period sample, the driver sample's trust label for the autonomous driving system at the end of the next second time period sample, each of the second event index samples, each of the second experience index samples, and each of the second behavior index samples, the state space parameters of the initial state space model are estimated to obtain personalized state space parameters; and based on the personalized state space parameters, the state space model corresponding to the driver sample is determined.

[0049] The second time period sample can be the same as or different from the first time period sample; this invention does not impose any limitation on this.

[0050] For example, the system collects second event indicator samples, second experience indicator samples, and second behavior indicator samples corresponding to driver samples in each second time period. The second event indicator samples include samples of the number of lane departures under the autonomous driving system, the number of times the autonomous driving system exits, the number of times the driver sample brakes suddenly under the autonomous driving system, and the number of times the vehicle makes a sudden stop. The second experience indicator samples include samples of the success rate of intelligent lane changing and the success rate of adaptive cruise control under the autonomous driving system. The second behavior indicator samples include samples of the percentage of mileage used by the driver sample using the autonomous driving system and the number of times the driver sample actively takes over the system within the second time period. The system also obtains the driver sample's trust label for the autonomous driving system at the end of the current second time period and the driver sample's trust label for the autonomous driving system at the end of the next second time period. Then, the state space parameters of the initial state space model are estimated using a linear mixed-effects model. Specifically, based on the driver's trust label for the autonomous driving system at the end of the current second time period and the trust label at the end of the next second time period, and combined with the second event indicator samples recorded within the current second time period (second lane departure times, second number of times the autonomous driving system exited, second number of times emergency braking, and second number of times the vehicle made sudden stops), second experience indicator samples (second intelligent lane change success rate and second adaptive cruise success rate), and second behavior indicator samples (second mileage percentage and second number of times active takeover occurred), the state space parameters of the initial state space model are statistically fitted and estimated. This yields personalized state space parameters that represent the driver's sample. These personalized state space parameters quantify the implicit relationship between each indicator sample and the driver's trust changes. Substituting these personalized state space parameters into the state space model yields the trained state space model corresponding to that driver's sample. Following the same method, the state space model for each driver sample can be obtained.

[0051] In this embodiment, the state space parameters of the initial state space model are estimated based on the trust labels of driver samples to the autonomous driving system at the end of the current second time period sample, the trust labels of each driver sample to the autonomous driving system at the end of the next second time period sample, each second event index sample, each second experience index sample, and each second behavior index sample. Then, a state space model for driver samples is trained, thus realizing the personalization of the state space model.

[0052] In one embodiment, the trust estimation method for an autonomous driving system based on a state-space model further includes the following steps: The estimation model is constructed based on the following formula (1): The updated model is constructed based on the following formula (2): in, This represents the end time of the driver's sample for the autonomous driving system in the current first time period. Trust rating labels This indicates the end time of the driver's sample for the autonomous driving system in the next first time period. Trust rating labels This represents the first event metric sample. This represents the first experience indicator sample. This represents the first behavioral indicator sample. All of these are state-space parameters to be estimated. Linear transition matrix, This is the noise term.

[0053] The combination of the estimated model and the updated model is determined as the initial state space model.

[0054] For example, Figure 3 This is a schematic diagram illustrating the evolution of various indicators and trust levels over time during the use of the autonomous driving system by driver samples provided in this embodiment of the invention. Figure 3 As shown, T is the time axis. Taking three samples from the first time period as an example, the end times of the three samples from the first time period are as follows: , and Within each first time period sample (or mileage window sample), drivers experience various safety and experience events while using the AD system (including first event indicator samples and first experience indicator samples). Through human-computer interaction, adjustments to the AD system are made, generating the driver sample's mileage percentage using the AD system (AD percentage) and the first number of proactive takeover attempts (first behavior indicator samples). These indicators influence the driver sample's trust level before the first time period sample, either increasing or decreasing it. Therefore, at the end of the first time period sample, the driver sample will have a new trust level. The trust level label corresponding to each moment is , The trust level label corresponding to each moment is , The trust level label corresponding to each moment is In this process, the trust level of each first time period sample is implicit. The observable records are the first event indicator sample, the first experience indicator sample, and the first behavior indicator sample that affect the trust level. By substituting the first event indicator sample, the first experience indicator sample, and the first behavior indicator sample of all driver samples into the above formulas (1) and (2), the following can be obtained: The value, and The mean and variance of, The value and The mean and variance are used as the average state-space parameters. Substituting the average state-space parameters into the above formulas (1) and (2) makes the mean and variance of the state-space parameters equal to the mean and variance of the state-space parameters. , Fixed, trust level The model is constructed as continuous state variables, and the final trained state-space model is used to describe the dynamic evolution of trust.

