Automatic driving automobile reliability analysis method based on section-by-section decision Markov

Abstract

The invention belongs to the technical field of automatic driving automobiles, and particularly relates to an automatic driving automobile reliability analysis method based on section-by-section decision Markov. Comprising the steps that a mixed state space model containing a discrete state space and a continuous state space is constructed, discrete states comprise an automatic driving mode, a manual driving mode and a takeover process mode, and the continuous state comprises a situation awareness level; defining a deterministic evolution mechanism of continuous state variables, and setting different situation awareness level descent parameters for different driving modes to simulate the influence of passive fatigue and active fatigue; a random transfer mechanism between discrete states is defined, and a transfer rate is set as a function of a situation awareness level. The method can accurately describe the influence of various factors such as driver fatigue, the take-over process, hand release reminding and non-driving related tasks on the reliability of the self-driving automobile.

Description

Technical Field

[0001] This invention belongs to the field of autonomous vehicle control technology, specifically relating to a reliability analysis method for autonomous vehicles based on a step-by-step decision Markov model. Background Technology

[0002] A piecewise determination Markov process (PDM) is a continuous-time hybrid stochastic process that combines deterministic evolution with stochastic jump behavior. When the system is in a given state, it initially evolves along a deterministic continuous trajectory, typically governed by ordinary differential equations or equations. Upon reaching a jump point determined by the stochastic jump rate, the process instantaneously jumps, and the state variables jump to a new value according to a given transition kernel. Subsequently, the system enters a new deterministic evolutionary phase until the next stochastic jump occurs. Piecewise determination Markov processes are suitable for modeling complex systems with continuous dynamics but subject to random events.

[0003] In recent years, due to the rapid development of technologies such as artificial intelligence, the Internet of Things, and sensors, autonomous vehicles have flourished and become an indispensable part of people's daily lives. However, current autonomous vehicles still have safety issues, with various autonomous driving safety accidents occurring frequently. Therefore, it is urgent to conduct reliability assessments of autonomous vehicles and analyze potential safety hazards in the system.

[0004] Components in autonomous vehicles (such as perception, decision-making, and planning components) still have a high failure rate under extreme conditions. Furthermore, human factors in autonomous vehicles cannot be ignored; excessive reliance on autonomous driving functions by drivers can also lead to safety accidents.

[0005] When driving manually, drivers are typically affected by active fatigue, leading to decreased driver attention and an increased risk of accidents. While active fatigue is alleviated in autonomous driving, drivers still need to be aware of their surroundings and take over when the system is unable to handle the situation to prevent accidents. Therefore, drivers in autonomous driving mode are primarily affected by passive fatigue. Both passive and active driver fatigue need to be considered in the reliability analysis of autonomous vehicles within autonomous driving human-machine systems.

[0006] The driver takeover process is complex. First, the driver needs to acknowledge the takeover request from the autonomous vehicle and promptly formulate a correct takeover plan. After making the plan, the driver needs to execute it quickly. Furthermore, for a period after the driver takeover is completed, the vehicle remains in an unstable state, thus posing a potential risk of accidents. This complex takeover process must be considered in the reliability analysis of autonomous vehicles.

[0007] Hands-off warning is a crucial feature for enhancing situational awareness in autonomous vehicles. When the system detects that the driver's hands have not been on the steering wheel for a period of time, it issues visual, audible, or textual reminders to prompt the driver to place their hands on the steering wheel, thus improving situational awareness and increasing the safety of the autonomous vehicle. Hands-off warning is an important human-machine interaction feature in autonomous vehicles and cannot be ignored in reliability analysis.

[0008] Furthermore, the impact of non-driving-related tasks on drivers cannot be ignored. Non-driving-related tasks can alleviate passive driver fatigue and improve driver safety. However, when drivers are engaged in non-driving-related tasks, their attention is focused on the task at hand, and their situational awareness is lower. If a takeover request occurs at this time, the driver may not be able to respond in time, potentially leading to a safety accident. Summary of the Invention

[0009] The purpose of this invention is to address the shortcomings of the aforementioned background technology and provide a reliability analysis method for autonomous vehicles based on a segmented decision Markov model.

