Driver fatigue state estimation method and system based on vehicle motion

Abstract

The invention discloses a driver fatigue state estimation method and system based on vehicle motion, and the method comprises the following steps: obtaining energy consumption historical data caused by different operation actions of a driver in a driving task, and constructing an energy consumption model; constructing and training a fatigue accumulation model based on fuzzy control by combining a fuzzy control method based on the energy consumption model and the GMM model; and obtaining energy consumption data of the driver by using the fatigue accumulation model, and obtaining a fatigue state of the driver based on the fatigue accumulation model. According to the method, a modeling method based on energy consumption accumulation is adopted, the energy consumption per minute of hand operation (steering wheel steering) and foot operation (acceleration and braking) is modeled, and a fatigue accumulation model is constructed through a fuzzy control method, so that the fatigue state of a driver is dynamically evaluated.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent transportation technology, and in particular relates to a method and system for estimating a driver's fatigue state based on vehicle motion. Background Art

[0002] Road traffic accidents pose a significant threat to life and safety, resulting in numerous casualties and substantial property damage. Driver fatigue is a key contributing factor. While intelligent driving has been a reality, due to limited attention paid to intelligent cockpits, no suitable assessment methods for driver fatigue have been established.

[0003] Currently, existing driver fatigue detection or evaluation technologies rely on external behavioral characteristics. For example, CN202211498557.3 uses facial recognition via surveillance cameras to detect driver fatigue in real time. However, this detection method may fail in scenes with suboptimal lighting conditions or occlusion. Others rely on physiological signals, such as CN202211300753.5, which uses a time-frequency attention mechanism and an adaptive feature fusion module to fully explore and capture key features related to driver fatigue and identify driver fatigue. However, using physiological signals as a basis makes the device complex and inefficient. To be convenient and effective, driver fatigue evaluation still requires a method to predict the changing process of driver fatigue through vehicle motion behavior. Summary of the Invention

[0004] The present invention aims to solve the deficiencies of the prior art and provides the following solutions:

[0005] A method for estimating a driver's fatigue state based on vehicle motion comprises the following steps:

[0006] Obtain historical data on the energy consumption caused by different operating actions of the driver during the driving task and build an energy consumption model;

[0007] Based on the energy consumption model and the GMM model, combined with the fuzzy control method, a fatigue accumulation model based on fuzzy control is constructed and trained;

[0008] The energy consumption data of the driver using the fatigue accumulation model is obtained, and the fatigue state of the driver is obtained based on the fatigue accumulation model.

[0009] Preferably, the energy consumption history data includes: upper limb energy consumption history data and lower limb energy consumption history data; the energy consumption model includes: upper limb movement energy consumption model and lower limb movement energy consumption model.

[0010] Preferably, the method for constructing the upper limb exercise energy consumption model includes:

[0011] Get the steering torque and steering wheel angle rate of the steering wheel operation:

[0012]

[0013] Among them, τ steer represents steering torque, m represents vehicle mass, r represents kingpin offset, G represents steering gear ratio, represents the steering wheel angular rate, L represents the wheelbase, v(t) represents the vehicle speed, t represents the time, R(t) represents the turning radius, and δ(t) represents the wheel angle;

[0014] Based on the steering torque and the steering wheel angular rate, the upper limb movement energy consumption model is constructed:

[0015]

[0016] Among them, a lat (τ) represents the lateral acceleration, v(τ) represents the absolute velocity, t0 represents the initial time, C hand Represents the proportional coefficient of the upper limb movement energy consumption model.

[0017] Preferably, the method for constructing the lower limb exercise energy consumption model includes:

[0018] Get the actual pedal force:

[0019]

[0020] Among them, F foot (t) represents the actual pedal force, a(t) represents the longitudinal acceleration, F max Indicates the maximum pedal force, a max Indicates the maximum acceleration;

[0021] Get the actual pedal stroke:

[0022]

[0023] Among them, d foot (t) represents the actual pedal stroke, d max Indicates the maximum pedal travel;

[0024] Based on the actual pedal force and the actual pedal stroke, the lower limb exercise energy consumption model is constructed:

[0025]

[0026] Where a(τ) represents the longitudinal acceleration, C foot Represents the proportional coefficient of the lower limb movement energy consumption model.