[0055] In one embodiment, step 103 above inputs the current trust level, the event indicator, the experience indicator, and the behavior indicator into the state space model to obtain the driver's trust level in the autonomous driving system in the next time period, which can be achieved in the following way: The current trust level, the event metric, and the experience metric are input into the estimation model of the state space model corresponding to the driver to obtain the driver's initial trust level towards the autonomous driving system in the next time period, output by the estimation model. The initial trust level and the behavior metric are input into the update model of the state space model corresponding to the driver, and the initial trust level is corrected based on the behavior metric using a Kalman filter to obtain the driver's trust level towards the autonomous driving system in the next time period, output by the update model.

[0056] For example, a pre-trained state-space model corresponding to a driver sample is deployed in an electronic device such as an in-vehicle device or a cloud server corresponding to that driver sample. When the driver's event indicators, experience indicators, and behavior indicators are obtained, if a state-space model corresponding to that driver is deployed, the current trust level, event indicators, and experience indicators are input into the estimation model of the driver's state-space model to obtain the driver's initial trust level for the autonomous driving system in the next time period output by the estimation model. Then, the initial trust level and behavior indicators are input into the update model of the driver's state-space model. The initial trust level is corrected based on the behavior indicators by a Kalman filter to obtain the driver's trust level for the autonomous driving system in the next time period output by the update model.

[0057] If no state space model is deployed for the driver, it indicates that the driver is a new driver. In this case, the current trust level, event indicators, and experience indicators are input into the estimation model of the general state space model determined based on the average state space parameters. This yields the driver's initial trust level in the autonomous driving system for the next time period, as output by the estimation model. The initial trust level and behavioral indicators are then input into the update model of the general state space model. The initial trust level is corrected based on the behavioral indicators using a Kalman filter, resulting in the driver's trust level in the autonomous driving system for the next time period, as output by the update model. This enables cold start prediction for new drivers. As trust levels accumulate over different time periods, the average state space parameters can be adjusted based on the accumulated trust levels to ultimately obtain the state space model corresponding to the new driver.

[0058] In this embodiment, based on the driver's event indicators and experience indicators in the current time period, and the behavioral indicators in the next time period, the implicit relationship between each indicator and the driver's trust level changes can be automatically captured using a pre-trained state space model corresponding to the driver. This enables the automatic inference of the driver's trust level in the autonomous driving system through the state space model corresponding to the driver, improving the trust level estimation efficiency of the autonomous driving system and further improving the accuracy of the trust level estimation.

[0059] The effects of the state-space model of this invention are described below: After tracking 20 drivers for four weeks, a total of 220 driver trust labels were collected, which, along with event metrics, experience metrics, and behavioral metrics, constituted the dataset sample. To better evaluate the performance of the state-space model, 5-fold cross-validation was used. The dataset sample was divided into five parts, and training and testing were performed five times in a 4:1 ratio. The mean performance of the test set was used to evaluate the effectiveness of the state-space model. The results are as follows: 1) Improved accuracy: Accuracy: The state-space model's predictions are highly close to the true values. On the test set, the mean absolute error (MAE) of the confidence prediction is 0.10, and the mean absolute percentage error (MAPE) is 13%, which is better than other models such as Transformer.

[0060] Consistency: The Pearson correlation coefficient between the predicted results and the true values ​​is 0.6, and the intragroup consistency correlation coefficient ICC(C,1) is 0.41, indicating that the state-space model can effectively capture the trend changes in trust.

[0061] Generalization ability: The state-space model is stable for different drivers using the autonomous driving system, and it is particularly adaptable to atypical changing trends.

[0062] 2) Practical value: Provides a single trust metric for AD systems, replacing multi-dimensional atomic metrics and simplifying the evaluation process of AD systems.

[0063] It supports real-time monitoring and historical backtracking, assisting in version decision-making and scenario optimization of the AD system.