[0010] The technical solution adopted in this invention is: a reliability analysis method for autonomous vehicles based on piecewise decision Markov, comprising: Step S1: Construct a hybrid state space model of the autonomous vehicle human-machine system. The hybrid state space model consists of a discrete state space and a continuous state space. The discrete state space contains multiple discrete states that characterize the system's operating mode. The discrete states include at least the autonomous driving mode, the manual driving mode, and the takeover process mode. The continuous state space contains at least continuous state variables that characterize the level of situational awareness. Step S2: Define the deterministic evolution mechanism of the continuous state variable; determine the evolution of the continuous state variable over time based on the current discrete state of the system; wherein, for the autonomous driving mode and the manual driving mode, different situational awareness level decline parameters are set respectively to simulate the different attenuation effects of passive fatigue and active fatigue on the driver's situational awareness level. Step S3: Define the random transition mechanism between the discrete states; set the boundary conditions and transition rates that trigger the discrete state switching; wherein at least a portion of the transition rates is defined as a function of the continuous state variables to establish the influence of the driver's situational awareness level on the system state switching probability.

[0011] Preferably, the hybrid state-space model is a piecewise decision Markov process, and its state is defined as a vector. , where X is a discrete state representing different operating modes of the autonomous vehicle's human-machine system; τ represents a parameter in the rate of change of the driver's situational awareness level; s0 represents the situational awareness level before performing non-driving-related tasks; r represents the time the current system is in the takeover request state; s represents the driver's situational awareness level; d represents the time when the mitigating effect of non-driving-related tasks on the decline in situational awareness disappears; t represents the system time.

[0012] Preferably, the range of values ​​for the discrete state X includes: Where A represents the autonomous driving state, HOD represents the autonomous driving state after hands-free warning, NDRT represents the state of non-driving related tasks, AN represents the autonomous driving state after non-driving related tasks, HODN represents the state after non-driving related tasks after hands-free warning, TOR represents the takeover request state, TOP represents the takeover plan state, TOA represents the takeover action state, TOS represents the takeover stabilization state, H represents the manual driving state, AES represents the automatic emergency braking state, and F represents the failure state.

[0013] Preferably, the deterministic evolution mechanism for the continuous state variable defined in step S2 specifically includes: when the discrete state When the driver is affected by passive fatigue, the driver's situational awareness level *s* decreases over time; when the discrete state... When a driver is affected by active fatigue, the driver's situational awareness level *s* decreases over time; the parameter λ represents the decrease in the driver's situational awareness level caused by active fatigue. a The parameter λ represents the decrease in driver situational awareness caused by passive fatigue. p .

[0014] Preferably, the takeover process mode in step S1 is refined into a multi-stage sequence model, including: the zero stage corresponds to the state TOR, which represents the process of the driver capturing the takeover request; the first stage corresponds to the state TOP, which represents the process of the driver formulating a takeover plan; the second stage corresponds to the state TOA, which represents the process of the driver executing the corresponding driving action according to the takeover plan; and the third stage corresponds to the state TOS, which represents the process of the vehicle remaining in an unstable state after the driver completes the driving action, until the vehicle returns to stability or an accident occurs.

[0015] Preferably, the random transition mechanism between discrete states defined in step S3 includes a hands-off warning mechanism: a hands-off warning trigger threshold t1 is set, and an alarm is triggered when the system detects that the driver's hands have been off the steering wheel for more than t1; the alarm trigger is defined as a forced jump event, which instantly raises the driver's situational awareness level s to a preset high value L.

[0016] Preferably, the method further includes modeling the dual impact of non-driving-related tasks: when the discrete state X = NDRT, the driver's attention is focused on non-driving-related tasks, and the level of situational awareness regarding driving tasks is in a low range; when the discrete state X = AN, non-driving-related tasks have the effect of alleviating passive fatigue, and the driver's situational awareness level decreases by a parameter λ. e And satisfying λ e <λ p Until the time variable d returns to zero, the system state jumps back to A, and the driver's situational awareness level recovers to λ. p The rate decreases.