[0027] Preferably, the method for constructing the fatigue accumulation model includes:

[0028] Based on the energy consumption model and in combination with the fuzzy control method, an initial fatigue accumulation model based on fuzzy control is constructed;

[0029] Constructing the GMM model and training the GMM model using an expectation-maximization algorithm to obtain a trained GMM model;

[0030] A membership function is constructed based on the trained GMM model, and a rule base of the initial fatigue accumulation model is established based on the membership function to obtain the fatigue accumulation model.

[0031] Preferably, the membership function includes: an upper limb high energy consumption membership function and a lower limb low energy consumption membership function;

[0032] The upper limb high energy consumption membership function is:

[0033]

[0034] Among them, μ high (E hand ) represents the upper limb high energy consumption membership function, μ3 represents the high energy consumption level, and k represents the proportional coefficient;

[0035] The lower limb low energy consumption membership function is:

[0036]

[0037] Among them, μ low (E foot ) represents the low energy consumption membership function of the lower limbs, μ1 represents the low energy consumption level, and σ1 represents the standard deviation of the low energy consumption level.

[0038] The present invention also provides a driver fatigue state estimation system based on vehicle motion, the system applying any of the above methods, comprising: an energy consumption model construction module, a fatigue accumulation model construction module and a fatigue state estimation module;

[0039] The energy consumption model building module is used to obtain historical data on energy consumption caused by different operating actions of the driver during the driving task and build an energy consumption model;

[0040] The fatigue accumulation model construction module is based on the energy consumption model and the GMM model, combined with the fuzzy control method, to construct and train a fatigue accumulation model based on fuzzy control;

[0041] The fatigue state estimation module is used to obtain the driver's energy consumption data using the fatigue accumulation model, and obtain the driver's fatigue state based on the fatigue accumulation model.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] Through the study of driving fatigue levels and transition processes, the present invention found that the generation of driving fatigue is the cumulative result of the driver's energy consumption, and different driving operations (such as steering wheel turning, accelerator and brake pedal use) will lead to the continuous consumption of the driver's energy. When the driving task continues without proper recovery, the degree of fatigue will gradually deepen, affecting the driver's operating ability. In order to quantitatively describe this process, the present invention adopts a modeling method based on energy consumption accumulation, and models the energy consumption per minute of hand operation (steering wheel steering) and foot operation (acceleration and braking) respectively, and constructs a fatigue accumulation model through a fuzzy control method to achieve a dynamic assessment of the driver's fatigue state. The present invention can more reasonably simulate the driver's fatigue evolution process and successfully reflect the impact of fatigue status on driving behavior, including features such as prolonged reaction time, slow speed adjustment, and steady acceleration changes. The present invention can add behaviors caused by driving fatigue to traffic vehicles in simulation scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0045] Figure 1 Schematic diagram of a method flow in an embodiment of the present invention;

[0046] Figure 2 Schematic diagram of the fatigue accumulation model structure of an embodiment of the present invention. DETAILED DESCRIPTION

[0047] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0048] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0049] Example 1

[0050] In this embodiment, if Figure 1As shown, a method for estimating a driver's fatigue state based on vehicle motion includes the following steps:

[0051] S1. Obtain historical data on energy consumption caused by different operating actions of the driver during the driving task and build an energy consumption model.

[0052] The energy consumption history data includes: upper limb energy consumption history data and lower limb energy consumption history data; the energy consumption model includes: upper limb movement energy consumption model and lower limb movement energy consumption model.

[0053] In this embodiment, an energy consumption model caused by different operating actions of the driver during the driving task (such as steering wheel turning, accelerator and brake pedal use) is established to quantify the impact of different driving operations on physical fitness over a period of time.

[0054] Fatigue Accumulation Model Construction: Based on the cumulative pattern of energy consumption and incorporating fuzzy control methods, the evolution of the driver's fatigue state is inferred. Different fatigue accumulation rates are generated according to different minute-by-minute energy consumption levels, and fatigue values ​​are updated in real time, thereby enabling dynamic assessment of driver fatigue transitions. When analyzing the driver's energy consumption during driving tasks, this paper divides the driver's operations into two main parts: hand operations and foot operations. This classification not only helps to gain a deeper understanding of the driver's energy consumption characteristics during driving from different perspectives, but also provides a clear theoretical basis for subsequent determination of minute-by-minute energy consumption levels.