[0064] Reduce reliance on questionnaires and improve data collection efficiency.

[0065] The following describes the autonomous driving system trust estimation device based on the state space model provided by the present invention. The autonomous driving system trust estimation device based on the state space model described below and the autonomous driving system trust estimation method based on the state space model described above can be referred to in correspondence.

[0066] Figure 4 This is a schematic diagram of the structure of the trust estimation device for an autonomous driving system based on a state-space model provided in an embodiment of the present invention, as shown below. Figure 4 As shown, the autonomous driving system trust estimation device 400 based on the state-space model includes a first acquisition unit 401, a second acquisition unit 402, and an estimation unit 403; wherein: The first acquisition unit 401 is used to acquire event indicators, experience indicators and behavior indicators within the current time period. The event indicators include indicators that affect the driver's trust in the autonomous driving system. The experience indicators include indicators related to the interaction between the driver and the autonomous driving system. The behavior indicators include indicators related to the driver's use of the autonomous driving system. The second acquisition unit 402 is used to acquire the driver's current level of trust in the autonomous driving system during the current time period; Estimation unit 403 is used to input the current trust level, the event index, the experience index and the behavior index into the state space model to obtain the driver's trust level in the autonomous driving system in the next time period output by the state space model; The state space model is trained based on at least one driver sample in different first time periods, including first event index samples, first experience index samples, first behavior index samples, and trust labels.

[0067] The present invention provides a trust estimation device for autonomous driving systems based on a state-space model. It acquires event indicators, experience indicators, and behavioral indicators within the current time period, and obtains the driver's current trust level in the autonomous driving system within the current time period. The current trust level, event indicators, experience indicators, and behavioral indicators are input into a pre-trained state-space model to obtain the driver's trust level in the autonomous driving system in the next time period, output by the state-space model. Therefore, the present invention can automatically capture the implicit relationship between the driver's event indicators, experience indicators, and behavioral indicators within the current time period using a pre-trained state-space model, thereby achieving automatic inference of the driver's trust level in the autonomous driving system through the state-space model and improving the efficiency of trust estimation for autonomous driving systems.

[0068] Based on any of the above embodiments, the event indicators include at least one of the following: the number of lane departures under the autonomous driving system, the number of times the autonomous driving system disengages, the number of times the driver brakes suddenly under the autonomous driving system, and the number of times the vehicle makes a sudden stop under the autonomous driving system.

[0069] Based on any of the above embodiments, the experience metrics include the success rate of intelligent lane changing and / or the success rate of adaptive cruise control under the autonomous driving system.

[0070] Based on any of the above embodiments, the behavioral indicators include the percentage of mileage the driver uses the autonomous driving system during the current time period and / or the number of times the driver actively takes over control of the vehicle.

[0071] Based on any of the above embodiments, the state-space model is trained in the following manner: Acquire first event indicator samples, first experience indicator samples, and first behavior indicator samples from multiple driver samples in different first time periods; The average state space parameters are obtained by estimating the state space parameters of the initial state space model based on the trust labels of each driver sample to the autonomous driving system at the end of the current first time period sample, the trust labels of each driver sample to the autonomous driving system at the end of the next first time period sample, each first event index sample, each first experience index sample, and each first behavior index sample. The state-space model is determined based on the average state-space parameters.

[0072] Based on any of the above embodiments The state-space model is trained in the following manner: Obtain the second event indicator sample, the second experience indicator sample, and the second behavior indicator sample of the driver sample in different second time periods; Personalized state space parameters are obtained by estimating the state space parameters of the initial state space model using a linear mixed-effects model based on the driver's trust label for the autonomous driving system at the end of the current second time period sample, the driver's trust label for the autonomous driving system at the end of the next second time period sample, each second event indicator sample, each second experience indicator sample, and each second behavior indicator sample. Based on the personalized state space parameters, the state space model corresponding to the driver sample is determined.

[0073] Based on any of the above embodiments, the device further includes: The first building unit is used to construct the estimation model based on the following formula (1): The second building unit is used to build the updated model based on the following formula (2): in, This represents the end time of the driver's sample for the autonomous driving system in the current first time period. Trust rating labels This indicates the end time of the driver's sample for the autonomous driving system in the next first time period. Trust rating labels This represents the first event metric sample. This represents the first experience indicator sample. This represents the first behavioral indicator sample. All of these are state-space parameters to be estimated. Linear transition matrix, Noise term; The determining unit is used to determine the combination of the estimated model and the updated model as the initial state space model.