[0017] Preferably, the time variable d is a time window initialized when the system transitions from state HODN to AN. This window characterizes the duration of the mitigating effect of non-driving-related tasks on the decline in the driver's situational awareness level; within this window, the decay rate of the driver's situational awareness level remains at the relatively small λ. e .

[0018] Preferably, determining the state transition rate in step S3 specifically includes: for discrete states related to the takeover process. Its state transition rate is not constant, but is constructed as a function of the current situational awareness level s to simulate the dynamic impact of the driver's situational awareness level on the takeover success rate.

[0019] Preferably, when At that time, the system was transferred to The rate is ,in For the driver The status successfully captured the initial rate of the takeover request. For parameters; when At that time, the system was transferred to The rate is , For a constant, when At that time, the rates of human error occurring in the system were respectively , ,in The system is respectively in and The initial rate at which the state occurs due to human error. It is a constant.

[0020] Preferably, step S3 further includes: determining the remaining rate; the remaining rate is determined based on the characteristics of the driving environment and the reliability parameters of the autonomous driving system, and is used to characterize the random state transition rate of the system excluding driver factors.

[0021] Compared with the prior art, the beneficial effects of the present invention are: This invention employs a stepwise decision Markov process to construct a human-machine system model for autonomous vehicles, which can simultaneously characterize the deterministic continuous evolution and random jump behavior of the system. Compared with the traditional pure discrete state Markov model, it more accurately describes the continuous change process of the driver's situational awareness level and improves the accuracy of reliability analysis.

[0022] This invention distinguishes between the different effects of passive fatigue and active fatigue on the driver's situational awareness level, and establishes attenuation models of situational awareness level in autonomous driving mode and manual driving mode respectively, so that reliability analysis can reflect the differences in driver cognitive load under different driving modes.

[0023] This invention refines the takeover process into a sequence model of four stages: takeover request, takeover plan, takeover action, and takeover stabilization. It fully depicts the entire process from the driver receiving the takeover request to the vehicle regaining stability, and can more accurately assess the failure risk at each stage.

[0024] This invention models the state transition rate as a function of the situational awareness level, establishes a dynamic coupling relationship between the driver's cognitive state and the system state switching probability, and can reflect the real-time impact of driver fatigue on takeover success rate.

[0025] This invention models the hands-free warning mechanism, enabling quantitative analysis of its effect on improving the driver's situational awareness, and providing a reference for the human-machine interaction design of autonomous vehicles.

[0026] This invention models the dual impact of non-driving-related tasks, taking into account both the negative impact of reduced situational awareness caused by non-driving-related tasks and their positive effect of alleviating passive fatigue, making the reliability analysis results more comprehensive and objective. Attached Figure Description

[0027] Figure 1 This is a flowchart illustrating a method for reliability analysis of autonomous vehicles based on a segmented decision Markov model, according to an embodiment of the present invention. Figure 2 This is a diagram illustrating the changes in the driver's situational awareness. Figure 3 Diagram of the driver takeover process; Figure 4A diagram illustrating the changes in situational awareness under the influence of the handover warning; Figure 5 This diagram illustrates the changes in situational awareness under the influence of non-driving-related tasks. Figure 6 A reliability model diagram of the Markov process for determining autonomous vehicles segment by segment. Detailed Implementation