[0055] As for hand energy consumption, it mainly comes from turning the steering wheel.

[0056] Methods for constructing upper limb exercise energy consumption models include:

[0057] Steering wheel operation energy consumption comes from overcoming the steering torque τ steer (τ) and steering wheel angle satisfy:

[0058]

[0059] Establishing steering torque τ steer (τ) and steering wheel angle Relationship with vehicle lateral motion: lateral acceleration a lat (t) is generated by the lateral force of the tire and is related to the steering wheel torque. According to vehicle dynamics, the steering torque and steering wheel angle rate of the steering wheel operation are obtained:

[0060]

[0061]

[0062] Among them, τsteer represents steering torque, m represents vehicle mass, r represents kingpin offset (mechanical lever ratio), G represents steering gear ratio, represents the steering wheel angular rate, L represents the wheelbase, v(t) represents the vehicle speed, t represents the time, R(t) represents the turning radius, and δ(t) represents the wheel angle;

[0063] The steering wheel angle and wheel angle satisfy the following relationship:

[0064]

[0065] The relationship between the steering wheel angle and the lateral acceleration is obtained by combining:

[0066]

[0067] The steering wheel angle rate is:

[0068]

[0069] in, represents the rate of change of vehicle speed; if the rate of change of vehicle speed is small (steady-state steering), the high-order terms can be ignored and the approximation is:

[0070]

[0071] The steering torque τ steer (τ) and steering wheel angle Substitute into the energy formula to construct the upper limb exercise energy consumption model:

[0072]

[0073] Among them, a lat (τ) represents the lateral acceleration, v(τ) represents the absolute velocity, t0 represents the initial time, C hand Represents the proportional coefficient of the upper limb movement energy consumption model.

[0074] The energy consumption of the foot mainly comes from the force and displacement overcome when stepping on the accelerator or brake pedal. Assumptions: 1. Pedal force F foot (t) is proportional to the vehicle longitudinal acceleration a(t); 2. Pedal travel d foot (t) is linearly related to acceleration a(t); 3. Energy consumption is the time integral of the product of force and displacement. The methods for constructing the lower limb movement energy consumption model include:

[0075] The longitudinal acceleration is driven by the pedal force according to Newton's second law:

[0076] F foot (t) = m·a(t)

[0077] However, the actual pedal force must be amplified by the transmission system, and the proportional coefficient k is introduced. foot :

[0078]

[0079] Based on the pedal force and the proportional coefficient, get the actual pedal force:

[0080]

[0081] Among them, F foot (t) represents the actual pedal force, a(t) represents the longitudinal acceleration, F max Indicates the maximum pedal force, a max Indicates the maximum acceleration;

[0082] Since the displacement is linearly related to the acceleration, the actual pedal stroke is obtained:

[0083]

[0084] Among them, d foot (t) represents the actual pedal stroke, d max Indicates the maximum pedal travel;

[0085] Based on the actual pedal force and actual pedal stroke, a lower limb exercise energy consumption model is constructed:

[0086]

[0087]

[0088] Where a(τ) represents the longitudinal acceleration, C foot Represents the proportional coefficient of the lower limb movement energy consumption model.

[0089] From the formulas of the two models, it can be seen that in longitudinal operation, sudden acceleration or braking (high |a|) will rapidly increase the driver's energy consumption, and in lateral operation, frequent and large-scale steering (high a|) will increase the driver's energy consumption. lat ) will increase the driver's fatigue; due to the existence of points, energy consumption is a cumulative process. Even if the driving is relatively smooth, there will still be significant physical energy consumption after a long time of driving. 2 Appearing in the denominator, when the vehicle speed is high, the energy required for steering wheel adjustment is relatively reduced, which is consistent with real-world driving experience (when driving at high speed, fine-tuning the direction has less impact on the vehicle); therefore, these two formulas can be used to evaluate the driver's energy consumption status in real time, and then infer the driver's fatigue level. If real-time statistics are required, energy consumption per minute can be used to describe the energy consumption during this period.

[0090] S2. Based on the energy consumption model and GMM model, combined with the fuzzy control method, a fatigue accumulation model based on fuzzy control is constructed and trained.