[0074] Based on any of the above embodiments, the estimation unit 403 is specifically used for: The current trust level, the event index, and the experience index are input into the estimation model of the state space model corresponding to the driver to obtain the driver's initial trust level of the autonomous driving system in the next time period output by the estimation model. The initial trust level and the behavioral indicators are input into the update model of the state space model corresponding to the driver. The initial trust level is corrected by a Kalman filter based on the behavioral indicators to obtain the driver's trust level in the autonomous driving system in the next time period, which is output by the update model.

[0075] Figure 5 This is a schematic diagram of the physical structure of the electronic device provided in the embodiments of the present invention, such as... Figure 5 As shown, the electronic device may include a processor 510, a communications interface 520, a memory 530, and a communication bus 540, wherein the processor 510, communications interface 520, and memory 530 communicate with each other via the communication bus 540. The processor 510 can call logical instructions in the memory 530 to execute a state-space model-based autonomous driving system trust estimation method. This method includes: acquiring event indicators, experience indicators, and behavioral indicators within the current time period; the event indicators include indicators affecting the driver's trust in the autonomous driving system; the experience indicators include indicators related to the interaction between the driver and the autonomous driving system; and the behavioral indicators include indicators related to the driver's use of the autonomous driving system. Obtain the driver's current level of trust in the autonomous driving system during the current time period; The current trust level, the event index, the experience index, and the behavior index are input into the state space model to obtain the driver's trust level in the autonomous driving system in the next time period, which is output by the state space model. The state space model is trained based on at least one driver sample in different first time periods, including first event index samples, first experience index samples, first behavior index samples, and trust labels.

[0076] Furthermore, the logical instructions in the aforementioned memory 530 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0077] On the other hand, the present invention also provides a computer program product, the computer program product including a computer program, the computer program being able to be stored on a non-transitory computer-readable storage medium, the computer program being executed by a processor, the computer being able to execute the autonomous driving system trust estimation method based on the state space model provided by the above methods, the method including: acquiring event indicators, experience indicators and behavioral indicators in the current time period, the event indicators including indicators affecting the driver's trust in the autonomous driving system, the experience indicators including indicators related to the interaction between the driver and the autonomous driving system, the behavioral indicators including indicators related to the driver's use of the autonomous driving system; Obtain the driver's current level of trust in the autonomous driving system during the current time period; The current trust level, the event index, the experience index, and the behavior index are input into the state space model to obtain the driver's trust level in the autonomous driving system in the next time period, which is output by the state space model. The state space model is trained based on at least one driver sample in different first time periods, including first event index samples, first experience index samples, first behavior index samples, and trust labels.

[0078] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the state-space model-based autonomous driving system trust estimation method provided by the above methods. The method includes: acquiring event indicators, experience indicators, and behavioral indicators in the current time period, wherein the event indicators include indicators that affect the driver's trust in the autonomous driving system, the experience indicators include indicators related to the interaction between the driver and the autonomous driving system, and the behavioral indicators include indicators related to the driver's use of the autonomous driving system. Obtain the driver's current level of trust in the autonomous driving system during the current time period; The current trust level, the event index, the experience index, and the behavior index are input into the state space model to obtain the driver's trust level in the autonomous driving system in the next time period, which is output by the state space model. The state space model is trained based on at least one driver sample in different first time periods, including first event index samples, first experience index samples, first behavior index samples, and trust labels.

[0079] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0080] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0081] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for estimating the trust level of an autonomous driving system based on a state-space model, characterized in that, include: The system acquires event metrics, experience metrics, and behavioral metrics within the current time period. The event metrics include metrics that affect the driver's trust in the autonomous driving system. The experience metrics include metrics related to the interaction between the driver and the autonomous driving system. The behavioral metrics include metrics related to the driver's use of the autonomous driving system. Obtain the driver's current level of trust in the autonomous driving system during the current time period; The current trust level, the event index, the experience index, and the behavior index are input into the state space model to obtain the driver's trust level in the autonomous driving system in the next time period, which is output by the state space model. The state space model is trained based on at least one driver sample in different first time periods, including first event index samples, first experience index samples, first behavior index samples, and trust labels.