[0028] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings. It should be noted that these descriptions are for the purpose of aiding understanding the present invention, but do not constitute a limitation thereof. Furthermore, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0029] like Figure 1 , Figure 6 As shown, this application provides a method for reliability analysis of autonomous vehicles based on a segmented decision Markov model, including: Step S1: Construct a hybrid state space model of the autonomous vehicle human-machine system. The hybrid state space model consists of a discrete state space and a continuous state space. The discrete state space contains multiple discrete states that characterize the system's operating mode. The discrete states include at least the autonomous driving mode, the manual driving mode, and the takeover process mode. The continuous state space contains at least continuous state variables that characterize the level of situational awareness. Step S2: Define the deterministic evolution mechanism of the continuous state variable; determine the evolution of the continuous state variable over time based on the current discrete state of the system; wherein, for the autonomous driving mode and the manual driving mode, different situational awareness level decline parameters are set respectively to simulate the different attenuation effects of passive fatigue and active fatigue on the driver's situational awareness level. Step S3: Define a random transition mechanism between discrete states; set boundary conditions and transition rates to trigger discrete state switching; wherein at least a portion of the transition rates are defined as functions of the continuous state variables to establish the influence of the driver's situational awareness level on the probability of system state switching. The hybrid state-space model can be used to calculate the probability that the system will enter a failure state within a predetermined time range, serving as an indicator for evaluating the reliability of autonomous vehicles.

[0030] In one embodiment, the hybrid state-space model is a piecewise decision Markov process, where the state is defined as a vector. , where X is a discrete state representing different operating modes of the autonomous vehicle's human-machine system; τ represents a parameter in the rate of change of the driver's situational awareness level; s0 represents the situational awareness level before performing non-driving-related tasks; r represents the time the current system is in the takeover request state; s represents the driver's situational awareness level; d represents the time when the mitigating effect of non-driving-related tasks on the decline in situational awareness disappears; t represents the system time.

[0031] In one embodiment, the range of values ​​for the discrete state X includes: Where A represents the autonomous driving state, HOD represents the autonomous driving state after hands-free warning, NDRT represents the state of non-driving related tasks, AN represents the autonomous driving state after non-driving related tasks, HODN represents the state after non-driving related tasks after hands-free warning, TOR represents the takeover request state, TOP represents the takeover plan state, TOA represents the takeover action state, TOS represents the takeover stabilization state, H represents the manual driving state, AES represents the automatic emergency braking state, and F represents the failure state.

[0032] In one embodiment, defining the deterministic evolution mechanism of the continuous state variable in step S2 specifically includes: when the discrete state When the driver is affected by passive fatigue, the driver's situational awareness level *s* decreases over time; when the discrete state... When a driver is actively fatigued, their situational awareness level (s) decreases over time. Due to fatigue, the driver's situational awareness level declines with driving duration, such as... Figure 2 As shown. The parameter λ represents the decrease in driver situational awareness caused by active fatigue. a The parameter λ represents the decrease in driver situational awareness caused by passive fatigue. p .

[0033] In one embodiment, the driver takeover process is as follows: Figure 3 As shown, the takeover process mode in step S1 is refined into a multi-stage sequence model, including: the zero stage corresponds to the state TOR, which represents the process of the driver capturing the takeover request; the first stage corresponds to the state TOP, which represents the process of the driver formulating a takeover plan; the second stage corresponds to the state TOA, which represents the process of the driver executing the corresponding driving action according to the takeover plan; and the third stage corresponds to the state TOS, which represents the process of the vehicle remaining in an unstable state after the driver completes the driving action, until the vehicle returns to stability or an accident occurs.

[0034] In one embodiment, the random transition mechanism between discrete states defined in step S3 includes a hands-off alert mechanism: the driver's situational awareness changes under the influence of the hands-off alert system as follows: Figure 4As shown, a hands-off warning trigger threshold t1 is set. When the system detects that the driver's hands have been off the steering wheel for more than t1, an alarm is triggered. This alarm trigger is defined as a forced jump event, which instantly raises the driver's situational awareness level s to a preset high value L. After a period of time t2-t1, the driver's hands are off the steering wheel again, the hands-off warning effect disappears, and the driver's situational awareness level continues to decline.