[0091] Methods for constructing fatigue accumulation models include:

[0092] S2.1 Based on the energy consumption model and combined with the fuzzy control method, an initial fatigue accumulation model based on fuzzy control is constructed.

[0093] In this embodiment, fuzzy control theory is an important control method widely used in engineering fields, particularly suitable for handling nonlinear, uncertain, and fuzzy problems in complex systems. The core concept of fuzzy control is derived from fuzzy set theory and the simulation of human logical reasoning process, achieving effective control of the target system through fuzzy logic reasoning. In recent years, with the development of computer technology and artificial intelligence, fuzzy control methods have been continuously enriched and improved, and have been successfully applied in many fields such as automation, robotics, aerospace, intelligent transportation, and process control.

[0094] The classification of fuzzy control methods has gradually become more refined, currently falling into three main categories: rule-based fuzzy control, model-based fuzzy control, and learning-based fuzzy control. A review of existing technologies indicates that, since the research subject of this invention is human driving behavior, designing rules based on human experience can better reflect human states. Therefore, rule-based fuzzy control methods can more effectively reflect the changing characteristics of a driver's state.

[0095] Rule-based fuzzy control is a classic approach in the fuzzy control system, with the most famous example being the fuzzy controller. This control method, based on expert knowledge, establishes a fuzzy rule base by summarizing the relationships between system inputs and outputs. Each rule typically uses an "if...then..." format to express the fuzzy mapping between inputs and outputs. These rules are derived using fuzzy inference mechanisms to generate control decisions for the system.

[0096] Fatigue itself is a vague concept, difficult to accurately describe through precise mathematical analytical models. Simply categorizing a driver's state into normal and fatigued states is insufficient to reflect actual driving conditions. To more scientifically and rationally describe the degree of driver fatigue, this embodiment subdivides the driver's physiological state into four fatigue levels. This classification not only conforms to the natural law of gradual accumulation of driver fatigue, but also more specifically reflects the changes in the driver's fatigue state.

[0097] Given that the concept of fatigue itself is ambiguous and difficult to describe using traditional analytical models, fuzzy reasoning can effectively transform human experience into automatic control strategies and is suitable for objects that are difficult to establish clear mathematical models. Therefore, this embodiment adopts a fuzzy reasoning system to achieve intelligent decision-making through human experience. The driver fatigue generation fuzzy controller designed in this embodiment adopts a two-input single-output structure, in which the two input variables are the energy consumption of hand operation per minute and the energy consumption of foot operation per minute, and the output is the driver's fatigue level. The specific structure diagram of the fatigue accumulation model based on fuzzy control is shown in the figure below. Figure 2 shown.

[0098] In this model, this embodiment first calculates the driver's minute-by-minute energy expenditure for both hands and feet and assesses their minute-by-minute energy expenditure level based on predefined rules. Subsequently, a classification method is used to categorize minute-by-minute energy expenditure into different levels, and fatigue values ​​are incremented accordingly. Within each time period, the accumulated fatigue values ​​are further mapped to specific fatigue levels, dynamically reflecting the driver's fatigue evolution.

[0099] S2.2 constructs a GMM model and uses the expectation maximization algorithm to train the GMM model to obtain a trained GMM model.

[0100] The key to constructing a membership function lies in setting appropriate classification criteria and establishing a suitable curve to describe the membership. While membership functions are typically constructed through empirical design, this project considers using statistical methods, selecting an appropriate data set, and constructing it through analysis of the distribution of minute-by-minute energy expenditure.

[0101] This example focuses on vehicle speed and lateral and longitudinal acceleration. The vehicle's lateral and longitudinal accelerations can be derived and calculated based on the changes in the vehicle's position coordinates over time. Specifically, longitudinal acceleration can be obtained by the rate of change of velocity between adjacent moments, while lateral acceleration requires calculation based on the vehicle's motion along the lateral coordinates. This acceleration information plays a key role in analyzing driving behavior characteristics.

[0102] For foot energy, the frequency of energy consumption per minute decreases as consumption increases. This distribution generally indicates that the majority of samples are concentrated in the lower energy consumption range, while the frequency decreases as energy consumption increases, consistent with the low frequency of high-energy-consuming actions by drivers. The frequency distribution shows that the data can be divided into two distinct regions and a subtle growth peak: these three areas can be considered the three energy consumption zones. The frequency of the low-energy consumption zone is higher, and as consumption increases, the frequency decreases.