2. The trust estimation method for autonomous driving systems based on state-space models according to claim 1, characterized in that, The event metrics include at least one of the following: the number of lane departures under the autonomous driving system, the number of times the autonomous driving system disengages, the number of times the driver brakes suddenly under the autonomous driving system, and the number of times the vehicle makes a sudden stop under the autonomous driving system.

3. The trust estimation method for autonomous driving systems based on state-space models according to claim 1, characterized in that, The experience metrics include the success rate of intelligent lane changing and / or the success rate of adaptive cruise control under the autonomous driving system.

4. The trust estimation method for autonomous driving systems based on state-space models according to claim 1, characterized in that, The behavioral metrics include the percentage of mileage the driver uses the autonomous driving system during the current time period and / or the number of times the driver actively takes over control of the vehicle.

5. The trust estimation method for autonomous driving systems based on state-space models according to claim 1, characterized in that, The state-space model is trained in the following manner: Acquire first event indicator samples, first experience indicator samples, and first behavior indicator samples from multiple driver samples in different first time periods; The average state space parameters are obtained by estimating the state space parameters of the initial state space model based on the trust labels of each driver sample to the autonomous driving system at the end of the current first time period sample, the trust labels of each driver sample to the autonomous driving system at the end of the next first time period sample, each first event index sample, each first experience index sample, and each first behavior index sample. The state-space model is determined based on the average state-space parameters.

6. The trust estimation method for an autonomous driving system based on a state-space model according to claim 1, characterized in that, The state-space model is trained in the following manner: Obtain the second event indicator sample, the second experience indicator sample, and the second behavior indicator sample of the driver sample in different second time periods; Personalized state space parameters are obtained by estimating the state space parameters of the initial state space model using a linear mixed-effects model based on the driver's trust label for the autonomous driving system at the end of the current second time period sample, the driver's trust label for the autonomous driving system at the end of the next second time period sample, each second event indicator sample, each second experience indicator sample, and each second behavior indicator sample. Based on the personalized state space parameters, the state space model corresponding to the driver sample is determined.

7. The trust estimation method for an autonomous driving system based on a state-space model according to claim 5, characterized in that, The method further includes: The estimation model is constructed based on the following formula (1): The updated model is constructed based on the following formula (2): in, This represents the end time of the driver's sample for the autonomous driving system in the current first time period. Trust rating labels This indicates the end time of the driver's sample for the autonomous driving system in the next first time period. Trust rating labels This represents the first event metric sample. This represents the first experience indicator sample. This represents the first behavioral indicator sample. All of these are state-space parameters to be estimated. Linear transition matrix, Noise term; The combination of the estimated model and the updated model is determined as the initial state space model.

8. The trust estimation method for an autonomous driving system based on a state-space model according to claim 7, characterized in that, The current trust level, the event metric, the experience metric, and the behavior metric are input into the state space model to obtain the driver's trust level in the autonomous driving system in the next time period, as output by the state space model. The current trust level, the event index, and the experience index are input into the estimation model of the state space model corresponding to the driver to obtain the driver's initial trust level of the autonomous driving system in the next time period output by the estimation model. The initial trust level and the behavioral indicators are input into the update model of the state space model corresponding to the driver. The initial trust level is corrected by a Kalman filter based on the behavioral indicators to obtain the driver's trust level in the autonomous driving system in the next time period, which is output by the update model.

9. A trust estimation device for an autonomous driving system based on a state-space model, characterized in that, include: The first acquisition unit is used to acquire event indicators, experience indicators and behavior indicators within the current time period. The event indicators include indicators that affect the driver's trust in the autonomous driving system. The experience indicators include indicators related to the interaction between the driver and the autonomous driving system. The behavior indicators include indicators related to the driver's use of the autonomous driving system. The second acquisition unit is used to acquire the driver's current level of trust in the autonomous driving system during the current time period; An estimation unit is used to input the current trust level, the event index, the experience index, and the behavior index into a state space model to obtain the driver's trust level in the autonomous driving system in the next time period, as output by the state space model. The state space model is trained based on at least one driver sample in different first time periods, including first event index samples, first experience index samples, first behavior index samples, and trust labels.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the autonomous driving system trust estimation method based on the state-space model as described in any one of claims 1 to 8.

Images