[0035] In one embodiment, the method further includes modeling the dual impact of non-driving-related tasks: when the discrete state X = NDRT, the driver's attention is focused on non-driving-related tasks, and the level of situational awareness regarding driving tasks is in a low range; when the discrete state X = AN, non-driving-related tasks alleviate passive fatigue, and the driver's situational awareness level decreases by a parameter λ. e And satisfying λ e <λ p Until the time variable d returns to zero, the system state jumps back to A, and the driver's situational awareness level recovers to λ. p The rate of decrease is observed. In autonomous driving mode, appropriate non-driving-related tasks can alleviate passive fatigue and offset the decline in situational awareness. However, when performing non-driving-related tasks, the driver's attention is focused on the non-driving-related task, the driver's situational awareness decreases, and the ability to respond to takeover requests weakens. The process of change in the driver's situational awareness level under the influence of non-driving-related tasks is as follows: Figure 5 As shown, at time At this time, the driver begins to perform non-driving related tasks. During this period, the driver's situational awareness is focused on non-driving related tasks, and the level of situational awareness of driving related tasks is relatively low. The execution of non-driving related tasks requires a certain amount of time. At that moment After completing the non-driving-related tasks, the driver resumes monitoring the autonomous vehicle's operation, and the level of situational awareness begins to decline. Since non-driving-related tasks can mitigate the decline in situational awareness caused by passive fatigue, the rate of decline in situational awareness after non-driving-related tasks is smaller than the rate of decline before non-driving-related tasks. After a period of time Afterward, the effect of non-driving-related tasks in alleviating passive fatigue disappeared, and the driver's situational awareness level declined again at the original rate.

[0036] In one embodiment, the time variable d is a time window initialized when the system transitions from state HODN to AN. This window characterizes the duration of the mitigating effect of non-driving-related tasks on the decline in driver situational awareness; within this window, the rate of decay of driver situational awareness remains at the relatively small λ. e .

[0037] In one embodiment, determining the rate of state transition in step S3 specifically includes: for discrete states related to the takeover process. Its state transition rate is not constant, but is constructed as a function of the current situational awareness level s to simulate the dynamic impact of the driver's situational awareness level on the takeover success rate.

[0038] In one embodiment, when At that time, the system was transferred to The rate is ,in For the driver The status successfully captured the initial rate of the takeover request. For parameters; when At that time, the system was transferred to The rate is , For a constant, when At that time, the rates of human error occurring in the system were respectively , ,in The system is respectively in and The initial rate at which the state occurs due to human error. It is a constant.

[0039] In one embodiment, step S3 further includes: determining the remaining rate; the remaining rate is determined based on the characteristics of the driving environment and the reliability parameters of the autonomous driving system, and is used to characterize the system's random state transition rate excluding driver factors.

[0040] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Contents not described in detail in this specification belong to prior art known to those skilled in the art.

Claims

1. A reliability analysis method for autonomous vehicles based on piecewise decision Markov, characterized in that: include: Step S1: Construct a hybrid state space model of the autonomous vehicle human-machine system. The hybrid state space model consists of a discrete state space and a continuous state space. The discrete state space contains multiple discrete states that characterize the system's operating mode. The discrete states include at least the autonomous driving mode, the manual driving mode, and the takeover process mode. The continuous state space contains at least continuous state variables that characterize the level of situational awareness. Step S2: Define the deterministic evolution mechanism of the continuous state variable; determine the evolution of the continuous state variable over time based on the current discrete state of the system; wherein, for the autonomous driving mode and the manual driving mode, different situational awareness level decline parameters are set respectively to simulate the different attenuation effects of passive fatigue and active fatigue on the driver's situational awareness level. Step S3: Define the random transition mechanism between the discrete states; set the boundary conditions and transition rates that trigger the discrete state switching; wherein at least a portion of the transition rates is defined as a function of the continuous state variables to establish the influence of the driver's situational awareness level on the system state switching probability.

2. The autonomous vehicle reliability analysis method based on a piecewise decision Markov process according to claim 1, characterized in that, The hybrid state-space model is a piecewise determination Markov process, where the state is defined as a vector. , where X is a discrete state representing different operating modes of the autonomous vehicle's human-machine system; τ represents a parameter in the rate of change of the driver's situational awareness level; s0 represents the situational awareness level before performing non-driving-related tasks; r represents the time the current system is in the takeover request state; s represents the driver's situational awareness level; d represents the time when the mitigating effect of non-driving-related tasks on the decline in situational awareness disappears; t represents the system time.