[0103] This distribution can be modeled using a Gaussian Mixture Model (GMM), which assumes that the data comes from multiple Gaussian distributions, each representing a low, medium, or high energy expenditure class. Similarly, hand consumption rates can also be classified using a GMM.

[0104] In a GMM, the samples in the dataset are assumed to follow a mixture of Gaussians, i.e., composed of multiple independent Gaussian distributions, each corresponding to a data cluster. By taking a weighted sum of a finite number of Gaussian probability density functions, the distribution of the sample dataset can be reasonably estimated, allowing the GMM to capture the complex structure of the data and depict the underlying patterns.

[0105] As a soft clustering method, GMM differs from hard clustering methods like K-means, which allows data points to belong to multiple clusters with certain probabilities, rather than a single category. During training, GMM determines the probability of each data point belonging to each Gaussian distribution using Maximum Likelihood Estimation (MLE) and assigns samples based on these probabilities. The training process typically uses the Expectation-Maximization (EM) algorithm, which iteratively optimizes the mean, covariance, and weight coefficients of each Gaussian distribution until it converges to the optimal solution. By superimposing multiple Gaussian distributions, GMM can approximate any complex probability distribution, making it valuable in tasks such as nonlinear data modeling, anomaly detection, speech recognition, and image segmentation.

[0106] The core idea of GMM is to assume that the data points x in the dataset are generated by a mixture of K different Gaussian distributions, each representing a category in the data. Mathematically, this can be expressed as follows:

[0107]

[0108] where K represents the number of Gaussian distributions, α k represents the weight of the kth Gaussian distribution, satisfying:

[0109] represents the probability density function of the kth Gaussian distribution with mean μ k and covariance matrix Σ k , which can be expressed as follows:

[0110]

[0111] where d represents the dimension of the data, and |∑ k | represents the determinant of the covariance matrix.

[0112] GMM training usually uses the Expectation Maximization (EM) algorithm, which solves the model parameters through iterative optimization to maximize the likelihood function of the data. The main steps of the EM algorithm are as follows:

[0113] 1. Initialization: Randomly initialize the mean μ of each Gaussian distribution k , covariance Σ k and weight α k .

[0114] 2. Step E (expectation step): Calculate the posterior probability of each data point belonging to each Gaussian distribution, that is, calculate the possibility of the data point belonging to each category:

[0115]

[0116] Among them, γ ik Indicates possibility;

[0117] 3.M step (maximization step): Based on the calculation results of the E step, update the parameters of GMM:

[0118] Mean update:

[0119]

[0120] Covariance update:

[0121]

[0122] Weight update:

[0123]

[0124] 4. Repeated iteration: Repeat the E-step and M-step continuously until the parameters converge, that is, the likelihood function no longer changes significantly. GMM is suitable for fitting multimodal data and can effectively model probability distributions with multiple peaks. Unlike hard clustering methods such as K-means, GMM allows data points to belong to multiple categories and classify them according to probability. GMM describes the statistical characteristics of data through parameters such as mean and covariance matrix, and can be used for tasks such as data generation, classification, and dimensionality reduction. In summary, GMM is a flexible and powerful probabilistic model that can play an important role in data analysis and pattern recognition tasks. Its EM training method makes GMM highly applicable in large-scale data distribution modeling and complex pattern recognition.

[0125] S2.3 constructs a membership function based on the trained GMM model, and establishes a rule base of the initial fatigue accumulation model based on the membership function to obtain the fatigue accumulation model.

[0126] In this embodiment, the Gaussian membership function is used to represent the membership of the clustered sets. The main advantage of using the Gaussian membership function is that it can naturally adapt to the distribution of data, especially data with a normal distribution or a bell curve. It uses a smooth curve and adjustable parameters (such as standard deviation) to represent the membership of the fuzzy set, which can effectively handle uncertainty and ambiguity:

[0127]

[0128] The central values ​​of these clusters are calculated, and then classified according to the size of the values ​​into three levels: low, medium and high, including: 1. Low energy consumption: mean μ1 = 200J, standard deviation σ1 = 50J; 2. Medium energy consumption: mean μ2 = 600J, standard deviation σ2 = 150J; 3. High energy consumption: mean μ3 = 1200J, standard deviation σ3 = 300J.

[0129] The membership functions include: upper limb high energy consumption membership function and lower limb low energy consumption membership function;

[0130] The membership function of upper limb high energy consumption is:

[0131]

[0132] Among them, μ high (E hand ) represents the upper limb high energy consumption membership function, μ3 represents the high energy consumption level, and k represents the proportional coefficient;

[0133] The membership function of low energy consumption of lower limbs is:

[0134]

[0135] Among them, μ low (E foot ) represents the low energy consumption membership function of the lower limbs, μ1 represents the low energy consumption level, and σ1 represents the standard deviation of the low energy consumption level.

[0136] After calculating the membership function, the rules of fuzzy control are defined, as shown in Table 1.

[0137] Table 1

[0138]

[0139] This set of fuzzy control rules describes the impact of F_state (foot energy expenditure per minute) and H_state (hand energy expenditure per minute) on E_level (overall energy expenditure per minute). Higher foot energy expenditure (F_state) is associated with higher overall energy expenditure (E_level), indicating that foot movement significantly impacts overall energy expenditure. Increasing hand energy expenditure (H_state) also increases overall energy expenditure (E_level) to a certain extent, but the magnitude of this impact depends on F_state.

[0140] In general, foot energy expenditure (F_state) is the primary factor determining overall energy expenditure (E_level), and increasing it significantly increases E_level. Hand energy expenditure (H_state) also enhances E_level; when H_state increases, E_level also increases accordingly, but F_state remains the dominant factor.

[0141] When foot energy consumption is low, overall energy consumption remains low regardless of changes in hand energy consumption, indicating that hand movement alone has a limited impact on overall energy consumption. When foot energy consumption is high, the impact of hand energy consumption is more pronounced, and energy consumption levels increase more rapidly, indicating that at high F_states, H_states further amplify E_levels.

[0142] It is also necessary to define the fatigue value increment per minute. Under normal circumstances, the driver will enter a state of severe fatigue in about three hours, and repeated operation will shorten this time. Therefore, in this embodiment, the fatigue value increment per minute is set according to Table 2.

[0143] Table 2

[0144]

[0145] Once the setup is complete, you can proceed to fatigue generation.

[0146] S3. Obtain the driver's energy consumption data using the fatigue accumulation model, and obtain the driver's fatigue state based on the fatigue accumulation model.

[0147] In this embodiment, the fatigue value is accumulated and calculated as follows:

[0148]

[0149] Where ΔF represents the fatigue increment per minute, ω i represents fatigue weight, R i Indicates the fatigue increment; fatigue value attenuation calculation:

[0150] F(t+1)=F(t)·e -λt

[0151] Where F(t+1) represents the fatigue value decay per hour, and λ represents the decay coefficient.

[0152] According to the calculated fatigue value, the fatigue level is divided into: Level 1 (awake): F(t) < 30% (steering wheel fine adjustment range < 5, brake response time < 0.8s); Level 2 (mild fatigue) 30% <F(t)<70%(方向盘调整幅度)5,刹车响应时间> Level 3 (dangerous fatigue): F(t) > 70% (> 3 consecutive instances of driving on the line). Thresholds are scaled based on individual driver differences (age, gender): For older drivers, the thresholds for each level are lowered by 20%.

[0153] Real-time status output: The system outputs the corresponding fatigue level based on F(t). Based on the fatigue level, the system adopts the preset corresponding strategy. For example, when F(t)>60% is detected, the orange warning icon on the instrument panel will flash. When F(t)>80% and lasts for five minutes, the seat vibration will be activated and the driver will be navigated to the nearest service area.

[0154] Example 2

[0155] In this embodiment, a driver fatigue state estimation system based on vehicle motion includes: an energy consumption model construction module, a fatigue accumulation model construction module and a fatigue state estimation module.

[0156] The energy consumption model building module is used to obtain the historical data of energy consumption caused by different operating actions of the driver during the driving task and build an energy consumption model.

[0157] The fatigue accumulation model construction module is based on the energy consumption model and GMM model, combined with the fuzzy control method, to construct and train the fatigue accumulation model based on fuzzy control.