3. The autonomous vehicle reliability analysis method based on a piecewise decision Markov process according to claim 2, characterized in that, The range of values ​​for the discrete state X includes: Where A represents the autonomous driving state, HOD represents the autonomous driving state after hands-free warning, NDRT represents the state of non-driving related tasks, AN represents the autonomous driving state after non-driving related tasks, HODN represents the state after non-driving related tasks after hands-free warning, TOR represents the takeover request state, TOP represents the takeover plan state, TOA represents the takeover action state, TOS represents the takeover stabilization state, H represents the manual driving state, AES represents the automatic emergency braking state, and F represents the failure state.

4. The autonomous vehicle reliability analysis method based on a piecewise decision Markov process according to claim 3, characterized in that, The deterministic evolution mechanism for the continuous state variable defined in step S2 specifically includes: when the discrete state When the driver is affected by passive fatigue, the driver's situational awareness level *s* decreases over time; when the discrete state... When a driver is affected by active fatigue, the driver's situational awareness level *s* decreases over time; the parameter λ represents the decrease in the driver's situational awareness level caused by active fatigue. a The parameter λ represents the decrease in driver situational awareness caused by passive fatigue. p .

5. The method for reliability analysis of autonomous vehicles based on a piecewise decision Markov process according to claim 3, characterized in that, The takeover process mode in step S1 is refined into a multi-stage sequence model, including: the zero stage corresponds to the state TOR, which represents the process of the driver capturing the takeover request; the first stage corresponds to the state TOP, which represents the process of the driver formulating a takeover plan; the second stage corresponds to the state TOA, which represents the process of the driver executing the corresponding driving action according to the takeover plan; and the third stage corresponds to the state TOS, which represents the process of the vehicle remaining in an unstable state after the driver completes the driving action, until the vehicle returns to stability or an accident occurs.

6. The method for reliability analysis of autonomous vehicles based on a piecewise decision Markov process according to claim 3, characterized in that, The random transition mechanism between discrete states defined in step S3 includes a hands-off warning mechanism: a hands-off warning trigger threshold t1 is set, and an alarm is triggered when the system detects that the driver's hands have been off the steering wheel for more than t1; the alarm trigger is defined as a forced jump event, which instantly raises the driver's situational awareness level s to a preset high value L.

7. The method for reliability analysis of autonomous vehicles based on a piecewise decision Markov process according to claim 3, characterized in that, The method also includes modeling the dual impact of non-driving-related tasks: when the discrete state X=NDRT, the driver's attention is focused on non-driving-related tasks, and the level of situational awareness of driving tasks is in the low range. When the discrete state X=AN, non-driving-related tasks have the effect of alleviating passive fatigue, and the parameter for the decrease in the driver's situational awareness level is λ. e And satisfying λ e <λ p Until the time variable d returns to zero, the system state jumps back to A, and the driver's situational awareness level recovers to λ. p The rate decreases.

8. The method for reliability analysis of autonomous vehicles based on a piecewise decision Markov process according to claim 7, characterized in that, The time variable d is the time window initialized when the system transitions from state HODN to AN. This window characterizes the duration of the mitigating effect of non-driving-related tasks on the decline in the driver's situational awareness level. Within this window, the rate of decay of the driver's situational awareness level remains at the relatively small λ. e .

9. The method for reliability analysis of autonomous vehicles based on a piecewise decision Markov process according to claim 3, characterized in that, Determining the state transition rate in step S3 specifically includes: for the discrete states related to the takeover process. Its state transition rate is not constant, but is constructed as a function of the current situational awareness level s to simulate the dynamic impact of the driver's situational awareness level on the takeover success rate.

10. The method for reliability analysis of autonomous vehicles based on a piecewise decision Markov process according to claim 1, characterized in that, Step S3 further includes: determining the remaining rate; the remaining rate is determined based on the characteristics of the driving environment and the reliability parameters of the autonomous driving system, and is used to characterize the random state transition rate of the system excluding driver factors.

Images