[0158] The fatigue state estimation module is used to obtain the driver's energy consumption data using a fatigue accumulation model and obtain the driver's fatigue state based on the fatigue accumulation model.

[0159] The embodiments described above are merely descriptions of preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by persons skilled in the art should fall within the scope of protection defined by the claims of the present invention.

Claims

1. A method for estimating driver fatigue state based on vehicle motion, characterized in that: The following steps are involved: Obtain historical data on the energy consumption caused by different operating actions of the driver during the driving task and build an energy consumption model; Based on the energy consumption model and the GMM model, combined with the fuzzy control method, a fatigue accumulation model based on fuzzy control is constructed and trained; The energy consumption data of the driver using the fatigue accumulation model is obtained, and the fatigue state of the driver is obtained based on the fatigue accumulation model.

2. The method for estimating driver fatigue state based on vehicle motion according to claim 1, characterized in that: The energy consumption history data includes: upper limb energy consumption history data and lower limb energy consumption history data; the energy consumption model includes: upper limb movement energy consumption model and lower limb movement energy consumption model.

3. The method for estimating driver fatigue state based on vehicle motion according to claim 2, characterized in that: The method for constructing the upper limb exercise energy consumption model includes: Get the steering torque and steering wheel angle rate of the steering wheel operation: Among them, τ steer represents steering torque, m represents vehicle mass, r represents kingpin offset, G represents steering gear ratio, represents the steering wheel angular rate, L represents the wheelbase, v(t) represents the vehicle speed, t represents the time, R(t) represents the turning radius, and δ(t) represents the wheel angle; Based on the steering torque and the steering wheel angular rate, the upper limb movement energy consumption model is constructed: Among them, a lat (τ) represents the lateral acceleration, v(τ) represents the absolute velocity, t0 represents the initial time, C hand Represents the proportional coefficient of the upper limb movement energy consumption model.

4. The method for estimating driver fatigue state based on vehicle motion according to claim 3, characterized in that: The method for constructing the lower limb exercise energy consumption model includes: Get the actual pedal force: Among them, F foot (t) represents the actual pedal force, a(t) represents the longitudinal acceleration, F max Indicates the maximum pedal force, a max Indicates the maximum acceleration; Get the actual pedal stroke: Among them, d foot (t) represents the actual pedal stroke, d max Indicates the maximum pedal travel; Based on the actual pedal force and the actual pedal stroke, the lower limb exercise energy consumption model is constructed: Where a(τ) represents the longitudinal acceleration, C foot Represents the proportional coefficient of the lower limb movement energy consumption model.

5. The method for estimating driver fatigue state based on vehicle motion according to claim 4, characterized in that: The method for constructing the fatigue accumulation model includes: Based on the energy consumption model and in combination with the fuzzy control method, an initial fatigue accumulation model based on fuzzy control is constructed; Constructing the GMM model and training the GMM model using an expectation-maximization algorithm to obtain a trained GMM model; A membership function is constructed based on the trained GMM model, and a rule base of the initial fatigue accumulation model is established based on the membership function to obtain the fatigue accumulation model.

6. The method for estimating driver fatigue state based on vehicle motion according to claim 5, characterized in that: The membership functions include: upper limb high energy consumption membership function and lower limb low energy consumption membership function; The upper limb high energy consumption membership function is: Among them, μ high (E hand ) represents the upper limb high energy consumption membership function, μ3 represents the high energy consumption level, and k represents the proportional coefficient; The lower limb low energy consumption membership function is: Among them, μ low (E foot ) represents the low energy consumption membership function of the lower limbs, μ1 represents the low energy consumption level, and σ1 represents the standard deviation of the low energy consumption level.

7. A driver fatigue state estimation system based on vehicle motion, the system applying the method according to any one of claims 1 to 6, characterized in that: include: Energy consumption model building module, fatigue accumulation model building module and fatigue state estimation module; The energy consumption model building module is used to obtain historical data on energy consumption caused by different operating actions of the driver during the driving task and build an energy consumption model; The fatigue accumulation model construction module is based on the energy consumption model and the GMM model, combined with the fuzzy control method, to construct and train a fatigue accumulation model based on fuzzy control; The fatigue state estimation module is used to obtain the driver's energy consumption data using the fatigue accumulation model, and obtain the driver's fatigue state based on the fatigue accumulation model.

Images